CHAPTER 1

ABSTRACT

Nanotechnology is a multidisciplinary field that covers a vast and diverse array of devices
derived from engineering, physics, chemistry, and biology. Nanotechnology has opened up by
rapid advances in science and technology, creating new opportunities for advances in the fields
of medicine, electronics, foods, and the environment. Nanoscale structures and have been
explored in many biological applications because their novel properties and functions differ
drastically from their bulk counterparts. Their high volume/surface ratio, improved solubility,
and multi-functionality open many new possibilities. The objective of this review is to describe
the potential benefits and impacts of the nanobiotechnology in different areas. Artificial
intelligence (AI) has been developing rapidly in recent years in terms of software algorithms,
hardware implementation, and applications in a vast number of areas. In this review, we
summarize the latest developments of applications of AI in biomedicine, including disease
diagnostics, living assistance, biomedical information processing, and biomedical research[1].

The aim of this review is to keep track of new scientific accomplishments, to understand the
availability of technologies, to appreciate the tremendous potential of AI in biomedicine, and to
provide researchers in related fields with inspiration. It can be asserted that, just like AI itself, the
application of AI in biomedicine is still in its early stage. Advances in biological and medical
technologies have been providing us explosive volumes of biological and physiological data,
such as medical images, electroencephalography, genomic and protein sequences. Learning from
these data facilitates the understanding of human health and disease. Developed from artificial
neural networks, deep learning-based algorithms show great promise in extracting features and
learning patterns from complex data. The aim of this paper is to provide an overview of deep
learning techniques and some of the state-of-the-art applications in the biomedical field. We first
introduce the development of artificial neural network and deep learning[2].

Keywords: Biosensors, Medicine, Foods, Pathogens, Biosurfactants, Nanobiotechnology,

Toxicity, Deep learning; Big data; Bioinformatics; Biomedical informatics; Medical image;
High-throughput sequencing, Artificial intelligence Machine learning Deep learning Neural
network Biomedical research Healthcare applications.
 CHAPTER 2

INTRODUCTION

Artificial intelligence (AI) is defined as the intelligence of machines, as opposed to the

intelligence of humans or other living species . AI can also be defined as the study of ‘‘intelligent
agents that is, any agent or device that can perceive and understand its surroundings and
accordingly take appropriate action to maximize its chances of achieving its objectives. AI also
refers to situations wherein machines can simulate human minds in learning and analysis, and
thus can work in problem solving. This kind of intelligence is also referred to as machine
learning (ML)[2] .

Deep learning is a recent and fast-growing field of machine learning. It attempts to model
abstraction from large-scale data by employing multi-layered deep neural networks (DNNs), thus
making sense of data such as images, sounds, and texts .

Deep learning in general has two properties:

(1) multiple layers of nonlinear processing units, and

(2) supervised or unsupervised learning of feature presentations on each layer .

The early framework for deep learning was built on artificial neural networks (ANNs) in the
1980s while the real impact of deep learning became apparent in 2006. Since then, deep learning
has been applied to a wide range of fields, including automatic speech recognition, image
recognition, natural language processing, drug discovery, and bioinformatics.

The past decades have witnessed a massive growth in biomedical data, such as genomic
sequences, protein structures, and medical images, due to the advances of high-throughput
technologies. This deluge of biomedical big data necessitates effective and efficient
computational tools to store, analyze, and interpret such data . Deep learning-based algorithmic
frameworks shed light on these challenging problems. The aim of this paper is to provide the
bioinformatics and biomedical informatics community an overview of deep learning techniques
and some of the state-of-the-art applications of deep learning in the biomedical field. We hope
this paper will provide readers an overview of deep learning, and how it can be used for
analyzing biomedical data. The rapid development of knowledge in the field of advanced
materials and nanomaterials has fueled a discussion on the best means to develop this emerging
technology both safely and sustainably, without limiting the incredible potential benefits that
these advancements bring about in material design and formulation. One of the first difficulties
encountered in this domain pertains to how we organize and utilize the massive volume of
information that is being produced, in relation to the performance and environmental and health
and safety (EHS) implications of these nanoscale materials. Nanotechnology, machine learning
(ML), and artificial intelligence (AI) are a few leading technologies in this domain; although ML
and AI have recently surpassed nanotechnology in popularity, they have largely complemented
each other. We have been conditioned to expect the development of AI in a wide range of
applications such as in flying drones for home delivery, traffic routing, and small-scale robotic
assistance in performing daily chores. We are probainteracting with AI more than we realize
due to a prominent upsurge in the use of AI in electronic gadgets and digital media, and with AI
grabbing the attention of the consumer industry[3].

the present contribution provides an interdisciplinary review of the existing research from the
areas of nano-engineering, biomedical engineering and ML. To the best of the authors
knowledge no such review exists in the technical literature, that focuses on the ML-related
methodologies that are employed in nano-scale biomedical engineering.

