Artificial Intelligence in Label free

Microscopy Biological Cell

Classification by Time Stretch 1st
Edition Ata Mahjoubfar
Visit to download the full and correct content document:
https://textbookfull.com/product/artificial-intelligence-in-label-free-microscopy-biologic
al-cell-classification-by-time-stretch-1st-edition-ata-mahjoubfar/
More products digital (pdf, epub, mobi) instant
download maybe you interests ...

Label free and multi parametric monitoring of cell

based assays with substrate embedded sensors
Oberleitner

https://textbookfull.com/product/label-free-and-multi-parametric-
monitoring-of-cell-based-assays-with-substrate-embedded-sensors-
oberleitner/

Artificial Intelligence By Example Develop machine

intelligence from scratch using real artificial
intelligence use cases Denis Rothman

https://textbookfull.com/product/artificial-intelligence-by-
example-develop-machine-intelligence-from-scratch-using-real-
artificial-intelligence-use-cases-denis-rothman/

Atomic Force Microscopy in Molecular and Cell Biology

Jiye Cai

https://textbookfull.com/product/atomic-force-microscopy-in-
molecular-and-cell-biology-jiye-cai/

Real time Speech and Music Classification by Large

Audio Feature Space Extraction 1st Edition Florian
Eyben (Auth.)

https://textbookfull.com/product/real-time-speech-and-music-
classification-by-large-audio-feature-space-extraction-1st-
edition-florian-eyben-auth/
Intelligence is not Artificial A History of Artificial
Intelligence and Why the Singularity is not Coming any
Time Soon Expanded Edition Piero Scaruffi

https://textbookfull.com/product/intelligence-is-not-artificial-
a-history-of-artificial-intelligence-and-why-the-singularity-is-
not-coming-any-time-soon-expanded-edition-piero-scaruffi/

Biological classification a philosophical introduction

1st Edition Richard A. Richards

https://textbookfull.com/product/biological-classification-a-
philosophical-introduction-1st-edition-richard-a-richards/

Teaching music to students with special needs : a

label-free approach Second Edition Hammel

https://textbookfull.com/product/teaching-music-to-students-with-
special-needs-a-label-free-approach-second-edition-hammel/

Fluorescence Microscopy From Principles to Biological

Applications 2nd Edition Ulrich Kubitscheck

https://textbookfull.com/product/fluorescence-microscopy-from-
principles-to-biological-applications-2nd-edition-ulrich-
kubitscheck/

Artificial Intelligence in Healthcare 1st Edition Adam

Bohr (Editor)

https://textbookfull.com/product/artificial-intelligence-in-
healthcare-1st-edition-adam-bohr-editor/
Ata Mahjoubfar · Claire Lifan Chen
Bahram Jalali

Artificial
Intelligence
in Label-free
Microscopy
Biological Cell Classification by Time
Stretch
Artificial Intelligence in Label-free Microscopy
Ata Mahjoubfar • Claire Lifan Chen • Bahram Jalali

Artificial Intelligence
in Label-free Microscopy
Biological Cell Classification by Time Stretch

123
Ata Mahjoubfar Claire Lifan Chen
University of California Los Angeles University of California Los Angeles
Los Angeles, CA, USA Los Angeles, CA, USA

Bahram Jalali
University of California Los Angeles
Los Angeles, CA, USA

ISBN 978-3-319-51447-5 ISBN 978-3-319-51448-2 (eBook)

DOI 10.1007/978-3-319-51448-2

Library of Congress Control Number: 2017937310

© Springer International Publishing AG 2017

This work is subject to copyright. All rights are reserved 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. The publisher remains neutral with regard to jurisdictional
claims in published maps and institutional affiliations.

Printed on acid-free paper

This Springer imprint is published by Springer Nature

The registered company is Springer International Publishing AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Preface

High-throughput real-time imaging and vision systems for capture and identification
of fast phenomena are among the most essential tools for scientific, industrial,
military, and most importantly biomedical applications. The key challenge in these
instruments is the fundamental trade-off between speed and sensitivity of the
measurement system due to the limited signal energy collected in each measurement
window. Based on two enabling technologies, namely, photonic time stretch and
optical amplification, several novel high-throughput optical measurement tools are
recently developed for applications such as volumetric scanning, vibrometry, and
flow cytometry. Here, we introduce time stretch imaging, a high-content computer
vision system developed for big data acquisition and analysis of images. Time
stretch imaging is able to capture quantitative optical phase and intensity images
simultaneously, enabling accurate surface inspection, volumetric scans, defect
detection, cell analysis, and cancer diagnostics.
We further describe a complete artificial intelligence pipeline for time stretch
microscopy that performs optical phase measurement, image processing, feature
extraction, and classification. Multiple biophysical features such as morphological
parameters, optical loss characteristics, and protein concentration are measured on
individual biological cells. These biophysical measurements form a hyperdimen-
sional feature space in which supervised learning is performed for cell classification.
The technology is in clinical testing for blood screening and circulating tumor
cell detection, as well as studying lipid-accumulating algal strains for biofuel
production. By integrating machine learning with high-throughput quantitative
imaging, this system achieves record-high accuracy in label-free cellular phenotypic
screening and opens up a new path to data-driven diagnosis.
Furthermore, we explained a real-time image compression technique performed
in the optical domain to solve the big data challenge created by ultrafast mea-
surement systems. Many ultrafast and high-throughput data acquisition equipment,
including time stretch imaging, produce a torrent of data in a short time, e.g.,
several gigabytes per second. Such a data volume and velocity place a burden on
data acquisition, storage, and processing and call for technologies that compress
images in optical domain and in real-time. As a solution, we have experimentally

