Artificial Intelligence in Financial

Markets: Cutting Edge Applications for

Risk Management, Portfolio
Optimization and Economics 1st
Edition Christian L. Dunis
Visit to download the full and correct content document:
https://textbookfull.com/product/artificial-intelligence-in-financial-markets-cutting-edge
-applications-for-risk-management-portfolio-optimization-and-economics-1st-edition-c
hristian-l-dunis/
More products digital (pdf, epub, mobi) instant
download maybe you interests ...

Computational Intelligence Techniques for Trading and

Investment 1st Edition Christian Dunis

https://textbookfull.com/product/computational-intelligence-
techniques-for-trading-and-investment-1st-edition-christian-
dunis/

Bayesian risk management a guide to model risk and

sequential learning in financial markets 1st Edition
Sekerke

https://textbookfull.com/product/bayesian-risk-management-a-
guide-to-model-risk-and-sequential-learning-in-financial-
markets-1st-edition-sekerke/

Financial Risk Management Applications in Market Credit

Asset and Liability Management and Firmwide Risk 1st
Edition Jimmy Skoglund

https://textbookfull.com/product/financial-risk-management-
applications-in-market-credit-asset-and-liability-management-and-
firmwide-risk-1st-edition-jimmy-skoglund/

Artificial Intelligence and Industrial Applications:

Artificial Intelligence Techniques for Cyber-Physical,
Digital Twin Systems and Engineering Applications
Tawfik Masrour
https://textbookfull.com/product/artificial-intelligence-and-
industrial-applications-artificial-intelligence-techniques-for-
cyber-physical-digital-twin-systems-and-engineering-applications-
Artificial Intelligence for Marketing Practical
Applications 1st Edition Jim Sterne

https://textbookfull.com/product/artificial-intelligence-for-
marketing-practical-applications-1st-edition-jim-sterne/

Web Semantics: Cutting Edge and Future Directions in

Healthcare 1st Edition Sarika Jain

https://textbookfull.com/product/web-semantics-cutting-edge-and-
future-directions-in-healthcare-1st-edition-sarika-jain/

Cutting-Edge Technology for Carbon Capture,

Utilization, and Storage 1st Edition Karine Ballerat-
Busserolles

https://textbookfull.com/product/cutting-edge-technology-for-
carbon-capture-utilization-and-storage-1st-edition-karine-
ballerat-busserolles/

Artificial Intelligence and Blockchain for Future

Cybersecurity Applications 1st Edition Yassine Maleh

https://textbookfull.com/product/artificial-intelligence-and-
blockchain-for-future-cybersecurity-applications-1st-edition-
yassine-maleh/

Winning with Risk Management Financial Engineering and

Risk Management Volume 2 1st Edition Russell Walker

https://textbookfull.com/product/winning-with-risk-management-
financial-engineering-and-risk-management-volume-2-1st-edition-
russell-walker/
 new developments in
quantitative trading
and investment

CHRISTIAN L . DUNIS
PETER W . MIDDLETON
ANDREAS KARATHANASOPOULOS
KONSTANTINOS THEOFILATOS

ARTIFICIAL
INTELLIGENCE
IN FINANCIAL
MARKETS
Cutting-Edge Applications
for Risk Management, Portfolio
Optimization and Economics
New Developments in Quantitative Trading and
Investment
Christian L. Dunis • Peter W. Middleton • Konstantinos Theofilatos
Andreas Karathanasopoulos
Editors

Artificial Intelligence
in Financial Markets
Cutting-Edge Applications for Risk Management,
Portfolio Optimization and Economics
Editors
Christian L. Dunis Peter W. Middleton
ACANTO Holding University of Liverpool
Hannover, Germany Liverpool, England

Konstantinos Theofilatos Andreas Karathanasopoulos

University of Patras American University of Beirut (AUB)
Patras, Greece Beirut, Lebanon

ISBN 978-1-137-48879-4 ISBN 978-1-137-48880-0 (eBook)

DOI 10.1057/978-1-137-48880-0

Library of Congress Control Number: 2016941760

© The Editor(s) (if applicable) and The Author(s) 2016

The author(s) has/have asserted their right(s) to be identified as the author(s) of this work in accordance with
the Copyright, Designs and Patents Act 1988.
This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether the
whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations,
recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or
information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar
methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does
not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective
laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this book are
believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors
give a warranty, express or implied, with respect to the material contained herein or for any errors or omissions
that may have been made.

Cover illustration: © Ioana Martalogu / Alamy

Printed on acid-free paper

This Palgrave Macmillan imprint is published by Springer Nature

The registered company is Macmillan Publishers Ltd. London
 Preface

The aim of this book is to focus on Artificial Intelligence (AI) and to provide
broad examples of its application to the field of finance. Due to the popu-
larity and rapid emergence of AI in the area of finance this book is the first
volume in a series called ‘New Developments in Quantitative Trading and
Investment’ to be published by Palgrave Macmillan. Moreover, this particular
volume targets a wide audience including both academic and professional
financial analysts. The content of this textbook targets a wide audience who
are interested in forecasting, modelling, trading, risk management, econom-
ics, credit risk and portfolio management. We offer a mixture of empirical
applications to different fields of finance and expect this book to be beneficial
to both academics and practitioners who are looking to apply the most up to
date and novel AI techniques. The objective of this text is to offer a wide vari-
ety of applications to different markets and assets classes. Furthermore, from
an extensive literature review it is apparent that there are no recent textbooks
that apply AI to different areas of finance or to a wide range of markets and
products.
Each Part is comprised of specialist contributions from experts in the field
of AI. Contributions offer the reader original and unpublished content that
is recent and original. Furthermore, as the cohort of authors includes various
international lecturers and professors we have no doubt that the research will
add value to many MA, MSc, and MBA graduate programmes. Furthermore,
for the professional financial forecaster this book is without parallel a compre-
hensive, practical and up-to-date insight into AI. Excerpts of programming
code are also provided throughout in order to give readers the opportunity to
apply these techniques on their own.

v
vi Preface

Authors of this book extend beyond the existing literature in at least three
ways. The first contribution is that we have included empirical applications
of AI in four different areas of finance: time-series modelling, economics,
credit and portfolio management. Secondly, the techniques and methodolo-
gies applied here are extremely broad and cover all areas of AI. Thirdly, each
chapter investigates different datasets from a variety of markets and asset
classes. Different frequencies of data are also investigated to include daily,
monthly, macroeconomic variables and even text data from different sources.
We believe that the Parts presented here are extremely informative and practi-
cal while also challenging existing traditional models and techniques many
of which are still used today in financial institutional and even in other areas
of business. The latter is extremely important to highlight since all of the
applications here clearly identify a benefit of utilizing AI to model time-series,
enhance decision making at a government level, assess credit ratings, stock
selection and portfolio optimization.

Contents

Part I
Following the introduction, the first part focuses on numerous time-series,
which will include commodity spreads, equities, and exchange traded funds.
For this part the objective is to focus on the application of AI methodologies
to model, forecast and trade a wide range of financial instruments. AI method-
ologies include, Artificial Neural Networks (ANN), Heuristic Optimization
Algorithms and hybrid techniques. All of the submissions provide recent
developments in the area of financial time-series analysis for forecasting and
trading. A review of publications reveals that existing methodologies are either
dated or are limited in scope as they only focus on one particular asset class at
a time. It is found that the majority of the literature focuses on forecasting for-
eign exchange and equities. For instance, Wang et al. [14] focus their research
and analysis on forecasting the Shanghai Composite index using a Wavelet-
Denoising-based back propagation Neural Network (NN). The performance
of this NN is benchmarked against a traditional back propagation NN. Other
research is now considered redundant as the field of AI is evolving at a rapid
rate. For instance, Zirilli [19] offers a practical application of neural networks
to the prediction of financial markets however, the techniques that were used
are no longer effective when predicting financial variables. Furthermore, data
 Preface vii

has become more readily available so input datasets can now be enriched to
enable methodologies to capture the relationships between input datasets and
target variables more accurately. As a result, more recent research and techno-
logical innovations have rendered such methodologies obsolete.
While numerous journal publications apply AI to various assets our search
did not uncover recent textbooks that focus on AI and in particular empirical
applications to financial instruments and markets. For this reason we believe
that an entire section dedicated to time-series modelling, forecasting and trad-
ing is justified.

Part II
The second part focuses on economics as a wider subject that encompasses the
prediction of economic variables and behavioural economics. Both macro-
and micro-economic analysis is provided here. The aim of this part is to pro-
vide a strong case for the application of AI in the area of economic modelling
and as a methodology to enhance decision making in corporations and also
at a government level. Various existing work focuses on agent-based simu-
lations such as Leitner and Wall [16] who investigate economic and social
systems using agent-based simulations. Teglio et al. [17] also focus on social
and economic modelling relying on computer simulations in order to model
and study the complexity of economic and social phenomena. Another recent
publication by Osinga et al. [13] also utilizes agent-based modelling to cap-
ture the complex relationship between economic variables. Although this part
only provides one empirical application we believe that it goes a long way to
proving the benefits of AI and in particular ‘Business Intelligence’.
With extensive research being carried out in the area of economic model-
ling it is clear that a whole section should also be devoted to this particular
area. In fact we expect this section to draw a lot of attention given its recent
popularity.

Part III
The third part focuses on analyzing credit and the modelling of corporate struc-
tures. This offers the reader an insight into AI for evaluating fundamental data
and financial statements when making investment decisions. From a prelimi-
nary search our results do not uncover any existing textbooks that exclusively
focus on credit analysis and corporate finance analyzed by AI methodologies.
However, the search uncovered a few journal publications that provide an
insight into credit analysis in the area of bankruptcy prediction. For instance,
Loukeris and Matsatsinis [9] research corporate finance by attempting to pre-
viii Preface

dict bankruptcy using AI models. From results produced by these journal

publications we believe that corporate finance could benefit from more recent
empirical results published in this part.
Earlier research in the area of credit analysis is carried out by Altman et al.
[1] who examine the use of layer networks and how their use has led to an
improvement in the reclassifying rate for existing bankruptcy forecasting
models. In this case, it was found that AI helped to identify a relationship
between capital structure and corporate performance.
The most recent literature reviewed in the area of corporate finance applies
AI methodologies to various credit case studies. We suspect that this was
inspired by the recent global credit crisis in 2008 as is the case with most
credit-based research published after the 2008 ‘credit crunch’. For instance,
Hajek [6] models municipal credit ratings using NN classification and genetic
programs to determine his input dataset. In particular, his model is designed
to classify US municipalities (located in the State of Connecticut) into rating
classes based on their levels of risk. The model includes data pre-processing, the
selection process of input variables and the design of various neural networks'
structures for classification. Each of the explanatory variables is extracted
from financial statements and statistical reports. These variables represent the
inputs of NNs, while the rating classes from Moody’s rating agency are the
outputs. Experimental results reveal that the rating classes assigned by the NN
classification to bond issuers are highly accurate even when a limited sub-set
of input variables is used. Further research carried out by Hajek [7] presents
an analysis of credit rating using fuzzy rule-based systems. A fuzzy rule-based
system adapted by a feed-forward neural network is designed to classify US
companies (divided into finance, manufacturing, mining, retail trade, ser-
vices, and transportation industries) and municipalities into the credit rating
classes obtained from rating agencies. A genetic algorithm is used again as a
search method and a filter rule is also applied. Empirical results corroborate
much of the existing research with the classification of credit ratings assigned
to bond issuers being highly accurate. The comparison of selected fuzzy rule-
based classifiers indicates that it is possible to increase classification perfor-
mance by using different classifiers for individual industries.
León-Soriano and Muñoz-Torres [8] use three layers feed-forward neural
networks to model two of the main agencies’ sovereign credit ratings. Their
results are found to be highly accurate even when using a reduced set of pub-
licly available economic data. In a more thorough application Zhong et al.
[20] model corporate credit ratings analyzing the effectiveness of four different
learning algorithms. Namely, back propagation, extreme learning machines,
incremental extreme learning machines and support vector machines over
 Preface ix

a data set consisting of real financial data for corporate credit ratings. The
results reveal that the SVM is more accurate than its peers.
With extensive research being carried out in the area of bankruptcy predic-
tion and corporate/sovereign credit ratings it is clear that the reader would
benefit from a whole section being devoted to credit and corporate finance.
In fact the first chapter provides an interesting application of AI to discover
which areas of credit are most popular. AI is emerging in the research of credit
analysis and corporate finance to challenge existing methodologies that were
found to be inadequate and were ultimately unable to limit the damage caused
by the 2008 ‘credit crisis’.