In more detail, the contribution of this paper is as follows:

• The main challenges-problems in nano-scale biomedical engineering, which can be tackled

with ML techniques, are identified and classified based on the discipline in three main
categories: structure and material design and simulations, communications and signal processing,
and bio-medicine applications. • SOTA ML methodologies, which are used in the field of nano-
scale biomedical engineering, are reviewed, and their architectures are described. For each one of
the presented ML methods, we report its principles and building blocks. Finally, their compelling
applications in nano-scale biomedicine systems are surveyed for aiding the readers in refining
the motivation of ML in these systems, all the way from analyzing and designing new materials
and structures to holistic therapy development.

• Finally, the advantages and limitations of each ML approach are highlighted, and future
research directions are provided[4].
Nanotechnology offers promise, as a broad spectrum of highly innovative approaches emerges
for the overcoming of this challenge. Four emerging approaches are reviewed below:
nanostructured surfaces for the enhancement of proteomic analysis via mass spectrometry (MS)
and reverse-phase protein microarrays; the bio-bar code method for the amplification of protein
signatures via the use of two-particle, sandwich assay; nanowires as biologically gated
transistors, transducing molecular binding events into real-time electrical signals; and silicon
cantilevers for the mechanics-based recognition of biomolecular populations. In simple terms, AI
is a broad area of computer science that attempts to impart to machines human-like intelligence
to learn and perform the given tasks.In 1956, John McCarthy, a Dartmouth professor, first coined
the term “Artificial Intelligence” when he observed that machines can solve problems such as
understanding language semantics and forming abstractions and concepts, which were thought to
be limited to humans.McCarthy along with a group of computer scientists and mathematicians
demonstrated that machines are capable of formal reasoning using trial and error, thus paving the
way for a new era of AI over 60 years ago. Since then, AI has mostly remained limited to the
Internet, university classrooms, and exclusive labs. The timeline of advances in computer
programming indicates that a wealth of applications has been created along with uncertainties in
different areas (Figure 1). AI and ML are growing exponentially and can soon become
ubiquitous.Over the past few years, two factors have led to the skyrocketing of AI worldwide,
i.e., data availability and a faster processing capacity. The amount of data being generated is
growing exponentially, which can be seen from the fact that 90% of the data globally has been
generated over the past 2 years alone.With high processing speeds, computers can process all of
this information more quickly and effectively, thus steadily rendering AI more real than
artificial, and significantly more intelligent. In this review, we aim to address the developments
in ML implemented in theoretical approaches and simulations used in characterizing nanoscale
materials over the last decade. However, by incorporating AI into its core, the ML process has
reached an all-time high. In this article, we review ML algorithms, which are continually being
applied in new areas based on the widely distributed branches of AI, for classifying the diverse
properties of nanomaterials, as well as correlation, validation, and grouping algorithms (Figure
2)[6].
 CHAPTER 3

LITERATURE REVIEW

Nanotechnology-based platforms for the high-throughput, multiplexed detection of proteins and

nucleic acids in heretofore unattainable abundance ranges promise to bring substantial advances
in molecular medicine. The emerging approaches reviewed in this article, with reference to their
diagnostic potential, include nanotextured surfaces for proteomics, a two-particle sandwich assay
for the biological amplification of low-concentration biomolecular signals, and silicon-based
nanostructures for the transduction of molecular binding into electrical and mechanical signals,
respectively.

MACHINE LEARNING CHALLENGES IN NANO-SCALE BIOMEDICAL

ENGINEERING

1) Structure and Material Design and Simulation

1) Biological and chemical systems simulation
2) Inverse design:
3) Experiments planning and autonomous research.

2) Communications and Signal Processing

1) Channel modeling:
2) Signal detection:
3) Routing and mobility management:
4) Event detection:
5) Security.

3) Biomedicine Applications
1) Disease detection
2) Therapy development.
1) Structure and Material Design and Simulation
One of the fundamental challenges in material science and chemistry is the understanding
of the structure properties . The complexity of this problem grows dramatically in the
case of nanomaterials because:
i) they adopt different properties from their bulk components; and
ii) they are usually heterostructures, consisting of multiple materials.

As a result, the design and optimization of novel structures and materials, by discovering
their properties and behavior through simulations and experiments, lead to multi-
parameter and multi-objective problems, which in most cases are extremely difficult or
impossible to be solved through conventional approaches; ML can be an efficient
alternative choice to this challenge.

1) Biological and chemical systems simulation:

In atomic and molecular systems, there exist complex relationships between the
atomistic configuration and the chemical properties, which, in general, cannot be
described by explicit forms. In these cases, ML aims to the development of associate
configurations by means of acquiring knowledge from experimental data.
Specifically, in order to incorporate quantum effects on molecular dynamics
simulations, ML can be employed for the derivation of potential energy surfaces
(PESs) from quantum mechanic (QM) evaluations. Another use of ML lies in the
simulation of molecular dynamic trajectories. For example, the authors formulated
ML problems for discovering the optimum reaction coordinates in molecular
dynamics, whereas, the problem of estimating free energy surfaces, was reported.
Furthermore, in the ML problem of creating Markov state models, which take into
account the molecular kinetics, was investigated. Finally, the ML use in generating
samples from equilibrium distributions, that describe molecular systems, was studied
in.