v
vi Preface

demonstrated warped time stretch, which offers variable spectral-domain sampling

rate, as well as the ability to engineer the time-bandwidth product of the signal’s
envelope to match that of the data acquisition systems. We also show how to
design the kernel of the transform and, specifically, the nonlinear group delay
profile governed by the signal sparsity. Such a kernel leads to smart detection
with nonuniform spectral resolution, having direct utility in improvement of data
acquisition rate, real-time data compression, and enhancement of ultrafast data
capture accuracy.

Los Angeles, CA, USA Ata Mahjoubfar

Claire Lifan Chen
Bahram Jalali
Acknowledgements

We would like to thank all of our publication coauthors, which our collaborative
works consist many chapters of this book. Chapter 2 is a version of IEEE 49th
Annual Conference on Information Sciences and Systems (CISS) Proceedings, pp.
7086896, 2015 [1]. Chapter 3 is a version of Applied Physics Letters, Vol. 98, No.
10, 101107, 2011 [12]. Chapter 4 is a version of Proceedings of SPIE, Frontiers
in Ultrafast Optics: Biomedical, Scientific, and Industrial Applications XIII, San
Francisco, CA, 2013, 86110N [16]. Chapter 5 is a version of Biomedical Optics
Express, Vol. 4, No. 9, pp. 1618–1625, 2013 [17]. Chapters 6 and 8 are based on
Scientific Reports, Vol. 6, pp. 21471, 2016 [3]. Chapter 7 is a version of Proceedings
of SPIE, High-Speed Biomedical Imaging and Spectroscopy: Toward Big Data
Instrumentation and Management II, San Francisco, CA, 2017, 100760J [2].
Chapter 9 is a version of PLoS ONE, Vol. 10, No. 4, pp. e0125106, 2015 [104].
Finally, Chapter 10 is a version of Scientific Reports, Vol. 5, pp. 17148, 2015 [4].

vii
Contents

Part I Time Stretch Imaging

1 Introduction .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 3
2 Time Stretch.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 7
2.1 Time Stretch Imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 7
2.2 Cell Classification Using Time Stretch Imaging .. . . . . . . . . . . . . . . . . . . 9
2.3 Label-Free Phenotypic Screening . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 10
2.4 Warped Time Stretch for Data Compression .. . .. . . . . . . . . . . . . . . . . . . . 10

Part II Inspection and Vision

3 Nanometer-Resolved Imaging Vibrometer . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 15
3.1 Introduction .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 15
3.2 Experimental Demonstration . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 16
3.3 Theoretical Study of the Vibrometer Performance . . . . . . . . . . . . . . . . . 18
3.4 Experimental Results. . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 18
3.5 Conclusion .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 19
4 Three-Dimensional Ultrafast Laser Scanner . . . . . . . .. . . . . . . . . . . . . . . . . . . . 21
4.1 Introduction .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 21
4.2 Principle of Hybrid Dispersion Laser Scanner ... . . . . . . . . . . . . . . . . . . . 22
4.3 Applications of Hybrid Dispersion Laser Scanner.. . . . . . . . . . . . . . . . . 23

Part III Biomedical Applications

5 Label-Free High-Throughput Phenotypic Screening .. . . . . . . . . . . . . . . . . . 33
5.1 Introduction .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 33
5.2 Experimental Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 35
5.3 Results and Discussion .. . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 38
5.4 Conclusion .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 41

ix
x Contents

6 Time Stretch Quantitative Phase Imaging .. . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 43

6.1 Background .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 43
6.2 Time Stretch Quantitative Phase Imaging .. . . . . .. . . . . . . . . . . . . . . . . . . . 45
6.2.1 Overview .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 45
6.2.2 Imaging System. . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 47
6.2.3 System Performance and Resolvable Points . . . . . . . . . . . . . . 49
6.2.4 Microfluidic Channel Design and Fabrication . . . . . . . . . . . . 49
6.2.5 Coherent Detection and Phase Extraction . . . . . . . . . . . . . . . . . 51
6.2.6 Cell Transmittance Extraction . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 56
6.2.7 Image Reconstruction . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 57
6.3 Image Processing Pipeline . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 59
6.3.1 Feature Extraction . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 59
6.3.2 Multivariate Features Enabled by Sensor Fusion . . . . . . . . . 61
6.3.3 System Calibration . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 63
6.4 Conclusion .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 63