Part IV
The final section of the book focuses on portfolio theory by providing exam-
ples of security selection, portfolio construction and the optimization of
asset allocation. This will be of great interest to portfolio managers as they
seek optimal returns from their portfolios of assets. Portfolio optimization
and security selection is a heavily researched area in terms of AI applications.
However, our search uncovered only a few existing journal publications and
textbooks that focus on this particular area of finance. Furthermore, research
in this area is quickly made redundant as AI methodologies are constantly
being updated and improved.
Existing journal publications challenge the Markowitz two-objective
mean-variance approach to portfolio design. For instance, Subbu et al. [15]
introduce a powerful hybrid multi-objective optimization approach that
combines evolutionary computation with linear programming to simultane-
ously maximize return, minimize risk and identify the efficient frontier of
portfolios that satisfy all constraints. They conclude that their Pareto Sorting
Evolutionary Algorithm (PSEA) is able to robustly identify the Pareto front
of optimal portfolios defined over a space of returns and risks. Furthermore
they believe that this algorithm is more efficient than the 2-dimensional and
widely accepted Markowitz approach.
An older textbook, which was co-authored by Trippi and Lee (1995),
focuses on asset allocation, timing decisions, pattern recognition and risk
assessment. They examine the Markowitz theory of portfolio optimization
and adapt it by incorporating it into a knowledge-based system. Overall this
is an interesting text however it is now almost 20 years old and updated appli-
cations/methodologies could be of great benefit to portfolio managers and
institutional investors.
x Preface

The Editors
All four editors offer a mixture of academic and professional experience in
the area of AI. The leading editor, Professor Christian Dunis has a wealth of
experience spanning more than 35 years and 75 publications, both in aca-
demia and quantitative investments. Professor Dunis has the highest expertise
in modelling and analyzing financial markets and in particular an extensive
experience with neural networks as well as advanced statistical analyses. Dr
Peter Middleton has recently completed his PhD in Financial Modelling and
Trading of Commodity Spreads at the University of Liverpool. To date he has
produced five publications and he is also a member of the CFA institute and
is working towards the CFA designation having already passed Level I. He
is also working in the finance industry in the area of Asset Management. Dr
Konstantinos possesses an expertise in technical and computational aspects
with backgrounds in evolutionary programming, neural networks, as well as
expert systems and AI. He has published numerous articles in the area of com-
puter science as well being an editor for Computational Intelligence for Trading
and Investment. Dr Andreas Karathanasopoulos is currently an Associate
Professor at the American University of Beirut and has worked in academia
for six years. He too has numerous publications in international journals for
his contribution to the area of financial forecasting using neural networks,
support vector machines and genetic programming. More recently he has also
been an editor for Computational Intelligence for Trading and Investment.

Acknowledgements
We would like to thank the authors of who have contributed original and
novel research to this book, the editors who were instrumental in its prepara-
tion and finally the publishers who have ultimately helped provide a showcase
for it to be read by the public.

Final Words
We hope that the publication of this book will enhance the spread of AI
throughout the world of finance. The models and methods developed here
have yet to reach their largest possible audience, partly because the results
are scattered in various journals and proceedings volumes. We hope that this
 Preface xi

book will help a new generation of quantitative analysts and researchers to

solve complicated problems with greater understanding and accuracy.

France Christian L. Dunis

UK Peter Middleton
Greece Konstantinos Theofilatos
Lebanon Andreas Karathanasopoulos

References
1. E.I. Altman, G. Marco, F. Varetto, Corporate distress diagnosis: Comparisons
using linear discriminant analysis and neural networks (the Italian experience),
Journal of Banking and Finance 18 (1994) 505±529.
2. Hájek, P. (2011). Municipal credit rating modelling by neural networks. Decision
Support Systems, 51(1), 108–118.
3. Hajek, P. (2012). Credit rating analysis using adaptive fuzzy rule-based systems:
An industry-specific approach. Central European Journal of Operations Research,
20(3), 421–434.
4. León-Soriano, R. and Muñoz-Torres, M. J. (2012). Using neural networks to
model sovereign credit ratings: Application to the European Union. Modeling
and Simulation in Engineering, Economics and Management: Lecture Notes in:
Business Information Processing, 115, 13–23.
5. Loukeris, N. and Matsatsinis, N. (2006). Corporate Financial Evaluation and
Bankruptcy Prediction Implementing Artificial Intelligence Methods. Proceedings of
the 10th WSEAS International Conference on COMPUTERS, Vouliagmeni,
Athens, Greece, July 13–15, 2006. Pp. 884–888.
6. Osinga, E. C., Leeflang, P. S. H., Srinivasan, S., & Wieringa, J. E. (2011). Why
do firms invest in consumer advertising with limited sales response? A share-
holder perspective. Journal of Marketing, 75(1), 109−124.
7. QIAO Yu-kun,WANG Shi-cheng,ZHANG Jin-sheng,ZHANG Qi,SUN Yuan
(Department of Automatic Control,The Second Artillery Engineering
College,Xi’an 710025,Shaanxi,China);Simulation Research on Geomagnetic
Matching Navigation Based on Soft-threshold Wavelet Denoising Method[J];Acta
Armamentarii;2011-09.
8. Subbu, R., Bonissone, P. P., Eklund, N., Bollapragada, S., and Chalermkraivuth,
K. (2005). Multiobjective Financial Portfolio Design: A Hybrid Evolutionary
Approach. In 2005 IEEE Congress on Evolutionary Computation (CEC’2005), vol.
2. Edinburgh, Scotland: IEEE Service Center, September 2005, pp. 1722–1729.
xii Preface

9. Leitner S.and F. Wall. Multi objective decision-making policies and coordination

mechanisms in hierarchical organizations: Results of an agent-based simulation.
Working Paper, Alpen-Adria Universit¨at Klagenfurt (in submission), 2013.
10. Teglio, A., Raberto, M., Cincotti, S., 2012. The impact of banks’ capital adequacy
regulation on the economic system: an agent-based approach. Advances in
Complex Systems 15 (2), 1250040–1 – 1250040–27.
11. Zirilli, J. S., 1997: Financial Prediction Using Neural Networks. International
Thomson, 135 pp.
12. Zhong, H., Miao, C., Shen, Z., and Feng, Y. (2012). Comparing the learning
effectiveness of BP, ELM, I-ELM, and SVM for corporate credit ratings.
Neurocomputing, 128, 285–295.
 Contents

Part I Introduction to Artificial Intelligence 1

1 A Review of Artificially Intelligent Applications

in the Financial Domain 3
Swapnaja Gadre Patwardhan, Vivek V. Katdare,
and Manish R. Joshi

Part II Financial Forecasting and Trading 45

2 Trading the FTSE100 Index: ‘Adaptive’ Modelling

and Optimization Techniques 47
Peter W. Middleton, Konstantinos Theofilatos,
and Andreas Karathanasopoulos

3 Modelling, Forecasting and Trading the Crack: A Sliding

Window Approach to Training Neural Networks 69
Christian L. Dunis, Peter W. Middleton,
Konstantinos Theofilatos, and Andreas Karathanasopoulos

4 GEPTrader: A New Standalone Tool for Constructing Trading

Strategies with Gene Expression Programming 107
Andreas Karathanasopoulos, Peter W. Middleton,
Konstantinos Theofilatos, and Efstratios Georgopoulos

xiii
xiv Contents

Part III Economics 123

5 Business Intelligence for Decision Making in Economics 125

Bodislav Dumitru-Alexandru

Part IV Credit Risk and Analysis 159

6 An Automated Literature Analysis on Data Mining

Applications to Credit Risk Assessment 161
Sérgio Moro, Paulo Cortez, and Paulo Rita

7 Intelligent Credit Risk Decision Support: Architecture

and Implementations 179
Paulius Danenas and Gintautas Garsva

8 Artificial Intelligence for Islamic Sukuk Rating

Predictions 211
Tika Arundina, Mira Kartiwi, and Mohd. Azmi Omar

Part V Portfolio Management, Analysis and Optimisation 243

9 Portfolio Selection as a Multi-period Choice Problem

Under Uncertainty: An Interaction-Based Approach 245
Matjaz Steinbacher

10 Handling Model Risk in Portfolio Selection

Using Multi-Objective Genetic Algorithm 285
Prisadarng Skolpadungket, Keshav Dahal,
and Napat Harnpornchai

11 Linear Regression Versus Fuzzy Linear Regression:

Does it Make a Difference in the Evaluation
of the Performance of Mutual Fund Managers? 311
Konstantina N. Pendaraki and Konstantinos P. Tsagarakis

Index 337
 Notes on Contributors

Christian L. Dunis is a Founding Partner of Acanto Research (www.acantore-

search.com) where he is responsible for global risk and new products. He is
also Emeritus Professor of Banking and Finance at Liverpool John Moores
University where he directed the Centre for International Banking, Economics
and Finance (CIBEF) from February 1999 to August 2011.
Christian Dunis holds an MSc, a Superior Studies Diploma in International
Economics and a PhD in Economics from the University of Paris.
Peter W. Middleton completed a Phd at the University of Liverpool. His
working experience is in Asset Management and he has published numerous
articles on financial forecasting of commodity spreads and equity time-series.
Andreas Karathanasopoulos studied for his MSc and Phd at Liverpool John
Moores University under the supervision of Professor Christian Dunis. His
working experience is academic having taught at Ulster University, London
Metropolitan University and the University of East London. He is currently
an Associate Professor at the American University of Beirut and has published
more than 30 articles and one book in the area of artificial intelligence.
Konstantinos Theofilatos completed his MSc and Phd in the University of
Patras Greece. His research interests include computational intelligence,
financial time-series forecasting and trading, bioinformatics, data mining and
web technologies. He has so far published 27 publications in scientific peer
reviewed journals and he has also published more than 30 articles in confer-
ence proceedings.

xv
 Part I
Introduction to Artificial Intelligence
 1
A Review of Artificially Intelligent
Applications in the Financial Domain
Swapnaja Gadre-Patwardhan, Vivek V. Katdare,
and Manish R. Joshi

1 Introduction
Undoubtedly, the toughest challenge faced by many researchers and managers
in the field of finance is uncertainty. Consequently, such uncertainty intro-
duces an unavoidable risk factor that is an integral part of financial theory.
The manifestation of risk not only complicates financial decision making but
also creates profitable opportunities for investors who can manage and analyze
risk efficiently and effectively. In order to handle the complex nature of the
problem an interdisciplinary approach is advocated.
Computational finance is a division of applied computer science that deals
with practical problems in finance. It can also be defined as the study of data
and algorithms used in finance. This is an interdisciplinary field that combines

S. Gadre-Patwardhan (*)
Institute of Management and Career Courses, Pune, India
e-mail: swapnaja.gadre24@gmail.com
V.V. Katdare
I.M.R. College, Jalgaon, India
e-mail: vvkatdare@rediffmail.com
M.R. Joshi
School of Computer Sciences, North Maharashtra University, Jalgaon, India
e-mail: joshmanish@gmail.com

© The Editor(s) (if applicable) and The Author(s) 2016 3

C.L. Dunis et al. (eds.), Artificial Intelligence in Financial Markets,
New Developments in Quantitative Trading and Investment,
DOI 10.1057/978-1-137-48880-0_1
4 S. Gadre-Patwardhan et al.