2) Inverse design:
The availability of several high resolution lithographic techniques opened
the door to devising complex structures with unprecedented properties. However, the vast
choices space, which is created due to the large number of spatial degrees of freedom
complemented by the wide choice of materials, makes extremely difficult or even
impossible for conventional inverse design methodologies to ensure the existence or
uniqueness of acceptable utilizations. To address this challenge, nanoscience community
turned their eyes to ML. In more detail, several researchers identified three possible
methods, which are based on artificial neural networks (ANNs), deep neural networks
(DNNs), and generative adversarial networks (GANs). ANNs follow a trail-and-error
approach in order to design multilayer nanoparticles . Meanwhile, DNNs are used in
metasurface design . Finally, GANs can be used to design nanophotonics structures with
precise user-define spectral responses .
3) Experiments planning and autonomous research:
ML has been widely employed, in order to efficiently explore the vast parameter space
created by different combinations of nano-materials and experimental conditions and to
reduce the number of experiments needed to optimize hetero-structures and references
therein). Towards this direction, fully autonomous research can be conducted, in which
experiments can be designed based on insights extracted from data processing through
ML, without human in the loop.
B. Communications and Signal Processing In biomedical applications
nano-sensors can be utilized for a variety of tasks such as monitoring,
detection and treatment. The size of such nano-sensors ranges between 1 − 100 nm,
which refers to both macro-molecules and bio-cells . The proper selection of size and
materials is critical for the system performance, while it is constrainted by the target area,
their purpose, and safety concerns. Such nano-networks are inspired by living organisms
and, when they are injected into the human body, they interact with biological processes
in order to collect the necessary information . However, they are characterized by limited
communication range and processing power, that allow only short-range transmission
techniques to be used . As a consequence, conventional electromagneticbased
transmission schemes may not be appropriate for communications among molecules ,
since, in the latter the information is usually encoded in the number of released particles.
The simplest approach for the receiver to demodulate the symbol is to compare the
number of received particles with predetermined thresholds. In the absence of inter-
symbol interference (ISI), finding the optimal thresholds is a straight forward process.
However, in the presence of ISI the threshold needs to be extracted as a solution of the
error probability minimization problem .The aforementioned approaches require
knowledge of the channel model. However, in several practical scenarios, where the
molecular communications (MC) system complexity is high, this may not be possible. To
countermeasure this issue, ML methods can be employed to accurately model the channel
or perform data sequence detection. An alternative to MCs that has been used to support
nanonetworks is communications in the terahertz (THz) band. For these networks, apart
from their specifications, an accurate model for the THz communication between nano-
sensors is imperative for their simulation and performance assessment. In addition,
another problem that is entangled with novel nanosensor networks is their resilience
against attacks, which is of high importance since not only the system reliability is
threatened, but also the safety of the patients is at stake. Thus, it is imperative for any
possible threats to be recognized and for effective countermeasures to be developed. A
solution to the above problems appears to be relatively complex for conventional
computational methods. On the other hand, ML can provide the tools to model the space-
time trajectories of nano-sensors in the complex environments of the human body as well
as to draw strategies that mitigate the security risks of the novel network architectures.
1) Channel modeling:
One of the fundamental problems in MCs is to accurately model the channel in
different environments and conditions. Most of the MC models assume that a
molecule is removed from the environment after hitting the receiver ; hence, each
molecule can contribute to the received signal once. To model this phenomenon a
first-passage process is employed. Another approach was created from the
assumption that molecules can pass through the receiver . In this case, a molecule
contributes multiple times to the received signal. However, neither of the
aforementioned approaches are capable of modeling perfectly absorbing receivers,
when the transmitters are reflecting spherical bodies. Interistingly, such models
accommodate practical scenarios where the emitter cells do not have receptors at the
emission site and they cannot absorb the emitted molecules. An indicative example
lies in hormonal secretion in the synapses and pancreatic β−cell islets . To fill this
gap, ML was employed in , to model molecular channels in realistic scenarios, with
the aid of ANNs. Similarly, in THz nano-scale networks, where the in-body
environment is characterized by high path-loss and molecular absorption noise
(MAN), ML methods can be used in order to accurately model MAN. This opens the
road to a better understanding of the MAN’s nature and the design of new
transmission schemes and waveforms.
2) Signal detection:
To avoid channel estimation in MC, Farsal et al. proposed in a sequence
detection scheme, based on recurrent neural networks (RNNs). Compared with
previously presented ISI mitigation schemes, ML-based data sequence detection is
less complex, since they do not require to perform channel estimation and data
equalization. Following a similar approach, the authors presented an ANN capable of
achieving the same performance as conventional detection techniques, that require
perfect knowledge of the channel. In THz nano-scale networks, an energy detector is
usually used to estimate the received data . In more detail, if the received signal
power is below a predefined threshold, the detector decides that the bit 0 has been
sent, otherwise, it decides that 1 is sent. However, the transmission of 1 causes a
MAN power increase, usually capable of affecting the detection of the next symbols.
To counterbalance this, without increasing the symbol duration, a possible approach
is to design ML algorithms that are trained to detect the next symbol and take into
account the already estimated ones. Another ML challenge in signal detection at THz
nano-scale networks, lies with detecting the modulation mode of the transmission
signal by a receiver, when no prior synchronization between transmitter and receiver
has occurred. The solution to this problem will provide scalability to these networks.
Motivated by this, in, the authors provided a ML algorithm for modulation
recognition and classification.
3) Routing and mobility management:
 In THz nano-scale networks, the design of routing protocols capable of proactively
countermeasuring congestion has been identified as the next step for their utilization .
These protocols need to take into account the extremely constrained computational
resources, the stochastic nature of nano-nodes movements as well as the existence of
obstacles that may interrupt the line-of-sight transmission. The aforementioned
challenges can be faced by employing SOTA ML techniques for analyzing collected
data and modeling the nano-sensors’ movements, discovering neighbors that can be
used as intermediate nodes, identifying possible blockers, and proactively
determining the message root from the source to the final destination.In this context,
in the authors presented a multi-hop deflection routing algorithm based on
reinforcement learning and analyzed its performance in comparison to different
neural networks (NNs) and decision tree updating policies.