Part IV Big Data and Artificial Intelligence

7 Big Data Acquisition and Processing in Real-Time .. . . . . . . . . . . . . . . . . . . . 67
7.1 Introduction .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 67
7.2 Technical Description of the Acquisition System . . . . . . . . . . . . . . . . . . 68
7.3 Big Data Acquisition Results . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 70
7.4 Conclusion .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 71
8 Deep Learning and Classification .. . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 73
8.1 Background .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 73
8.2 Machine Learning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 74
8.3 Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 76
8.3.1 Blood Screening: Demonstration in Classification
of OT-II and SW-480 Cells . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 76
8.3.2 Biofuel: Demonstration in Algae Lipid Content
Classification. . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 77
8.4 Further Discussions in Machine Learning . . . . . .. . . . . . . . . . . . . . . . . . . . 80
8.4.1 Learning Curves . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 80
8.4.2 Principal Component Analysis (PCA) .. . . . . . . . . . . . . . . . . . . . 80
8.4.3 Cross Validation . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 83
8.4.4 Computation Time . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 84
8.4.5 Data Cleaning .. . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 84
8.5 Conclusion .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 85

Part V Data Compression

9 Optical Data Compression in Time Stretch Imaging . . . . . . . . . . . . . . . . . . . 89
9.1 Background .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 89
9.2 Warped Stretch Imaging . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 90
9.3 Optical Image Compression . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 91
Contents xi

9.4 Experimental Design and Results . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 95

9.5 Conclusion .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 99
10 Design of Warped Stretch Transform . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 101
10.1 Overview.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 101
10.2 Kernel Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 104
10.2.1 Spectral Resolution .. . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 106
10.2.2 Group Delay Profile Design .. . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 110
10.2.3 Simulation Model.. . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 114
10.2.4 Spectrograms . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 116
10.3 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 116
10.4 Conclusion .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 118
11 Concluding Remarks and Future Work . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 121

References .. .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 123

Index . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 131
List of Figures

Fig. 2.1 Operating principle of time stretch imaging. The key

innovations in STEAM that enable high-speed real-time
imaging are photonic time stretch for digitizing fast
images in real-time and the optical image amplification for
compensating the low number of photons collected during
the ultra-short shutter time . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 9
Fig. 2.2 Illustration of warped-stretch transform in imaging.
(a) In conventional time stretch imaging (STEAM),
the spectrum-to-time mapping is uniform, and the
pixels are assigned equally distanced across the field of
view. (b) Using a nonlinear group delay profile in the
spectrum-to-time mapping process results in a nonuniform
sampling of the line image, assigning more pixels to the
information-rich central part of field-of-view and less to
the low-entropy peripherals . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 11
Fig. 3.1 Schematic of the STEAM vibrometer. The principle
of the method is three-fold: (1) encoding of the lateral
and axial coordinates of the target into the different
frequencies and corresponding amplitudes of a spatially
dispersed broadband pulse which spectrally interferes
with a reference pulse, (2) amplified dispersive Fourier
transformation in which the spectrum is mapped into
a temporal waveform, time stretched so that it can be
digitized in real time, and simultaneously amplified in
the optical domain, and (3) Hilbert transformation on the
detected pulse in the digital domain to extract the axial
information of the target . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 17

xiii
xiv List of Figures

Fig. 3.2 Basic performance of the STEAM vibrometer. (a)

Temporal waveform of a single interfered pulse captured
by the photodiode in comparison with the optical spectrum
measured by a conventional optical spectrum analyzer. (b)
Repetitive pulses (scans) with a time interval of 27.2 ns
detected by the photodiode indicating that the STEAM
vibrometer operates at 36.7 MHz scan rate . . . .. . . . . . . . . . . . . . . . . . . . 19
Fig. 3.3 Surface vibration of the acoustic diaphragm captured by
the STEAM vibrometer with 1 nm axial resolution and
30 ps dwell time. The diaphragm was driven to vibrate at
30 kHz .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 20
Fig. 3.4 Axial velocity of the acoustic diaphragm obtained by the
STEAM vibrometer. The diaphragm was driven to vibrate
at 30 kHz (the same as in Fig. 3.3) . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 20
Fig. 4.1 Concept of HDLS. HDLS operation relies on two
high-speed mapping processes. First, the spectrum of
each broadband optical pulse generated by a mode-locked
laser is mapped to time. This wavelength-to-time
mapping is performed by a temporal dispersive element
such as a dispersive optical fiber or prism pair. Next,
wavelength-to-space transformation is used to direct each
wavelength component of the optical pulse to a unique
point on the target. Overall, a one-to-one mapping between
time and space is formed. Therefore, each point in HDLS’
field of view is sampled with an individual wavelength
component of the optical pulse at a specific time. The
repetition rate of the mode-locked laser determines the
sampling rate of the HDLS . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 23
Fig. 4.2 HDLS experimental setup. In a 2D demonstration of laser
scanning with the HDLS, optical pulses generated by a
mode-locked Ti:