Discriminant
Parametric Analysis
Sta s cal Methods
Logis c Regression
Financial Analysis Methods
Decision Tree
Non-Parametric
Sta s cal Methods
Nearest Neighbor

Ar ficial Neural Network

Fuzzy Logic
SoCompu ng
Support Vector
Machine

Gene c Algorithm

Fig. 1.1 Techniques for analysis of financial applications

numerical methods and mathematical finance. Computational finance uses

mathematical proofs that can be applied to economic analyses thus aiding the
development of finance models and systems. These models are employed in
portfolio management, stock prediction and risk management and play an
important role in finance management.
During past few years, researchers have aimed to assist the financial sector
through trend prediction, identifying investor behaviour, portfolio manage-
ment, fraud detection, risk management, bankruptcy, stock prediction, finan-
cial goal evaluation, finding regularities in security price movement and so
forth. To achieve this, different methods like parametric statistical methods,
non-parametric statistical methods and soft computing methods have been
used as shown in Fig. 1.1. It is observed that many researchers are exploring
and comparing soft computing techniques with parametric statistical tech-
niques and non-parametric statistical techniques. Soft computing techniques,
such as, Artificial Neural Network (ANN), Fuzzy Logic, Support Vector
Machine (SVM), Genetic Algorithm, are widely applied and accepted tech-
niques in the field of finance and hence are considered in this review.
(A) Parametric statistical methods: Parametric statistics is a division of sta-
tistics. It assumes that data is collected from various distributed systems and
 1 A Review of Artificially Intelligent Applications to Finance 5

integrated in order to draw inferences about the parameters of the distribu-

tion. There are two types of parametric statistical methods namely discrimi-
nant analysis and logistic regression:
(I) Discriminant analysis: Discriminant analysis is a statistical analysis car-
ried out with the help of a discriminant function to assign data to one of two or
more naturally occurring groups. Discriminant analysis is used to determine
the set of variables for the prediction of category membership. Discriminant
function analysis is a type of classification that distributes items of data into
classes or groups or categories of the same type.
(II) Logistic regression: Logistic regression is a method of prediction that
models the relationship between dependent and independent variables. It the
best-fit model to be found and also identifies the significance of relationships
between dependent and independent variables. Logistic regression is used to
estimate the probability of the occurrence of an event.
(B) Non-parametric statistical methods: These are the methods in which
data is not required to fit a normal distribution. The non-parametric method
provides a series of alternative statistical methods that require no, or limited,
assumptions to be made about the data. The techniques of non-parametric
statistical methods follow.
(I) Decision tree: A decision tree is a classifier that is a tree-like graph that
supports the decision making process. It is a tool that is employed in mul-
tiple variable analyses. A decision tree consists of nodes that a branching-tree
shape. All the nodes have only one input. Terminal nodes are referred to as
leaves. A node with an outgoing edge is termed a test node or an internal
node. In a decision tree, a test node splits the instance space into two or more
sub-spaces according to the discrete function.
(II) Nearest neighbour: The nearest neighbour algorithm is a non-­parametric
method applied for regression and classification. Nearest neighbour can also
be referred as a similar search, proximity search or closest-point search, which
is used to find the nearest or closest points in the feature space. The K-nearest
neighbour algorithm is a technique used for classification and regression.
(C) Soft computing: Soft computing is a set of methods that aims to handle
uncertainty, partial truth, imprecision and approximation that are fundamen-
tally are based on human neurology. Soft computing employs techniques like:
ANN, fuzzy logic, SVM, genetic algorithm [1].
(I) Artificial neural network: A neuron is a fundamental element of
ANN. These neurons are connected to form a graph-like structure, which are
also referred to as networks. These neurons are like biological neurons. A neu-
ron has small branches, that is, dendrites, which are used for receiving inputs.
Axons carry the output and connect to another neuron. Every neuron carries
a signal received from dendrites as shown in Fig. 1.2 [2]. When the strength
6 S. Gadre-Patwardhan et al.

Fig. 1.2 Structure of Artificial Neurons

of a signal exceeds a particular threshold value, an impulse is generated as an

output, this is known as the action signal.
Like biological neurons, artificial neurons accept input and generate out-
put but are not able to model automatically. In ANN information or data is
distributed and stored throughout the network in the form of weighted inter-
connections. Simulation of a neuron is carried out with the help of non-linear
function. Interconnections of artificial neurons are referred as weights. The
diagram below shows the structure of an artificial neuron in which xi is the
input to the neuron and wi is the weight of the neuron. The average input is
calculated by the formula [2].

n
a = ∑xiwi (1.1)
i =0

ANN has a minimum of three layers of artificial neurons: input, hidden

and output as shown in Fig. 1.3 [3]. The input layer accepts the input and
passes it to the hidden layer. The hidden layer is the most important layer
from a computational point of view. All the complex functions reside in this
layer.
(II) Fuzzy logic: Fuzzy logic is a type of many values logic that deals with
approximate values instead of exact or fixed reasoning. Fuzzy logic is a method
of computing based on the degree of truth rather than a crisp true or false
value. Its truth value ranges in between 0 and 1.
(III) Support vector machine: SVM is a supervised learning model with
related learning algorithms that is used for data analysis and pattern recogni-
tion in classification and regression. SVM uses the concept of a hyperplane,
 1 A Review of Artificially Intelligent Applications to Finance 7

Fig. 1.3 Three layer architecture of ANN

which defines the boundaries of a decision. The decision plane separates the
objects based on class membership and is able to handle categorical and con-
tinuous variables.
(IV) Genetic algorithm: A genetic algorithm is an artificial intelligence
technique that mimics a natural selection process. This technique is mostly
used for optimization and search problems using selection, crossover, muta-
tion and inheritance operations.
This chapter emphasizes the application of soft computing techniques
namely artificial neural network, expert system (ES) and hybrid intelligence
system (HIS) in finance management.
In recent years, it has been observed that an array of computer technologies
is being used in the field of finance; ANN is one of these. From the array of
available AI techniques, financial uncertainties are handled in a more efficient
manner by ANN. These uncertainties are handled by pattern recognition and
future trend analysis. The most difficult aspects to incorporate in finance anal-
ysis are changes in the interest rates and currency movements. Large ‘noisy’
data can be handled well by ANN. ANN are characterized as numeric in
nature. In statistical techniques, like discriminant analysis or regression analy-
sis, data distribution assumptions are required for input data. However, ANN
does not require any data distribution assumptions and hence could be appli-
cable to a wider range of problems than other statistical techniques. Statistical
techniques and symbolic manipulation techniques are batch oriented; old and
new data are submitted in a single batch to the model and later new mining
results are generated. In contrast, in ANN it is possible to add new data to a
trained ANN so as to update the existing result. Since financial markets are
8 S. Gadre-Patwardhan et al.

dynamic in nature, ANN can accommodate new data without reprocessing

old data and hence it is used in finance management [4].
An ES is knowledge-based system used to solve critical problems in a partic-
ular domain. These are rule-based systems with predefined sets of knowledge
used for decision making. Generic ES contain two modules—the inference
engine and the knowledge base. The inference engine combines and processes
the facts associated with the specific problem using the chunk of the knowl-
edge base relevant to it. The knowledge base is coded in the form of rules,
semantic nets, predicates and objects in the system. ES are characterized as
efficient, permanent, consistent, timely, complete decision-making systems
and hence their use in finance management. ES are characterized as intelli-
gent, capable of reasoning, able to draw conclusions from relationships, capa-
ble of dealing with uncertainties and so forth. ES are capable of reproducing
efficient, consistent and timely information so as to facilitate decision making
[5]. Furthermore Rich and Knight (1991) specified long ago that financial
analysis is an expert’s task.
HIS are software systems that combine methods and techniques of artificial
intelligence, for example, fuzzy expert systems, neuro-fuzzy systems, genetic-­
fuzzy systems. The integration of various learning techniques is combined
to overcome the limitation of an individual system. Because of its facility of
combined techniques, it can be used effectively for finance management.
With reference to the financial market, we identified portfolio manage-
ment, stock market prediction and risk management as the three most impor-
tant AI application domains. As investment is an important aspect of finance
management hence these three cases are considered. In this study, we consider
the contribution of researchers in financial domains from the past 20 years in
order to study and compare the applications of ANN, ES and HIS with tra-
ditional methods. The chapter is organized thus: the second, third and fourth
sections deal with the application of ANN, ES and HIS respectively. In the
fifth section conclusions are put forth. We enlist popularly used data min-
ing tools as set out in Appendix 1 that includes some sample coding of NN
techniques using MATLAB [6] in Finance Management. Code excerpts for
implementing typical statistical functions including regression analysis, naïve
Bayes classification, fuzzy c-means clustering extracted from different openly
available authentic sources [7] are also presented in Appendix 1.

Applications of ANN in Finance

ANN are computational tools and are used in various disciplines for model-
ling real-world complex problem [8]. ANN resemble biological neurons
acting as a source inspiration for a variety of techniques covering a vast field
 1 A Review of Artificially Intelligent Applications to Finance 9

of application [9]. In general, ANN are referred to as information processing sys-

tems that which use earning and generalization capabilities, which are adaptive
in nature. Due to their adaptive nature, ANN can provide solutions to problems
such as forecasting, decision making and information processing. In recent years,
ANN have proved to be a powerful tool for handling dynamic financial market
in terms of prediction [10], panning, forecasting [11] and decision making [12].
With reference to this various studies have been carried out in order to
classify and review the application of ANN in the finance domain [13, 14].
Mixed results have been obtained concerning the ability of ANN in finance
domain. It has been observed that financial classification like financial evalua-
tion, portfolio management, credit evaluation and prediction are significantly
improved with the application of ANN in the finance domain. We further
consider the application of ANN in the finance domain in portfolio manage-
ment, stock market prediction and risk management. The details of these
applications are presented as described previously.

Portfolio Management

The determination of the optimal allocation of assets into broad categories, for
example, mutual funds, bonds, stocks, which suits investment by financial insti-
tutions across a specific time with an acceptable risk tolerance is a crucial task.
Nowadays investors prefer diversified portfolios that contain a variety of securities.
Motiwalla et al. [15] applied ANN and regression analysis to study the
predictable variations in US stock returns and concluded that ANN models
are better than regression. Yamamoto et al. [16] designed a multi-layer Back
Propagation Neural Network (BPNN) for the prediction of the prepayment
rate of a mortgage with the help of a correlation learning algorithm. Lowe
et al. [17] developed an analogue Neural Network (NN) for the construction
of portfolio under specified constraints. They also developed a feed forward
NN for prediction of short-term equities in non-linear multi-channel time-­
series forecasting. Adedeji et al. [18] applied ANN for the analysis of risky
economic projects. For the prediction of the potential returns on investment,
an NN model could be used. On the basis of results obtained from the neural
network, financial managers could select the financial project by comparing
the results to those obtained from conventional models. The survey conducted
in this paper for portfolio management concludes that ANN performs better
in terms of accuracy. Without any time consuming and expensive simulation
experiments, accuracy can be obtained by combining conventional simula-
tion experiments with a neural network.
10 S. Gadre-Patwardhan et al.

Research papers surveyed for portfolio management demonstrates that

when compared to other traditional methods, ANN performs better particu-
larly BPNN. Zimmermann et al. [19] demonstrated the application of the
Back/Litterman portfolio optimization algorithm with the help of an error
correction NN. Optimization of the portfolio includes (1) allocation that
comply investors constraints and (2) controlled risk in the portfolio. The
method was tested with internationally diversified portfolios across 21 finan-
cial markets from G7 countries. They stated that their approach surpassed
conventional portfolio forecasts like Markowitz’s mean-variance framework.
Ellis et al. [20] performed a portfolio analysis by comparing BPNN with a
randomly selected portfolio method and a general property method conclud-
ing that ANN performs better.