4) Event detection:

Nano-sensor biomedicine networks can provide continuous monitoring solutions,

that can be used as compact, accurate, and portable diagnostic systems. Each nano-
sensor obtains a biological signal linked to a specific disease and is used for
detecting physiological change or various biological materials . Successful
applications in event detection include monitoring of DNA interactions, antibody,
and enzymatic interactions, or cellular communication processes, and are able to
detect viruses, asthma attacks and lung cancer . For example, the authors developed a
bio-transferrable graphene wireless nano-sensor that is able to sense extremely
sensitive chemicals and biological compounds up to single bacterium. Furthermore, a
Sandwich Assay was developed that combines mechanical and optoplasmonic
transduction in order to detect cancer biomarkers at extremely low concentrations.
Also, in a molecular communication-based event detection network was proposed,
that is able to cope with scenarios where the molecules propagate according to
anomalous diffusion instead of the conventional Brownian motion.
5) Security:
Although, the emergence of nano-scale networks based on both electromagnetic
and MCs opened opportunities for the development of novel healthcare applications, it
also generated new problems concerning the patients’ safety. In particular, two types of
security risks have been observed, namely blackhole and sentry attacks . In the former,
malicious nano-sensors emit chemicals to attract the legitimate ones and prevent them
from searching for their target. On the contrary, in the latter, the malicious nano-sensors
repel the legitimate ones for the same reason. Such security risks can be counterbalanced
with the use of threshold-based and bayesian ML techniques that have been proven to
counter the threats with minimal requirements. C. Biomedicine Applications Timely
detection and intervention are tied with successful treatment for many diseases. This is
the so-called proactive treatment and is one of the main objectives of the next generation
healthcare systems, in order to detect and predict diseases and offer treatment services
 seamlessly. Data analysis and nanotechnology progress simultaneously toward the
realization of these systems[5].
Recent breakthroughs in nanotechnology-enabled healthcare systems allow for the
exploitation of not only the data that already exist in medical databases throughout the
world, but also of the data gathered from millions of nano-sensors.
1) Disease detection:
One of the most common problems in healthcare systems is genome
classification, with cancer detection being the most popular. Various classification
algorithms are suitable for tackling this problem, such as Naive Bayes, k-Nearest
Neighbors, Decision tree, ANNs and support vector machine (SVM) . For example,
the authors in, predicted the risk of cerebral infarction in patients by using
demographic and cerebral infarction data. In addition, in a unique coarse-to-fine
learning method was applied on genome data to identify gastric cancer. Another
example is the research presented in , where SVM and convolution NNs (CNNs)
were used to classify breast cancer subcategory by performing analysis on
microscopic images of biopsy.
2) Therapy development: Therapy development and optimization can improve clinical
efficacy of treatment for various diseases, without generating unwanted outcomes.
Optimization still remains a challenging task, due to its requirement for selecting the right
combination of drugs, dose and dosing frequency. For instance, a quadratic phenotype
optimization platform (QPOP) was proposed in to determine the optimal combination
from 114 drugs to treat bortezomib-resistant multiple myeloma. Since its creation, QPOP
has been used to surpass the problems related to drug designing and optimization, as well
as drug combinations and dosing strategies. Also, the authors presented a platform called
CURATE.AI, which was validated clinically and was used to standardize therapy of
tuberculosis patients with liver transplant-related immunosuppression. Furthermore,
CURATE.AI was used for treatment development and patient guidance that resulted in
halted progression of metastatic castration resistant prostate cancer.
Applications in biomedicine Owing to advances in high-throughput technologies
a deluge of biological and medical data has been obtained in recent decades, including
data related to medical images, biological sequences, and protein structures. Some successful
applications of deep learning in biomedical fields are reviewed in this section and a summary of
application Medical image classification and segmentation Machine learning for medical images
has long been a powerful tool in the diagnosis or assessment of diseases. Traditionally,
discriminative features referring to medical image interpretation are manually designed for
classification and segmentation of regions of interest in different medical applications. This
requires the participation of physicians with expertise. Nonetheless, the complexity and 13
ambiguity of medical images, limited knowledge for medical image interpretation, and the
requirement of large amounts of annotated data have hindered the wide use of machine learning
in the medical image domain. Notably, deep learning methods have attained success in a variety
of computer vision tasks such as object recognition, localization, and segmentation in natural
images. These have soon brought about an active field of machine learning in medical image
analysis. Segmentation of tissues and organs is crucial for qualitative and quantitative assessment
of medical images. Pereira et al. used data augmentation, small convolutional kernels, and a
preprocessing stage to achieve accurate brain tumor segmentation . Their CNN-based
segmentation method won first place in the Brain Tumor Segmentation (BRATS) Challenge in
2013, and second place in 2015. Havaei et al. presented a fully automatic brain tumor
segmentation method based on DNNs in magnetic resonance (MR) images with a two-phase
training procedure , which obtained second place in the 2013 BRATS. Their methodology was
tested on the publicly available datasets INbreast and Digital Database for Screening
Mammography (DDSM) , outperforming in terms of accuracy and efficiency several stateof-the-
art methods when tested on DDSM. Additional medical applications employing a deep learning
architecture have been demonstrated in segmenting the left ventricle of the heart from the MR
data, the pancreas through computed tomography (CT) , tibial cartilage through magnetic
resonance imaging (MRI), the prostate through MRI, and the hippocampus through MR brain
images . The differentiation of tissues or organs in medical images has been termed semantic
segmentation in which each pixel of an image is assigned to a class or a label. The skeletal
muscles, organs, and fat in CT images are well delineated through semantic segmentation based
on a DNN architecture . Similarly, the semantic segmentation of MR images also attained
accurate segmentation results. Detection of lesion and abnormality is the major issue in medical
image analysis. Deep learning methods learn the representations directly instead of using hand-
crafted features from training data. A classifier is then used to assign the representations to a
probability that indicates whether or not the image contains lesions.
In other words, the deep learning schemas classify each pixel to be a lesion point or not, which
can be done in two ways:
(1) classifying the mini patch around the pixel with a deep network, and
(2) using a fully convolutional network to classify each pixel.
14 Debdoot et al. applied a DNN to histologically characterize healthy skin and healing wounds
to reduce clinical reporting variability. Two unsupervised pre-trained layers of denoising AEs
(DAEs) were used to learn features in their hybrid architecture, and subsequently the whole
network was learned using labelled tissues for characterization. Detection of cerebral
microbleeds and coronary artery calcification also produced better results when using deep
learning-based approaches. In addition, brain tumor progression prediction implemented with a
deep learning architecture has also shown a more robust tumor progression model in comparison
with a high-precision manifold learning approach . Detection of pathologies on stained
histopathology images exemplify the high precision of deep learning-based approaches. For
breast cancer detection in histopathology images, Cruz-Roa et al. established a deep learning
model to precisely delineate the invasive ductal carcinoma (IDC) regions to distinguish the
invasive tumor tissue and noninvasive or healthy tissue. Their 3-layer CNN architecture,
composed of two cascading convolutional and pooling layers, a full-connected layer, and a
logistic regression classifier for prediction, attained a better F-measure (71.8%) and higher
balanced accuracy (BAC; 84.23%) in comparison with an approach using handcrafted image
features and a machine learning classifier. The mammogram is one of the most effective imaging
modalities in early diagnosis and risk prediction of breast cancer. A deep learning model trained
on a large dataset of 45,000 images attained performance similar to that of certified screening
radiologists in mammographic lesion detection. Kallenberg et al.investigated the scoring of
percentage mammographic density (PMD) and mammographic texture (MT) related to
prediction of breast cancer risk. They employed a sparse AE to learn deep hierarchical features
from unlabeled mammograms. Multinomial logistic regression or softmax regression was then
used as a classifier in the supervised training. As a result, the performance of their approach was
comparable with that of the subjective and expensive manual PMD and MT scorings. Color
fundus photography is an important diagnostic tool for ophthalmic diseases. Deep learning-based
methods with fundus images have recently gained considerable interest as a key to developing
automated diagnosis systems. A DNN architecture was proposed by Srivastava et al. to
distinguish optic disc (OD) from parapapillary atrophy (PPA). A DNN consisting of SAEs
followed by a refined active shape model attained accurate OD segmentation. For image
registration, deep learning in combination with a multi-scale Hessian matrix was used to 15
detect vessel landmarks in the retinal image, whereas convolutional neural networks have also
produced excellent results in the detection of hemorrhages and exudates in color fundus images.
It is difficult to design an automatic screening system for retinal-based diseases such as age-
related molecular degeneration, diabetic retinopathy, retinoblastoma, retinal detachment, and
retinitis pigmentosa, because these diseases share similar characteristics. Through deep learning
methods, Arunkumar et al. successfully built a system to discriminate retina-based diseases only
using fundus images. First, a DBN composed of a stack of RBMs was designed for feature
extraction. Then a generalized regression neural network (GRNN) was employed to reduce
dimensionality. Finally, a multi-class SVM was used for classification. Interestingly, Kaggle
organized a competition on the staging of diabetic retinopathy from 35,126 training and 53,576
test color fundus images in 2015. Using convolutional neural networks, the top model
outperformed other machine learning methods with a kappa score of 0.8496 . In addition to static
images, time-series medical records such as signal maps from electroencephalography and
magnetoencephalography can also be analyzed using deep learning methods . These deep
learning schemas take coded features of signals or raw signals as input, and extract features
from the data for anomaly classification or understanding emotions. All the aforementioned
applications illustrate that as a frontier of machine learning, deep learning has made substantial
progress in medical image segmentation and classification. We expect that more clinical trials
and systematic medical image analytic applications will emerge to help achieve better
performance when applying deep learning in medicine[3].