Stock Market Prediction

In recent years with the help of online trading, the stock market is one of the
avenues where individual investors can earn sizeable profits. Hence there is a
need to predict stock market behaviour accurately. With this prediction inves-
tors can take decisions about where and when to invest. Because of the volatil-
ity of financial market building a forecasting model is a challenging task.
ANN are a widely used soft computing method for stock market predic-
tion and forecasting. White applied ANN on IBM daily stock returns and
concluded that the NN outperformed other methods [21]. Kimoto et al. [22]
reported the effectiveness of learning algorithms and prediction methods of
Modular Neural Networks (MNN) for the Tokyo Stock Exchange price index
prediction system. Kazuhiro et al. [23] investigated the application of prior
knowledge and neural networks for the improvement of prediction ability.
Prediction of daily stock prices was considered a real-world problem. They
considered some non-numerical features such as political and international
events, as well as a variety of prior knowledge that was difficult to incorporate
into a network structure (the prior knowledge included stock prices and infor-
mation about foreign and domestic events published in newspapers.) It was
observed that event knowledge combined with an NN was more effective for
prediction with a significance level of 5 %. Pai et al. [24] stated that ARIMA
(autoregressive integrated moving average) along with SVM can be combined
to deal with non-linear data. The unique strengths of ARIMA and SVM are
used for more reliable stock-price forecasting. Thawornwong et al. [25] dem-
onstrated that the NN model with feed-forward and probabilistic network
for the prediction of stock generated high profits with low risk. Nakayama
et al. [26] proposed a Fuzzy Neural Network (FNN) that contained a specific
 1 A Review of Artificially Intelligent Applications to Finance 11

structure for realizing a fuzzy inference system. Every membership function

consists of one or two sigmoid functions for inference rule. They concluded
that their FNN performed better. Duke et al. [27] used Back Propagation
Network (BPN) for the prediction of the performance of the German govern-
ment’s bonds

Risk Management

Financial risk management (FRM) is the process of managing economic value

in a firm with the help of financial instruments to manage risk exposure espe-
cially market risk and credit risk. Financial Risk Management (FRM) is the
process of identification of risk associated with the investments and possibly
mitigating them. FRM can be qualitative or quantitative. FRM focuses on
how and when hedging is to be done with the help of financial instruments
to manage exposure to risk.
Treacy et al. [28] stated that the traditional approach of banks for credit
risk assessment is to generate an internal rating that considers subjective as
well as qualitative factors such as earning, leverage, reputation. Zhang et al.
[29] compared Logistic Regression (LR), NN and five-fold cross validation
procedures on the database of manufacturing firms. They employed Altman’s
five functional ratios along with the ratio current assets/current liabilities as an
input to NN. They concluded that NN outperforms with accuracy 88.2 %.
Tam et al. [30] introduced an NN approach to implement discriminant anal-
ysis in business research. Using bank data, linear classification is compared
with a neural approach. Empirical results concluded that the neural model
is more promising for the evaluation of bank condition in terms of adapt-
ability, robustness and predictive accuracy. Huang et al. [31] introduced an
SVM to build a model with a better explanatory ability. They used BPNN as
a benchmark and obtained around 80 % prediction accuracy for both SVM
and BPNN for Taiwan and United States markets.
Table 1.1 provides details of the literature that considers the applica-
tion of ANN for portfolio management, stock market prediction and risk
management.

2 Application of Expert Systems in Finance

An expert system is a computer system that is composed of a well-organized
body of knowledge that emulates expert problem-solving abilities in a lim-
ited domain of expertise. Matsatsinis et al. [54] presented a methodology
12 S. Gadre-Patwardhan et al.

of acquisition of knowledge and representation of knowledge for the devel-

opment of ES for financial analysis. Development of FINEVA (FINancial
EVAluation), a multi-criterion knowledge base DSS (decision support soft-
ware) for assessment of viability and corporate performance and the applica-
tion of FINEVA was discussed. For a particular domain, a set of inference
rules are provided by a human expert. The knowledge base is a collection of
relevant facts, data, outcome and judgments [34]. Components of expert sys-
tems include the knowledge base, the user interface and the inference engine.
Knowledge is represented through the techniques such as, predicate logic,
frames and semantic nets but the most popular and widely used technique is
the IF-THEN rule also referred as the production rule.
Liao et al. [55] carried out a review of the use of an ES in a variety of areas
including finance during period 1995 to 2004. They observed that ES are
flexible and provide a powerful method for solving a variety of problems,
which can be used as and when required. Examples of the application of ES
in finance domain follow.

Portfolio Management

It is a difficult and time-consuming task to explore and analyze a portfolio in

relation to the requirements and objectives of the fund manager. Ellis et al. [34]
examined the application of rule-based ES in the property market and port-
folios randomly constructed from the market. They observed that rule-based
outperform the random portfolio or market on risk adjusted return basis.
Bohanec et al. [56] developed a knowledge-based tool for portfolio analysis
for evaluation of a project. This ES was developed for the Republic of Solvenia.
The model is demonstrated with a tree structure supplemented by IF-THEN
rules. Sanja Vraneš et al. [57] developed the Blackboard-based Expert Systems
Toolkit (BEST) for combining knowledge from different sources, using dif-
ferent methodologies for knowledge acquisition. As far as investment decision
making is concerned, information from proficient economist critical invest-
ment ranking might be combined with knowledge evolved from operational
research methods. When decisions are made based on information combined
from many sources, there is a probability of redundancy reduction and more
promising results. Varnes et al. [58] suggested INVEX (investment advisory
expert system) for investment management. This system assists investors and
project analysts to select a project for investment. Mogharreban et al. [59]
developed the PROSEL (PORtfolio SELection) system that uses a set of rules
for stock selection. PROSEL consists of three parts (1) an information centre
Table 1.1 A Brief Review of ANNs Applied to Portfolio Management, Stock Market Prediction and Risk Management
Approach Compared Evaluation
Author Objective Data Set Preprocessing used with metrics
Stoppiglia H, To develop a The database comprises Fifteen financial ANN Statistical Classification
1

Idan Y, neural network 398 companies, ratios such as, Method

Dreyfus G [32] -aided model with 172 A companies, working capital/
for portfolio 172 B companies,& 54 C fixedassets,
management companies profit after taxes
and interest/net
worth, per year
Hans Georg Portfolio Financial markets Monthly data ANN Mean-variance Forecasting
Zimmermann, Optimization of the G7 countries extracted from theory
Ralph Neuneier all databases
and Ralph
Grothmann,
Siemens AG [33]
Ellis C, Willson P To select Australian property sector Not mentioned BPNN Random Performance
[34] portfolios stocks selection measure
portfolio
Fernandez A, Portfolio Hang Seng in Hong Kong, Weekly prices ANN, GA Heuristic Portfolio
Gomez S [35] selection and DAX 100 in Germany, from data sets and SA methods selection and
portfolio FTSE 100 in UK, S&P 100 are extracted optimization
management in USA and Nikkei 225 in
Japan
Freitas FD, Portfolio selection IBOVESPA Selected a subset BPNN Mean-variance Prediction
De Souza AF, and Portfolio of 52 stocks with model
De Almeida AR optimization long enough
A Review of Artificially Intelligent Applications to Finance

[36] time-series for

training the
neural networks
(continued)
13
Table 1.1 (continued)
14

Approach Compared Evaluation

Author Objective Data Set Preprocessing used with metrics
Po-Chang Ko, Portfolio selection Taiwan stock exchange Not mentioned ANN Traditional Portfolio
Ping-Chen Lin and portfolio ANN model optimization
[37] optimization
Chiang W-C, Asset forecasting US mutual fund 15 economic BPNN Regression NAV prediction
Urban TL, variables are model
Baldridge GW identified
[38]
Chen A-S, Stock index Taiwan StockExchange Data is extracted PNN Random walk Return on
Leung MT, forecasting on the basis of model and the investment
Daouk H [39] length of parametric
S. Gadre-Patwardhan et al.

investment GMM models

horizon
O’Connor N, To predict stock New York Stock Exchange Daily opening and BPNN Simple Accuracy
Madden MG market and NASDAQ closing values of benchmark
[40] movement DJIA index functions
forecasting
De Faria EL, To predict stock Brazilian stock market Not mentioned BPNN Adaptive Accuracy
Marcelo P, market exponential
Albuquerque J, movement smoothing
Gonzalez L, forecasting method
Cavalcante JTP,
Marcio
Albuquerque P
[41]
Liao A, Wang J Stock index SP500, SAI, SBI, DJI, HIS Data BP Brownian Forecasting
[42] forecasting and IXIC normalization stochastic motion
and adjusted to time
remove the noise effective
NN
Hyun-jung Kim, To detect patterns Korea StockPrice Daily stock data ATNN TDNN Accuracy
Kyung-shik Shin in stock market Index 200 is extracted
[43]
Kuo RJ, Stock market Taiwan stock market Not mentioned ANN, Qualitative and Performance
Chen CH, forecasting GFNN quantitative evaluation
Hwang YC [44] factors of NN
Md. Rafiul Hassan, Stock market www.finance.yahoo.com Daily data is ANN, GA, ARIMA model Forecasting
Baikunth Nath, forecasting extracted HMM
Michael Kirley
[45]
Chye KH, Credit risk Australian and German For each applicant SVM Neural Accuracy
Tan WC, assessment credit data sets 24 variables are classifier networks,
Goh GP [46] selected genetic
programming,
and decision
tree classifiers
Eliana Angelini, Credit risk Bankin Italy Sample group is ANN classical feed Classification
Giacomo di Tollo, evaluation categorized into forward neural
Andrea Roli [47] two groups i.e. network and
boins and default special purpose
feed forward
architecture
Fanning, Cogger To develop Management database Not mentioned ANN Generalized Accuracy
Shrivastava [48] a model using adaptive neural
neural network network
to find architectures
managerial (GANNA) and
fraud the Adaptive
1 A Review of Artificially Intelligent Applications to Finance

Logic Network
(ALN)
(continued)
15
Another random document with
no related content on Scribd:
all. We thus have light discharges and severe discharges at different times, or, as the habit is
established, only the light or only the severe, the manifestation depending probably upon the number of
discharging cells and the importance of the exciting cause.
31 Traité de l'Épilepsie, etc., Paris, 1845, p. 55 et seq.

When the attacks occur in alarming frequency, as they sometimes do, the condition is known as the
status epilepticus. Leuret had a patient who had eighty in two hours, and Delasiauve reports the case of
a young man fifteen years old who had twenty-five hundred in one month. They may be so numerous as
to be apparently continuous. The patient remains in a state of coma (the status epilepticus), with very
high temperature. If he be not restored, he sinks into a deeper coma, and all the signs of collapse
manifest themselves. Bed-sores form, œdema of the lungs ensues, and the patient dies. Happily, this
condition of affairs is rare.

Delasiauve calls attention to the fact that the first two or three attacks that usher in this state do not
usually attract much attention, but the succeeding ones are so violent as to immediately suggest violent
consequences. In one of my cases the attacks, when they had once become numerous, were readily
excited by the least jarring, noise, or handling, just as we find in strychnine-poisoning or tetanus.

Irregular Forms.—There are occasional cases of psychical or masked epilepsy, the study of which is
intensely interesting. Such forms are characterized by perverted consciousness and a low degree of
volitional direction which may vary from automatism to the undoubted exercise of complex functions of
the mind, though badly co-ordinated. Mesnet's soldier, when subject to a paroxysm and apparently
unconscious, would perform a number of suggested acts in a rhythmical manner and with no
subsequent knowledge of the previous event: when started off by the word of command to march, he
would blindly go on, marking time when he met with an obstruction until stopped, or when a paper and
tobacco were placed in his hands he would proceed to roll an unlimited number of cigarettes.