Artificial intelligence applications in healthcare :

It is generally believed that AI tools will facilitate and enhance human work and not replace the
work of physicians and other healthcare staff as such. AI is ready to support healthcare personnel
with a variety of tasks from administrative workflow to clinical documentation and patient
outreach as well as specialized support such as in image analysis, medical device automation,
and patient monitoring. There are different opinions on the most beneficial applications of AI for
healthcare purposes. Forbes stated in 2018 that the most important areas would be administrative
workflows, image analysis, robotic surgery, virtual assistants, and clinical decision support . A
2018 report by Accenture mentioned the same areas and also included connected The rise of
artificial intelligence in healthcare applications 27 machines, dosage error reduction, and
cybersecurity . A 2019 report from McKinsey states important areas being connected and
cognitive devices, targeted and personalized medicine, robotics-assisted surgery, and
electroceuticals [10]. In the next sections, some of the major applications of AI in healthcare will
be discussed covering both the applications that are directly associated with healthcare and other
applications in the healthcare value chain such as drug development and ambient assisted living
(AAL).

2.Precision medicine

Precision medicine provides the possibility of tailoring healthcare interventions to individuals or

groups of patients based on their disease profile, diagnostic or prognostic information, or their
treatment response. The tailor-made treatment opportunity will take into consideration the
genomic variations as well as contributing factors of medical treatment such as age, gender,
geography, race, family history, immune profile, metabolic profile, microbiome, and
environment vulnerability. The objective of precision medicine is to use individual biology
rather than population biology at all stages of a patient’s medical journey. This means collecting
data from individuals such as genetic information, physiological monitoring data, or EMR data
and tailoring their treatment based on advanced models. Advantages of precision medicine
include reduced healthcare costs, reduction in adverse drug response, and enhancing effectivity
of drug action . Innovation in precision medicine is expected to provide great benefits to patients
and change the way health services are delivered and evaluated. There are many types of
precision medicine initiatives and overall, they can be divided into three types of clinical areas:
complex algorithms, digital health applications, and “omics”-based tests. Complex algorithms:
Machine learning algorithms are used with large datasets such as genetic information,
demographic data, or electronic health records to provide prediction of prognosis and optimal
treatment strategy. Digital health applications: Healthcare apps record and process data added by
patients such as food intake, emotional state or activity, and health monitoring data from
wearables, mobile sensors, and the likes. Some of these apps fall under precision medicine and
use machine learning algorithms to find trends in the data and make better predictions and give
personalized treatment advice. 28 Artificial Intelligence in Healthcare Omics-based tests:
Genetic information from a population pool is used with machine learning algorithms to find
correlations and predict treatment responses for the individual patient. In addition to genetic
information, other biomarkers such as protein expression, gut microbiome, and metabolic profile
are also employed with machine learning to enable personalized treatments . Here, we explore
selected therapeutic applications of AI including genetics-based solutions and drug discovery.

2.1 Genetics-based solutions

It is believed that within the next decade a large part of the global population will be offered full
genome sequencing either at birth or in adult life. Such genome sequencing is estimated to take
up 100150 GB of data and will allow a great tool for precision medicine. Interfacing the genomic
and phenotype information is still ongoing. The current clinical system would need a redesign to
be able to use such genomics data and the benefits hereof . Deep Genomics, a Healthtech
company, is looking at identifying patterns in the vast genetic dataset as well as EMRs, in order
to link the two with regard to disease markers. This company uses these correlations to identify
therapeutics targets, either existing therapeutic targets or new therapeutic candidates with the
purpose of developing individualized genetic medicines. They use AI in every step of their drug
discovery and development process including target discovery, lead optimization, toxicity
assessment, and innovative trial design. Many inherited diseases result in symptoms without a
specific diagnosis and while interpreting whole genome data is still challenging due to the many
genetic profiles. Precision medicine can allow methods to improve identification of genetic
mutations based on full genome sequencing and the use of AI.