Two cases of a more complex exercise of certain intellectual powers, while others were dormant, came
under my observation some time ago. One of them was a young man of twenty-three, who had had
irregular epileptic seizures for some years. He went to bed one evening as usual, arose, and
breakfasted with his family without creating any suspicion that he was at all unwell. He then went down
to his place of business, and after his arrival was sent to a distant part of the city for some tool. On his
return down town he stopped at a tobacconist's and became involved in a quarrel with one of the
persons in the shop. A policeman was called, who, more intelligent than many of his class, immediately
detected something queer about the man, arrested him, and afterward took him to Bellevue Hospital.
There he remained three days, and suddenly returned to consciousness and a knowledge of his
surroundings, but was entirely ignorant of his unfortunate experience. It is unnecessary to say his habits
were perfectly good and he was not drunk at the time of the quarrel or arrest. His last recollection was
that of going to bed the night before the day of his arrest.

Another case of unusual interest which came under my care, illustrating a phase of sensory epilepsy, is
worthy of reproduction:

C. O——, aged twenty-two, is a reporter attached to one of the New York afternoon papers, who
received a severe injury of the head when but three years old. He fell from the second story of an
unfinished building to the cellar, striking the upper and back part of his head upon a beam. He was
rendered unconscious, and remained so for a day or more. He recovered from the immediate bad effect,
but has suffered from severe general headaches, which recur every week or so, with an increase in the
amount of urine excreted. About six months ago he began to have epileptic convulsions of a violent
character almost every day, and sometimes more often. These were precipitated by excitement, and he
had a great many when worried about his wife at the time of her delivery. Upon one occasion he fell
down stairs and injured himself quite severely. The attacks were, as a rule, preceded by an epigastric
aura of long duration, and occasionally by a visual aura, and, according to the testimony of his
associates, he became strange and queer. When in such a dazed condition he would restlessly wander
about his office, and suddenly, without any cry, become convulsed. After the attack he slept soundly.
The bromides of sodium and ammonium and digitalis did little or no good, but the bromide of nickel
appeared to have some influence. During the past month he has had only two or three attacks, but
these have been of a quite irregular character. He told me that there were times when he felt like doing
himself an injury, and that he had impulses to kill some one else. His companions said he was irritable,
pugnacious, and easily thwarted, and his brother-in-law stated that upon several occasions he had
queer turns, when he would raise his hand to strike some member of the family—that he subsequently
knew nothing of his conduct, and when it was detailed to him he appeared greatly astonished.

Mr. O—— came to my office in company with a friend at ten o'clock in the morning of December 27,
1883. He had had one of his attacks at the newspaper office, of rather more severe character than
usual, at eight o'clock, with a psychical aura, during the existence of which he was very morose and
sullen. Upon recovery he was speechless, though he could communicate by signs. Upon his arrival at
my office his manner was composed and he appeared somewhat dazed. His pupils were dilated, but
contracted readily to light. I asked him one or more questions regarding his inability to speak, which he
perfectly understood, and when I gave him a pencil and a piece of paper he replied without difficulty in
writing. When told to make a great effort to speak he did so, and I thought I detected the word ‘To day,’
but he could not repeat it, though he tried and expressed great annoyance. He was unable to utter any
sound except a sort of groan, which could not in any way be taken as an element of speech. I examined
his larynx, but found nothing which could explain his impaired phonation, and I sent him to Dr. Asch,
who found absolutely no abnormal appearances to account for the speech difficulty. The patient could
not phonate, and though he made attempts to enunciate the vowel-sounds, and the vocal cords were
approximated, he made no orderly sound. Asch found a slight laryngitis of no importance.

The patient went home, and remained speechless all day, and was seen by my associate, G. de Forrest
Smith, in the evening. What occurred during and after that gentleman's visit is contained in his notes: “I
was called to see patient about 8.15 P.M. He was lying upon the bed, but had not slept; recognized me
and motioned that he could not speak, and I found that he could only say one or two words, and this
with the greatest effort, and so all my questions were put so that he could answer them by nodding or
shaking his head. He knew that he had had an attack in the morning, that he had seen Hamilton and
Asch, and recalled various incidents of the day, answering intelligently my questions in regard to them.
He indicated by motions that his inability to speak was due to a lump in his throat. When asked if he had
any trouble to think of the word he wanted, he shook his head, but shortly afterward hesitated in an
answer, and when asked if this was due to his inability to think of the word, said ‘Yes.’ Was asked if he
had any loss of power in either side, and he motioned to his right arm and leg, and said that he felt a
numbness and pricking on that side. On his grasping my hands with his, the right was perceptibly
weaker.

“At one time he seemed confused as to which was his right or left side, and put up both hands, and after
looking at first one and then the other in a puzzled manner, at last decided correctly, then smiled
apparently at his confusion.

“All this time he had been half lying on the bed. He now intimated that he was tired, put his head down
on the pillow and began to belch up wind, and as he appeared about to vomit I called for a basin; but
this was only the beginning of an attack; the muscles of the neck and right side assumed a state of tonic
spasm, the extensors predominating, so that the head was turned a little to the left and forcibly thrust
back into the pillow, and the right arm and leg were firmly extended. He remained in this position about
one minute; then, taking two or three full inspirations, put his hand to his throat and said plainly,
‘Something has fallen from there.’ On being asked ‘What?’ he replied, ‘A bone has fallen from my
throat.’ I told him it was well that the bone had fallen, as now he could speak. ‘Why,’ said he, ‘I have had
no difficulty in talking.’ On being asked why he had seen Asch, he said ‘Who is Dr. Asch? I never saw
any such person.’ Further questioning showed that all the occurrences of the day (except those which
had taken place immediately before the first attack) were an absolute blank, and he thought it still
morning. He asked the time, and I told him half-past eight o'clock in the evening. At this he seemed
much surprised and said, ‘Why, I went to work this morning; how did I come here?’ I then explained to
him that he had been ill. After further conversation he said he felt sleepy, and, after resting a few
minutes, he arose, put on his slippers, and came out into the room. He walked with difficulty, because of
the loss of power in the right side, which he said felt numb and sore, as if it had been pounded, also a
sensation of pins and needles. After the attack his mind was perfectly clear, and he could talk as well as
ever, and all that had happened before the attack in the morning he could remember perfectly well, but
the interval between the two was a complete blank. His inability to speak seemed due, not to lack of
knowledge of what he wanted to say, but rather to want of power to form the words, although there was
no paralysis of the vocal muscles. When he did manage to say a word, it was invariably the correct one,
but it was always done with the greatest effort. The day after the attacks he remained at home; the next
day he went to work, but his head felt heavy and confused. Two days after he complained of a pressure
on the left side and back part of the head; otherwise he was all right. At this visit he said that after I had
left him on the night of the attacks he intently thought, striving to recall the incidents of the day, and after
a time concluded he could remember being at Thirty-third street, but did not know how he got there. He
thought he could recall going to see Asch, but would not know him if he should see him. I then asked
him how questions were answered by him on that day; he answered he did not know, as he had not
thought of that; then, after a few moments' reflection, said he must have written the answers. He was
then shown some of the answers he had written, which he recognized, and by an effort of memory could
recall some of the incidents of writing them. He was still unable to remember anything that occurred
after his arrival home previous to the last convulsion.”

January 27, 1885: This patient subsequently suffered from several attacks in which the psychical
element predominated. His head presented a remarkable deformity, there being a prominence
posteriorly which might be compared to a caput succedaneum, only it was entirely osseous. The upper
margin was separated from the anterior parts by a deep sulcus.

Under such circumstances we find very often that acts of great violence are committed by such
epileptics for which they are entirely irresponsible. Two or three cases of the kind occur to me now. One
of them was a boy who always bit every one and everything—his family, the domestic animals, and
inanimate objects; another, a most dignified and lady-like woman, who violently struck different
members of her family; and within the past week a woman was brought to me who hurled a kerosene
lamp at a perfect stranger with whom she was quietly talking before the seizure was precipitated.
Numerous instances are related where individuals while in the masked epileptic state have wandered
for long distances and committed a variety of purposeless acts, and undoubtedly many of the
mysterious disappearances are of this order.

SENSORY EPILEPSY.—Some years ago Hammond referred to certain peculiar epileptic attacks in which
sensory manifestations were very pronounced. To this condition he gave the name thalamic epilepsy,
believing the condition to be one of the optic thalamus. Among the large number of unclassified and
irregular cases reported by various authors there are many so much resembling each other that I think
they should be relegated to a special place.32 The notable examples of Sommers, Bergmann, Tagges,
Guislain, and others belong to this category.
32 I shortly afterward, believing the term a misnomer, invented that in use: “On Cortical Sensory Discharging Lesions or Sensory
Epilepsy,” New York Med. Journal and Obstetrical Review, June, 1882; also see “A Contribution to the Study of Several Unusual
Forms of Sensory Epilepsy which are probably Dependent upon Lesions of the Occipital Cortex,” New York Med. Record, April 4,
1885.

The features of this form of epilepsy are (1) the expression of some hallucination (prodromal stage), or
hemiopia; (2) supraorbital neuralgia; (3) aphasia, formication; (4) slight loss of consciousness, and little
if any motor disturbance.

I may present two illustrative cases:

A few months ago I was consulted by a medical gentleman in regard to a patient who had for years
presented a curious train of nervous symptoms, which afterward assumed a form leading me to think
she might have sensory epilepsy. She would, in the presence of the gentleman who consulted me, who
was a personal friend of the patient and a medical man, stop short in the midst of an animated
conversation, look fixedly ahead, appearing momentarily lost, remaining abstracted for a short period,
possibly a minute, and on recovering herself go on, finishing the sentence she had commenced before
the seizure. At this time she constantly had hallucinations of a visual character, when she saw animals,
birds, figures of men and women, who approached her, as well as a variety of other objects. A common
hallucination, which had been repeated quite frequently, consisted in visions in which green leaves and
white rabbits and other objects familiar to her in childhood figured extensively. Upon one occasion, while
sitting in the drawing-room, opposite a door which communicated with the hall, she suddenly called her
companion's attention to the hand of a man which she saw clasping the baluster rail. The hand was
seemingly disconnected from the arm. She was somewhat agitated, and it was nearly half a minute
before the vision was dismissed. Sometimes she would call attention to the hallucinations before the
attack, but more often she became transfixed, apparently lost, and then recovering she described her
visions minutely. She has apparently been able to foresee the attacks and ward them off by a strong
voluntary effort. So far as can be learned, there is no hysterical element in the case, but her seizures are
more frequent at the time of menstruation. In a private note it is stated that “the family history of the
patient is very good, and she has always seemed remarkably healthy and robust, and has shown more
than usual intellectual ability. She has appeared to persons generally to be of a contented, happy
disposition.... At night, when she closed her eyes, she suffered from these hallucinations, especially
after a day of fatigue. Her pupils are usually dilated, but her color undergoes no change during the
seizure.”

The second case is one of a more complex type:

J. B——, a bright boy aged sixteen, was sent to me by F. H. Bosworth in April, 1883. He comes of
nervous stock, his mother being subject to epilepsy, and his father is an eccentric man who manifests
his mental peculiarities chiefly in a morbid restlessness and irascibility. There is a brother who is healthy.
The attention of the parents was first called to the boy's condition by his recital of a sudden attack which
occurred during the summer of 1882. While rowing upon a river he suddenly and completely lost the
visual use of the right eye, so that in looking at a number of ducks swimming near his boat he failed to
perceive those upon one side of the flock. This condition lasted for twenty minutes, and after a brief and
severe pain over the right eye he became unconscious, the unconsciousness being preceded by a
tingling and numbness of the hand, forearm, arm, and left side of the tongue. He has subsequently had
eight or ten of these attacks, of which the following is an example: Usually without any bad feelings,
physical or mental, he, while engaged in any duty or at any time, suffers a sudden unilateral blindness.
This is never gradual, and not like the form of amblyopia in which the visual field is gradually reduced.
There is some hemichromatopsia. It would seem as if the retinal anæsthesia was unequal, for while
usually the loss is complete and universal, it sometimes happens that there is only a limited loss. Upon
one occasion, while reading, he suddenly lost the printed matter of the right lower half of the page below
a diagonal line extending from the right upper corner to the left lower corner. There is never diplopia.
This deprivation lasts anywhere from ten to twenty minutes; meanwhile, a distal anæsthesia, coming
very gradually, involves at first the fingers of the opposite hand, and successively extends to the
forearm, arm, and other parts, as I have already mentioned. There seems to be analgesia as well as
anæsthesia, for a pin may be run into the muscles without producing pain, and upon one occasion the
gum was freely pricked without any discomfort to the patient. It invariably happened that the cutaneous
sensory trouble occurred upon the side opposite to the hemianopsia and neuralgia, and in the greater
number of instances the left side was that affected. The third stage of the attack consists in migrainous
headache of a very severe kind, and which sometimes lasts for an hour or more. There is a subjective
feeling as if the eye was pushed forward. This disappears with nausea and relaxation. More often he
loses consciousness when the anæsthesia reaches its limit, which seems to be the extension of the
anæsthesia to the gums. Occasionally there are slight convulsive movements upon the anæsthetic side.
While the attacks involve the left side of the body as a rule, it happens that when there is primary left
hemianopsia and right-sided anæsthesia the boy becomes very much confused in speech, and
sometimes is paraphasic, the trouble being but transitory. He is sometimes unable to speak at all,
though perfectly conscious and in possession of his faculties. No pupillary disturbance has been noticed
at any time. Upon two occasions there was a swelling of gums and tongue, which was not only
subjective, but perceived by the mother. Occasionally he sees prismatic colors and rays before the
blindness, but this has been only once or twice. During his early life he had attacks of slight numbness
of the hands and feet which were not thought much of, and he had headache as well. He has been a
somnambulist.
MORBID ANATOMY AND PATHOLOGY.—The literature of the experimental physiology of epilepsy is enriched
by the observations of a variety of careful students, among them Sir Astley Cooper, Kussmaul and
Tenner, Brown-Séquard, Nothnagel, Schroeder Van der Kolk, Pitres, Hughlings-Jackson, and the
followers of the localization school, as well as many others more or less distinguished.

The experiments of many of the early writers were directed for the purpose of ascertaining the relations
of circulatory variations to convulsive seizures, and the most notable were those of Burrows and
Kussmaul and Tenner. These latter produced compression of the carotid arteries, and instituted cerebral
anæmia by free and exhausting hemorrhages. As a consequence, the emptying of the cerebral vessels
was followed by a loss of consciousness and by epileptiform convulsions, and it was necessary to
produce the same result to compress all the great afferent vessels of the brain. The experience of
surgeons generally is, that ligation of the common carotid upon one side of the neck is sometimes very
apt to produce an alarming anæmia, with occasional convulsions, and sometimes fatal consequences.

The experiments of Hall, Landois, Hermann, and others, as well as those of the writers just mentioned,
show that carotid compression results in capilliary anæmia and venous hyperæmia, and that with
cessation of this pressure there is a sudden congestion of all vessels. The susceptibility of the brain is
greatest at its posterior part and between the optic thalami and the cord. When the bulb was subjected
to sudden changes in its nutrition—such, for instance, as followed the experiments of Hermann, who
ligated simultaneously the superior and inferior venæ cavæ of a rabbit—there were not only
convulsions, but various cardiac and other disturbances which were undoubtedly due to central
impairment of function. Kussmaul and Tenner conducted their experiments with watch-glasses luted into
the cranium—a procedure which, however, at best, is unreliable.

Brown-Séquard some years ago in part established an important pathological truth, the theory of
epileptic zones, and demonstrated in certain animals that bruising and injury of the great nerve-trunks,
especially the great sciatic, would give rise to epilepsy, and that irritation of certain tracts would
precipitate the paroxysms. He further announced that the progeny of animals in whom epilepsy had
been thus induced very frequently inherited the epilepsy of the parent. By some it was held that such
epilepsies were purely peripheral, and Brown-Séquard even believed in spinal epilepsy. His spinal
epilepsic theory has, for the most part, been explained by the anatomical researches of Hitzig and the
doctrine of interrupted spinal inhibition. In fact, many of the spinal epilepsies are examples of
exaggerated reflexes.

The epileptiginous zone theory, which, while it induced many to believe that the disease might have its
origin outside of the brain, gave rise to the false assumption that attacks with distal auræ were primarily
non-cerebral, has been discarded, and most observers have arrived at the conclusion that even in these
cases the first explosion is due to some cerebral cell-discharge.

Hughlings-Jackson's grand work has revolutionized the views held prior to his first published writings,
about twelve years ago. He believes that any part of the gray matter may, through over-excitability, give
rise to convulsive attacks.

The production of convulsions by cortical irritation is now an old story.

The experiment of Pitres and Frank33 bears upon the sensorial function of the cortex in showing that,
when the cortex is irritated, epileptiform convulsions follow, but if the exposed surface be subjected to
the ether spray the same irritation will only produce definite movements, but no convulsions.
33 Gazette des Hôpitaux, No. 38, 1883.

The investigations of Van der Kolk especially, and his followers, certainly give the medulla an important
place as the locus morbi of the malady; and it must be assumed, bearing in mind the existence of the
vaso-motor centres of Dieters and the presentation of symptoms indicative of disturbance at the floor of
the fourth ventricle, that the most important pathological changes must be looked for in this part of the
brain.
Jackson's cortical explanation is, however, fully in consonance with the medullary theory. If we study the
different stages of the attack, we shall find that there is probably a suspension of cortical inhibition—that
a derangement of the cortical cells or discharge may cause a resulting disturbance in the bulb. On the
other hand, a reflex irritation through the pneumogastric or from some distal part brings about the same
disturbance of equilibrium. There is anæmia due to irritation of the vaso-motor centre, an inhibition of the
great ganglion-cells, and a disturbance of function of the important cranial nerves. The primary anæmia
and unconsciousness are accounted for by this primary irritation of sympathetic filaments and vascular
constriction; the secondary hyperæmia is explained by the experiments of Kussmaul, which
demonstrated the succeeding congestion; or by irritation of the spinal accessory and contraction of the
muscles of the neck and compression of the large veins. The pupillary, ocular, respiratory, and other
symptoms indicate the disturbance of the nerve-nuclei in the bulb. The respiratory difficulty and the
interrupted decarbonization of the blood undoubtedly account for the secondary unconsciousness.

Van der Kolk34 in localizing the lesion in the medulla found capillary dilatations in the neighborhood of
the hypoglossal nuclei in tongue-biters. In epileptic patients who were in the habit of biting their tongues
during the fit the vessels were wider than in those who did not bite the tongue, on an average in the
course of the hypoglossus by 0.096; in the corpus olivare, which certainly here plays an important part,
by 0.098 mm.; and in the raphé by 0.055. In those who did not bite the tongue, on the contrary, the
vessels in the path of the vagus were 0.111 wider than in those in the first, Table A.35
34 “On the Minute Structure and Functions of the Spinal Cord,” by J. L. C. Schroeder Van der Kolk, New Syd. Soc. Trans.

35 TABLE.
Different Epileptics. Hypoglossus. Corpus olivare. Raphé. Vagus.
Table A—tongue biters 0.306 0.315 0.315 0.237
Table B—non biters 0.210 0.217 0.217 0.348
Difference +0.096 A. +0.098 A. +0.055 A. +0.111 B.

Nothnagel36 is of the opinion that the anæmia of the brain is not the cause of the convulsions, but that
the “excitation of the vaso-motor centre and that of the centre for the muscles are co-ordinate—that both
go on side by side, and are independent of each other.”
36 Ziemssen's Encyclopædia, vol. xiv. p. 268.

He by this theory explains the occurrence of those forms of petit mal in which there is loss of
consciousness without convulsions, and, on the other hand, twitchings before the coma.

The best argument in favor of this hypothesis is in Jacksonian epilepsy, when monospasms exist
oftentimes with a succeeding extension.

In those cases which are the outgrowth of migraine the pathological condition is probably an
exaggerated tendency to angio-spasm, the original impaired vascular tonus in the beginning giving rise
simply to pain and lesser troubles, while after repeated changes of calibre not only nutritive alterations
ensue, but hyperexcitability of the bulbar convulsion centres as well.

The labors of those who have endeavored to connect epilepsy with cerebral-tissue alterations have
been attended by nothing very definite or positive, so far as pathological explanation is concerned. The
post-mortem appearances have varied widely, and the only conclusion to be reached is that which
shows that almost any morbid gross alteration of the cerebral mass may be symptomatized by
convulsions, but such a production of paroxysmal trouble is much more likely to be the case, and in a
more definite manner, when the cortical motor-centres are subject to destructive disease or irritative
pressure. This is even not always the case, for numerous cases of injury of the paracentral lobe have
been recorded with no showing of resulting convulsions. The long list of autopsies which I will not here
consider show that an epilepsy may owe its origin to the pressure of a spicula of bone, or to the
pressure exercised by depressed fragments of the same—to tumors or adventitious products,
meningitis, cortical encephalitis, vascular degeneration, ventricular œdema, contusio-cerebri, and many
other morbid processes which result in rapid or tardy degeneration. Of course, in such cases the
genesis of the disease depends not so much upon the nature of the lesion as the location. The fruitful
collections of cases of Ogle and Jackson are full of examples of limited growth or disease involving the
cerebral cortex, while numerous cases collated by other writers show disease of the bulb or various
peripheral parts which have been closely connected with the growth and behavior of the affection.

Several able pathologists have independently and repeatedly found that sclerotic degeneration of the
hippocampal folds often existed. Delasiauve and Lébert first observed this lesion, but many modern
authorities—among them Meynert, Nothnagel, and Charcot—who have also found this appearance,
regard the change as of purely secondary, and consequently unimportant, character.

Tamburini37 reports a case of hemiplegic epilepsy with induration of the left optic thalamus and the left
cornu ammonis, in which aphasia existed. Pfleger38 and Henkes have also found the sole lesion to be
induration of the cornu ammonis. Of Pfleger's39 43 autopsies, atrophy and sclerosis of the cornu
ammonis were found twenty-five times, and it was noted that the extent of the morbid change bore
relation to the violence and frequency of the seizures.
37 Sallanzani, Modena, 1879, viii. 550-557.

38 Allg. Zeitschrift f. Psychiatrie, etc., Berlin, 1879, xxxvi. 359-365.

39 Ibid., lxxvi., and Arch. de Neurologie, No. 2, 1880, p. 299.

In many examples, especially where the disease has been found to be unilateral and associated with
more or less hemiatrophy, the autopsy disclosed a corresponding hemiatrophy of the brain. Many such
cases are reported. I have frequently found epilepsy in association with cerebral hypertrophy, and as a
symptom of cerebral tuberculosis it has long been recognized, and numerous cases are reported in
which for a long time the paroxysms were the only manifestations of the condition. In one of these
cases, reported by Luys,40 the bulb was found involved by tuberculous matter.
40 Archives gén. de Méd., 1869, ii. 641 et seq.

Convulsions have very frequently been noted in association with imperfect cerebral development, and
Echeverria laid great stress upon the hyperplastic increase in volume of certain parts of the brain.

Marie Bra41 has thus summed up her conclusions relating to the morbid anatomy of epilepsy:

“1. The mean weight of the brains of epileptics is less than the physiological mean.

“2. The cerebellum is greater than the physiological mean.

“3. There frequently exists an asymmetry between the lobes (not peculiar to epilepsy). The increase
of weight is sometimes found on the right and sometimes on the left side. Equality is the exception.

“4. In no form of mental disease (excepting perhaps general paresis, which is accompanied also by
epileptiform crises) have we met with so marked and constant a variation between the weights of
the hemispheres as exists in epilepsy.”
41 Referred to by Axenfeld.