2.2 Drug discovery and development

Drug discovery and development is an immensely long, costly, and complex process that can
often take more than 10 years from identification of molecular targets until a drug product is
approved and marketed. Any failure during this process has a large financial impact, and in fact
most drug candidates fail sometime during development and never make it onto the market. On
top of that are the ever-increasing regulatory obstacles and The rise of artificial intelligence in
healthcare applications 29 the difficulties in continuously discovering drug molecules that are
substantially better than what is currently marketed. This makes the drug innovation process both
challenging and inefficient with a high price tag on any new drug products that make it onto the
market . There has been a substantial increase in the amount of data available assessing drug
compound activity and biomedical data in the past few years. This is due to the increasing
automation and the introduction of new experimental techniques including hidden Markov model
based text to speech synthesis and parallel synthesis. However, mining of the largescale
chemistry data is needed to efficiently classify potential drug compounds and machine learning
techniques have shown great potential [15]. Methods such as support vector machines, neural
networks, and random forest have all been used to develop models to aid drug discovery since
the 1990s. More recently, DL has begun to be implemented due to the increased amount of data
and the continuous improvements in computing power. There are various tasks in the drug
discovery process where machine learning can be used to streamline the tasks. This includes
drug compound property and activity prediction, de novo design of drug compounds,
drugreceptor interactions, and drug reaction prediction [16]. The drug molecules and the
associated features used in the in silico models are transformed into vector format so they can be
read by the learning systems. Generally, the data used here include molecular descriptors (e.g.,
physicochemical properties) and molecular fingerprints (molecular structure) as well as
simplified molecular input line entry system (SMILES) strings and grids for convolutional neural
networks (CNNs)[7].
Figure 1. Timeline of AI and ML in nanomaterial development. Evolution timeline for both
the development of nanoparticles (NPs), starting after the first synthesis and quantum
effects as observed in 1853 by Faraday, and AI including statistical approaches. In 2010,
both timelines merged when AI was applied in tasks such as the identification of NP
properties or interaction partners, the grouping of NPs depending on their properties or
toxic effects, and the prediction of NP toxicity.

Figure 2. Interconnected algorithm network for exploring advanced material design. A

blueprint for AI and other important ML approaches used in nanotoxicology for the
prediction of safer nanomaterials.
 CHAPTER-4

CASE STUDY

Epileptic seizure prediction Epilepsy, a neurodegenerative disease, is one of the most common
neurological conditions and is characterized by spontaneous, unpredictable, and recurrent
seizures. While first lines of treatment consist of long-term medications-based therapy, more
than one third of patients are refractory. On the other hand, recourse to epilepsy surgery is still
relatively low due to very modest success rates and fear of complications. An interesting
research direction is to explore the possibility of predicting seizures, which, if made possible,
could result in the development of alternative interventional strategies. Although early seizure-
forecasting investigations date back to the 1970s, the limited number of seizure events, the
paucity of intracranial electroencephalography recordings, and the limited extent of interictal
epochs have been major hurdles toward an adequate evaluation of seizure prediction
performances. Interestingly, signals acquired from naturally epileptic canines implanted with the
ambulatory monitoring device have been made accessible through the ieeg.org online portal.
However, the seizure onset zone was not disclosed/ available. Our group investigated the
possibility of forecasting seizures using the aforementioned canine data. Subsequently, we
performed a directed transfer function (DTF)-based, quantitative identification of electrodes
located within the epileptic network. A genetic algorithm was employed to select the features
most discriminative of the preictal state. We proposed a new fitness function that is insensitive to
skewed data distributions. An average sensitivity of 84.82% at a time-in-warning of 10% was
reported on the held-out dataset, improving previous seizure prediction performances. Trying to
find new opportunities for seizure prediction, we also explored novel features to track the
preictal state based on higher order spectral analysis. Extracted features were then used as inputs
to a multilayer perceptron for classification. Our preliminary findings revealed significant
differences between interictal and preictal states using each of three bispectrum-extracted
characteristics (p < 0.05). Test accuracies of 73.26%, 72.64%, and 78.11% were achieved for the
mean of magnitudes, normalized bispectral entropy, and normalized squared entropy,
respectively. In addition, we demonstrated the existence of consistent differences between the
epileptic preictal and interictal states in mean phase– amplitude coupling on the same bilateral
canine iEEG recordings. In contrast, we also explored the possibility of using quantitative
effective connectivity measures to determine the network seizure activity in high-density
recordings. The ability of the DTF to quantify causal relations between iEEG recordings has
been previously validated. However, quasi-stationarity of the analyzed signals remains a must to
avoid spurious connections between iEEG contacts . Although the identification of stationary
epochs is possible when dealing with a relatively small number of contacts, it becomes more
challenging when analyzing highdensity iEEG signals. Recently, a time-varying version of the
DTF was proposed: the spectrum-weighted adaptive directed transfer function (swADTF). The
swADTF is able to cope with nonstationarity issues and automatically identify frequency ranges
of interest. Subsequently, we validated the possibility of finding seizure activity generators and
sinks by employing the swADTF on high-density recordings. The database consisted of patients
with refractory epilepsy admitted for pre-surgical evaluation at the University of Montreal
Hospital Center. Interestingly, the identified seizure activity sources were within the epileptic
focus and resected volume for patients who went seizure-free after surgical resection. In contrast,
additional or different generators were identified in non-seizure-free patients. Our findings
highlighted the feasibility of accurately identifying seizure generators and sinks using the
swADTF. Electrode selection methods based on effective connectivity measures are thus
recommended in future seizureforecasting investigations. Recent findings highlight the
feasibility of predicting seizures using iEEG recordings; the transition from interictal into ictal
states consists of a ‘‘buildup” that can be tracked using advanced feature extraction and AI
techniques. Nevertheless, before current approaches can be translated into actual clinical devices,
further research is needed on feature extraction, electrode selection, hardware implementation,
and deep learning algorithm.
 CONCLUSION