Drasche, Green, Greenhow, Löbel, and others have detailed cases in which tuberculous deposits were
undoubtedly the causes of the disease.

Kussmaul and Tenner, Hoffman, and others have held that a stenosis of the superior part of the
vertebral canal may explain, through pressure upon the cord, the genesis of the attack, and Kroon found
asymmetry of the medulla oblongata.
The microscopical changes that have been found in brains where no gross lesion was apparent are by
no means distinctive. I have myself examined the brains of many epileptics with discouraging results.
The varying granular cell-degeneration, capillary dilatation, and exudative changes are common
enough. In several cases of cortical epilepsy I found more or less advanced degeneration of the great
cells in limited regions.

By far the most important and exact changes are those observed in the cases of sensory epilepsy. I
have elsewhere collected some continental cases. In brief, areas of occipital softening or degeneration
have been discovered in those cases with hallucination, sensory expressions, and hemiopia. In one
case attended by hallucination of smell the autopsy disclosed the following:

M. M——, was a stout Irish woman about forty years of age. She had suffered from a light form of
epilepsy dating from the tenth year, and resulting, as she stated, from a fall, when she struck her head
and was unconscious thereafter for some hours. No scar was visible, however. No satisfactory history
could be obtained regarding her early life and the first paroxysms. In the beginning these were rather
frequent, and she had as many as four or five a month. They afterward diminished in number and
severity, and for many years she had but three or four in the course of the year. They were not very
severe, and she was enabled to pursue her work as a housemaid, but did not keep her places for any
great length of time. She rarely bit her tongue, but usually frothed at the mouth and became livid and
convulsed for a short time. There was no history of one-sided spasms. As I have stated, I could gain no
accurate account of the previous attacks, except that she nearly always had an aura of a peculiar
character, which was a prominent feature of the seizure and very pronounced. She suddenly perceived
a disagreeable odor, sometimes of smoke, sometimes of a fetid character, and quite uncomplicated by
other sensory warnings; and afterward became unconscious, and remained so for two or three minutes.
She was invariably able to describe her sensations when she recovered, which she always did when I
asked her, comparing her warning to the smell of burning rags, to the smell from a match, and, as she
expressed it, it sometimes rose up in her head and choked her. She was under my observation for one
or two years, but eventually developed phthisis, and died, her attacks occurring from time to time until
her death.

Besides well-marked tuberculous lesions in the lungs, there was little of interest so far as the visceral
examination was concerned. The brain was removed and its peculiarities were carefully observed. A
great quantity of fluid was found, especially at the dependent portions of the membranes and in the
ventricles, while the dura was thickened and pearly in spots. There was a condition that might be
likened to a low grade of hemorrhagic pachymeningitis, and at the base of the brain old plastic changes
were found, there being adhesions, especially in the region of the middle lobes, but more particularly on
the right side and near the median line. The brain as a whole was small, and weighed forty-one ounces
and a fraction. The sulci were deep and gaping, and the convolutions were distinct. There was no
atrophy of the fore-brain convolutions, and no other pathological appearance was presented except that
found in the meninges, but at the lower part of the temporo-sphenoidal lobe of the right side an
appearance was found of an exceedingly interesting nature. At this point a decided shrinkage of tissue
was discovered, with depression and adhesion of the pia, the induration involving the uncinate gyrus
and parts of the adjacent convolutions, as represented in the drawing. No induration or softening of the
great motor tracts was observed, and the optic thalamus and parts adjacent were uninvolved, as was
the cord. An attempted microscopic examination, undertaken some months subsequently, was
unsatisfactory, because of the bad condition of the brain, the preserving fluid having been improperly
made. The olfactory nerves were not involved. The third frontal convolution was examined, but no
disease was found there. Consequently, it is to be inferred that no lesion of the external root of the
olfactory nerve existed.
 FIG. 27.

Lower Face of Right Hemisphere.

DIAGNOSIS.—Having spoken of epilepsy as in most instances a

symptomatic disorder, it would be proper to confine this section to
the differentiation of the simpler and more classical form of the
idiopathic disorder from certain purely eclamptic attacks or those due
to cerebral tumor or coarse degeneration. The epileptic attack itself
is to be considered from its time of happening, its duration, the
element of unconsciousness, its associations, and the antecedent
history of the individual. It may be confounded with the similar
phenomenon dependent upon cardiac weakness, uræmic poisoning,
toxic or alcoholic saturation, etc.

Of course, when we find recurring seizures with a certain amount of

what Carter-Gray calls quasi-periodicity, preferring perhaps the night,
the early morning, or only the daytime, we are almost sure of
epilepsy. This supposition is strengthened by the association with
attacks of petit mal. The duration of an attack, which may be from a
few seconds to several minutes, is also a guide, for in certain toxic
and other paroxysms the rule is for a succession of attacks to occur.

The question of consciousness is one that has drawn forth a great

deal of discussion, especially with reference to medico-legal cases. I
think the majority of clinicians are agreed that loss of consciousness
is an absolute belonging of epilepsy, yet there are cases in which the
lapse is scarcely perceptible. It is a dangerous precedent to
establish, for the convulsive symptoms in such cases are taken from
the epileptoid category. It is quite true that there are many hemi-
epilepsies in which the intellectual condition is one that may easily
be mistaken. I have seen numerous cases in which an apparent
conservation of consciousness remained throughout a slight
monospasm, but I do not feel at all sure of this; and in cases of
aborted or masked epilepsy there is a dual mental state which would
readily deceive the lay observer. The case of Mrs. S—— is an
example of this kind. After the obvious subsidence of the dramatic
and conspicuous feature of the fit she remained for hours and days
in a state of undoubted transposition, performing acts which required
something more than a high degree of automatism—going to the
table, talking about certain subjects which were suggested, with
apparent ease, but not connecting them intelligently with her
surroundings, as she would before and after the epileptoid state.
After a time she apparently resumed her normal state, but was
entirely unconscious of the happenings of her previous hours or
days, not even recollecting her simplest actions. Julian Hawthorne's
hero in Archibald Malmaison, though not drawn by a physician's pen,
suggests the state of which I speak, and it has the merit of being
based upon one of the elder Forbes Winslow's interesting cases.

When we find paroxysmal attacks occurring in individuals with

atypical heads, thick swollen lips, scarred tongues, and irregular
teeth, we may strongly suspect the patient to be epileptic. To these
we may add the appearance of the eyes, the fishy, lack-lustre
expression which betokens old epilepsy. The hands are clammy and
the skin mud-colored; the hair is dry and coarse; and the body often
has a death-like odor.

In children, certain mental peculiarities are to be inquired into.

Unnatural brightness or dulness—what may be called the clumsy
organization—is often present, and the muscular use is often
imperfect. We find that there is often but little nicety in walking, in
using the hands, in speaking, or after learning to write there is an
incapacity, with ever so much teaching, to develop a character or
style. Such children can never become ambidextrous. These little
points may seem trifling, but to the physician who carefully studies
his cases they may prove of great help. The history of the nights may
often lead to the discovery perhaps of a long-existing nocturnal
epilepsy. Incontinence of urine, blood upon the pillow, nightmares,
morning headache, and petechiæ betoken unsuspected night
attacks; and Le Grand du Saulk mentions the case of a young
Englishman who committed a purposeless crime and was
discovered to be epileptic, the diagnosis being confirmed by an
antecedent history of nocturnal seizures, and subsequent watching
resulted in the discovery of many night attacks.

As to special conditions with which the epilepsy may be confounded,

I may refer to cardiac weakness. It not rarely happens that simple
fainting attacks are confounded with those of an epileptic nature.
Such is the case more often in heat-prostration, when some rigidity
attends the loss of consciousness. The duration of such a state, the
condition of the pulse and color, however, will easily clear up any
doubts upon the part of the observer. The existence of a cause
should also be considered, and the fact that usually the epileptic
paroxysm is sudden, while a feeling of depression and feebleness
precedes the fainting attack, should be remembered. I may present
in tabular form the points of difference:

EPILEPSY. SYNCOPE.
Loss of consciousness Loss of consciousness follows
sudden. feeling of faintness.
Period of complete Unconscious throughout, no
unconsciousness usually convulsions.
short.
The existence of auræ of a The existence of preliminary vague
well-defined type. prostration, nausea, and irregular
heart action.
Often involuntary discharge Quite rare or never.
from bowels and bladder.
Patient usually falls into heavy After slight weakness patient is
sleep or is indifferent after anxious and worried, and quickly
convulsion. seeks relief.

The difficulty of diagnosis, however, is only in cases of petit mal.

There are light forms of auditory vertigo that may resemble
vertiginous epilepsy. In the former there is never loss of
consciousness, and the patient refers to the rotary character of the
vertigo. A history of antecedent attacks, tinnitus, aural disease, and a
certain constancy which is not a feature of petit mal, may be
mentioned.

There are cases, however, which are puzzling, and come under the
head of auditory epilepsy rather than auditory vertigo; and in these
there is a multiplicity of expressions, the auditory symptoms
predominating.

Of uræmic convulsions it is hardly necessary to speak. There is a

previous history of renal disease which the microscope and less
delicate tests will reveal, and clinically there is antecedent headache,
some stupidity, and not unrarely thickness of speech and
somnolence. There are some cases, however, which are obscure. I
have known patients with chronic renal disease—such as waxy
kidney, for instance—to develop a species of epilepsy, the
paroxysms recurring from time to time and behaving very much as
the idiopathic disease would; and their occurrence would mark some
imprudence in diet or exposure, and their disappearance an
improvement in the patient's general condition. The attacks were not
classical, inasmuch as there seemed to be but one stage of violent
clonic convulsion, preceded by intellectual dulness, and followed by
a semi-comatose condition which was far mere profound than the
somnolent stage of epilepsy. The movements were not accompanied
by a great degree of opisthotonos or pleurosthotonos.

Alcoholic and absinthic epilepsies are usually preceded and followed

by symptoms indicative of profound saturation.

The consideration of hysterical epilepsy may be found elsewhere,

but it may do to briefly refer to some cases which do not present the
phenomenon first described by Charcot and Bourneville. The
ordinary hysterical attack is never attended by loss of
consciousness, by any of the pupillary changes so constant in
epilepsy, by the mobility of the pupil between the attacks which is
present in a large number of true epileptic individuals. There are
never the succeeding changes of color, and the seizures are
commonly produced or attended by some emotional disturbance, or
are associated with ovarian disturbance.

Epilepsy is occasionally simulated by malingerers, and sometimes

the skill of the subject is so great as to even deceive the practised
eye. Prisoners, soldiers, and litigants may counterfeit an epilepsy,
and go through with great personal suffering to accomplish their
purpose. “Clegg, the dummy-chucker,” whose remarkable case has
figured in the medical journals, upon one occasion threw himself
from an iron platform to the stone floor of the jail, nearly twenty feet
below, to convince a suspicious physician of his honesty. The
simulator rarely bears close watching. The dilatation and contraction
of the pupil cannot be simulated, nor can the corneal or pupillary
insensibility. The fraud cannot voluntarily change his color, as is the
case in true epilepsy, and as a rule the thumbs of the impostor are
never flexed, as they should be. Suggestions for a purpose are
readily heard, and sometimes adopted, by the apparently
unconscious man. Gottardi42 lays great stress upon the use of the
ophthalmoscope as a means of detecting simulated epilepsy. He
calls attention to the frequency of retinal changes with facial
asymmetry and other evidences that suggest cerebral disease or
non-development. Gottardi has found that the pulse in true epilepsy
is always lower after an attack, but it soon reaches its normal
standard.
42 Abstract in Journal of N. and M. Dis., Oct., 1881, p. 843.