In this paper, we reviewed the latest developments in the application of AI in biomedicine,

including disease diagnostics and prediction, living assistance, biomedical information
processing, and biomedical research. AI has interesting applications in many other biomedical
areas as well. It can be seen that AI plays an increasingly important role in biomedicine, not only
because of the continuous progress of AI itself, but also because of the innate complex nature of
biomedical problems and the suitability of AI to solve such problems. New AI capabilities
provide novel solutions for biomedicine, and the development of biomedicine demands new
levels of capability from AI. This match of supply and demand and coupled developments will
enable both fields to advance significantly in the foreseeable future, which will ultimately benefit
the quality of life of people in need[5]. In summary, in this article, we have reviewed how ML
algorithms bear fruits in nano-scale biomedical engineering. In more detail, we presented the
main challenges and problems in this field, which can be solved through ML, and classified
them, based on their discipline, into three distinctive categories. For each category, we have
provided insightful discussions that revealed its particularities as well as existing research gaps.
Moreover, we have surveyed a variate of SOTA ML methodologies and models, which have
been used as countermeasures to the aforementioned challenges. Special attention was payed to
the ML methodologies architecture, operating principle, advantages and limitations. Finally,
future research directions have been provided, which highlight the need of thorough
interdisciplinary research efforts for the successful realization of hitherto uncharted scenarios
and applications in the nano-scale biomedical engineering field[10].
 REFERNCES

1. Michele Greque de Morais , Vilásia Guimarães Martins, Daniela Steffens, Patricia Pranke, and
Jorge Alberto Vieira da Costa Journal of Nanoscience and Nanotechnology 2014, Vol. 14, 1007–
1017.
2. Guoguang Rong , Arnaldo Mendez , Elie Bou Assi c , Bo Zhao d , Mohamad Sawan Engineering
2020,Vol. 6,291-301.
3. Chensi Cao, Feng Liu, Hai Tan, Deshou Song, Wenjie Shu, Weizhong Li, Yiming Zhou, Xiaochen
Bo, Zhi Xie Genomics, Proteomics & Bioinformatics 2018, 0229(18)30002-0.
4. Ajay Vikram Singh,* Daniel Rosenkranz, Mohammad Hasan Dad Ansari, Rishabh Singh, Anurag
Kanase, Shubham Pratap Singh, Blair Johnston, Jutta Tentschert, Peter Laux, and Andreas Luch
Reviwe On Artificial Intelligence and Machine Learning Empower Advanced Biomedical Material
Design to Toxicity Prediction.
5. Alexandros–Apostolos A. Boulogeorgos, Senior Member, IEEE, Stylianos E. Trevlakis, Student
Member, IEEE, Sotiris A. Tegos, Student Member, IEEE, Vasilis K. Papanikolaou, Student
Member, IEEE, and George K. Karagiannidis, Fellow, 2020.
6. Mark Ming-Cheng Cheng1,* , Giovanni Cuda2,* , Yuri L Bunimovich3 , Marco Gaspari2 , James R
Heath3 , Haley D Hill4 , Chad A Mirkin4 , A Jasper Nijdam1 , Rosa Terracciano2 , Thomas
Thundat5 and Mauro Ferrari Nanotechnologies for biomolecular detection and medical
diagnostics
7. Weng J, McClelland J, Pentland A, Sporns O, Stockman I, Sur M, et al. Autonomous
mental development by robots and animals. Science 2001;291 (5504):599–60
8. Wooldridge M, Jennings NR. Intelligent agents: theory and practice. Knowl Eng Rev
1995;10(2):115–52.
9. Huang G, Huang GB, Song S, You K. Trends in extreme learning machines: a review.
Neural Netw 2015;61:32–48.
10. Hopfield JJ. Neural networks and physical systems with emergent collective
computational abilities. Proc Natl Acad Sci USA 1982;79(8):2554–8.
11. Watts DJ, Strogatz SH. Collective dynamics of ‘small-world’ networks.
Nature1998;393(6684):440–2.
12. Zucker RS, Regehr WG. Short-term synaptic plasticity. Annu Rev Physiol

2002;64:355–405.
13. Schmidhuber J. Deep learning in neural networks: an overview. Neural

Netw2015;61:85–117.
14. LeCun Y, Bengio Y, Hinton G. Deep learning. Nature 2015;521(7553):436–44.
15. Arel I, Rose DC, Karnowski TP. Deep machine learning—a new frontier in

artificial intelligence research. IEEE Comput Intell Mag 2010;5(4):13–8.