The differentiation of idiopathic epilepsy from that due to syphilis is

possible when we consider the element of pain. Besides the tibial
pains of syphilis, the epilepsy thus produced is often preceded by
intense frontal headache, while that of ordinary epilepsy follows the
attack. The syphilitic epilepsy is not attended by so great or
continued a loss of consciousness as the non-specific form, and the
movements are apt to be most violent on one side or the other. The
association of the attacks with various bodily signs, such as nodes,
old scars, alopecia, erosions, etc., and in connection with possible
cranial nerve-paralyses, will throw light upon its true character. The
paralyses referred to seem most frequently to involve the motor
ocularis, abducens, and patheticus. Syphilitic epilepsy, too, is quite
irregular in its time of manifestation, and is not unrarely followed by
aphasia; but the interparoxysmal mental state is one of extreme
dulness, memory being blunted and there being a peculiar hebetude.

PROGNOSIS.—Within the past quarter of a century the ideas of the

medical profession regarding the curability of epilepsy have certainly
undergone a change. The statistics of Bennett and others show that
since the introduction of the bromic salts the proportion of cures has
been decidedly increased. Nevertheless, the disease is a most
discouraging and troublesome one to manage, and especially is this
the case when it assumes the form of petit mal. The rapid recurrence
of light attacks is, as has already been said, very apt to lead not only
to mental enfeeblement, but is very often followed by very severe
paroxysms.

Epilepsy of a more or less constant form, in which the seizures

resemble each other, is far more incurable than that of variable type;
for example, we find that unilateral seizures are much more apt to be
associated with established cortical disease than when they are
general and simply explosive manifestations. It has been held that a
tendency to permanency is marked by a diminution in the extent of
the interval. This is by no means true. I have had cases under
observation for ten or twelve years in which attacks separated by
intervals of six months or one year marked the course of the
disease, in which frequent initial attacks were present. These cases I
regard as very bad so far as prognosis is concerned. I much prefer a
history of irregular and comparatively disorderly attacks. In female
subjects the menstrual influence is not always a bad factor. When we
are enabled to remove some production of an exciting cause in
connection with the catamenia the prognosis is more hopeful; but an
opinion must be expressed with great caution, especially in those
cases beginning at an early age and not after the establishment of
the menses. Traumatic cases are not always bad, but those in which
the element of heredity plays a part most certainly are, Herpin and
Gowers to the contrary; and though these cases for a time do well
under treatment, its good effects are not constant. Individuals with
misshapen heads, whose deformity suggests premature sutural
ossification, are not susceptible to the influence of treatment, and all
other osseous changes, such as exostosis, plaques in the dura, and
bony growths, whose existence can only be guessed at or inferred
from suggestive appearances elsewhere, give rise to a variety of
epilepsy which is beyond the reach of drugs. With symptomatic
epilepsies the case is sometimes different, for while the seizures
which arise from the irritation of a cerebral tumor are almost as
hopeless as the form I have just mentioned, we know from
experience that the epilepsy of syphilis and other allied diseases,
and those of toxic origin, with the exception sometimes of those
occasionally due to alcohol or lead, are curable. The meningeal
thickening of alcoholic origin or the encephalopathy of lead may be
the pathological bases of very intractable paroxysms.

So far as age is concerned, it may be stated that many eclamptic

seizures of young children which are due to well-recognized irritable
causes are promptly cured if there be no hydrops ventriculi or
preossification of the coronal sutures, and if the epileptic habit is not
established. The epilepsies of six or eight years' standing are not
encouraging from a therapeutic point of view, and those of advanced
life developing in aged persons are equally unfavorable.

The treatment of epilepsy due to heat-stroke is by no means

satisfactory, and, though the attacks are often separated by long
periods, they are apt to recur in spite of drugs.

Gowers has prepared several valuable tables which show the

influence of age upon recovery. He says: “The following table shows
that age has a distinct influence on prognosis. The percentage of the
unimproved cases to the whole is 30 (43:143::30:100). The
percentage of the cases commencing at each age arrested and
unimproved is stated, and between brackets is indicated the excess
of the arrested or unimproved cases at each period of life over the
proportion for the whole 30 and 70 per cent. respectively:

———Cases.——— ———Percentage.———
Unimproved. Arrested. Unimproved. Arrested.
Under 10 14 29 32.5(+2.5) 67.5
From 10-19 23 45 34 (+4) 66
20 and over 6 26 19 81 (+11)
43 100 30 70

Thus, the proportion of the cases commencing under twenty in which

arrest was obtained is considerably less than the proportion of cases
commencing over twenty, the difference amounting to about 13 per
cent. The period of the first twenty years of life at which the disease
commences has little influence, but the prognosis is little better in the
cases which commence under ten than in those which commence
between ten and twenty: arrest is more frequently obtained. The
cases which commence in women at the second climacteric period
are also obstinate, although not sufficiently numerous to be
separately given.”

He also finds, from an analysis of the same cases, the fact noted by
others, that the prognosis is favorable in inverse proportion to the
duration of the disease.

Attacks which chiefly occur in the daytime are much more amenable
to treatment than the nocturnal seizures, and especially is this the
case in the tongue-biting form. Sudden blows upon the head or falls
have been known in isolated cases to effect an amelioration in the
patient's disease, but these examples are rare.

The existence of an aura is much better than if none existed.

Death from the attack itself is rare, yet in the large pauper institution
with which I was connected for many years I have known of several
cases. More often the death results from asphyxia resulting from a
bolus of food which chokes the patient or from a fall in some
dangerous place—into the fire or elsewhere. Accidental death from
drowning is more common than any other form.

The status epileptica into which patients sometimes pass who have
had many convulsions is occasionally a fatal termination of the
malady, and is always a serious feature.

The influence of different epochs in life is worthy of consideration: of

menstruation, of marriage, of pregnancy, and of the menopause
there is much to be said. I have sufficiently spoken of the
establishment of the menstrual flow, and I would only add another
word of caution against giving a too favorable prognosis except in
those cases of very recent origin. Marriage appears to have very
little to do with changing the attacks, unless they be of an hysteroid
character. I have never known epilepsy to influence the course of a
pregnancy in any unfavorable way, and I think this has been the
observation of others. Gowers refers to cases in which the attacks
ceased during the time the mother was carrying the child.

The occasional bad influence of the pregnant state has been

illustrated by a case reported by Terrillon.43 This example was a
woman who had been the victim of epilepsy of hereditary origin since
her seventh year. At the commencement of menstruation her attacks
became periodic, and recurred every two months, and she had
several two or three days before the flow. Two pregnancies followed
several years afterward. During the periods they were increased in
number and severity, and occurred several times daily. She had
more attacks at this period than in all the time after delivery.
43 Annales de Gynécologie, June, 1881, p. 401.

I have found that the relief of uterine flexion or the establishment of

menstruation has exercised a decidedly modifying influence for the
better in several cases.

Sometimes the disease is interrupted by the menopause, but very

often in my experience it has changed in type and been followed by
mental degeneration.
The prognosis of the epileptoid mental state is serious in the
extreme, and even when in early life the attacks are aborted or
changed to perversion of the intellect or emotions existing as a
complication, the hope of cure dwindles almost to nothing.

The spontaneous cure of epilepsy is rare. Its course, however, is

more often interrupted by some intercurrent disease. Of 33 cases
observed by Delasiauve in which there was some complicating
disease, such as erysipelas, pneumonia, pleurisy, acute articular
rheumatism, burns and contusions, scarlatina, erythema, or the
condition of pregnancy, it was found that in 25 cases there was a
decided improvement (une heureuse influence), and in 8 only was
there no modification of the attacks. Axenfeld is disposed to believe
that acute and febrile disorders more decidedly influence the course
of epilepsy than those of a chronic nature. The influence of either
may be small or may diminish the number of attacks without curing
the disease.

TREATMENT.—The treatment of epilepsy depends so much upon the

form of the disease that no arbitrary rule can be laid down upon
general principles, and we must bear in mind the necessity for
removing the exciting causes if possible, the preservation of the
balance of cerebral blood-pressure and cell-equilibrium, and the
nutrition of the cerebral mass. It may be our purpose to remove
various toxæmic or cachectic states as well. As a minor division of
treatment we must consider the abortion of the attack when auræ
are detected. It has been the custom, I think among too many, to
manage the disease in an empirical manner, depending upon some
routine course of treatment, such as the indiscriminate use of the
bromides, for example. I am convinced that the intractable character
of the disease has come to be greatly exaggerated through failures
attendant upon the wholesale use of the bromic salts, without regard
to the indication in each particular case. It behooves us first to select
a reliable bromide, and then to give it with relation to the time of the
fit, its severity, and the condition of the individual. The importance of
this has impressed me very often. An equally divided daily dose will
not do a patient who has matutinal attacks the same good as will a
large dose at night, and in certain anæmic individuals the bromides
very often increase the attack. Then, too, the cases in which
seizures of petit mal predominate are not benefited to the extent that
those are in which the repetition of severe attacks is the feature. The
bromide should always be well diluted and given when the stomach
is empty or nearly so. The bromide of sodium is, to my mind, the
most serviceable salt, and when given alone or combined with the
bromide of ammonium is better than the potassic salt or the various
others. It should be carefully kept in solution or in waxed-paper
powders in a tight preserve-jar. As to its method of administration, I
much prefer the use of a solution which shall combine other
adjuvants which I will presently mention, and separate powders as
well, which may be used to reinforce the dose. The latter are to be
employed by the patient in the event of an unlooked-for series of
attacks, and are to be used to the point of producing mild bromism at
the time. In certain cases it is all-important to take into consideration
the condition of the heart. In some cases where there is manifest
and continued cerebral congestion, with full vessels and hard pulse, I
have found that the combination with chloral or aconite was
desirable. In other cases where the heart's action was irregular and
weak my experience with digitalis and nitro-glycerin or with
strychnine was most happy. In those cases that passed large
quantities of clear urine of low specific gravity the action of nitro-
glycerin and digitalis has been most prompt, and the same has been
true of epilepsies of migrainous origin.

There have been various methods of using the bromide suggested

which are more or less practical. The writers of a few years ago
suggested the prompt production of bromism—a state in which the
patient should be kept for a long time. This I strongly disapprove of,
not only because the mental and physical depression is a very
disagreeable and sometimes permanent condition, but because I
have found that the attacks are often increased after a time, though
at first they may have been suppressed. Others believe in giving the
bromides at intervals, with periods when no medicine at all is
administered.
For my own part, I am in favor of the establishment of a mild
bromism which does not extend beyond a slight anæsthesia of the
fauces or the appearance of slight acne. If I find it necessary to
increase the dose, I give cod-liver oil, iron, and bark or some of the
many preparations of the hypophosphites for a time; and they do not
diminish the specific effects of the drug to any great extent. In the
event of a series of attacks I direct the patient to take an extra dose
at such time as will anticipate the seizure.

This treatment should be kept up for at least two or three years after
the attacks have disappeared, and it may be even necessary to
continue a bromide course in a small way for an unlimited period.

If there be an hysterical element, or if ovarian excitement is

supposed to have anything to do with the attacks, the combination of
cannabis indica is strongly recommended. This suggestion holds
good in the cases where migraine is associated with the epilepsy, or
the latter is an outgrowth of the former.

Next to the bromides of potassium and sodium I have been very

successful with the nickel bromide. I find that it is retained with little
trouble, producing no gastric derangement if taken after eating. A
syrup prepared by most of the good pharmacists is preferable to any
other method of administering the salt. Quite recently Leaman44 has
reported two cases of severe epilepsy which were greatly benefited.
His conclusion is that it does most good in the form of the disease
when the attacks are separated by long intervals.
44 Med. News, Apr. 18, 1885.

The bromate of potassium, which has been used by Weir Mitchell45

and Hinsdale, may be worthy of a trial. It should never be given,
however, in larger doses than five or ten grains thrice daily. These
investigators found that thirty grains slowed the heart very
considerably, and forty grains produced watery discharges from the
bowels and drowsiness. In their hands, notwithstanding these
disagreeable effects, it controlled the seizures. The hydrobromate of
conia is a comparatively new remedy which has been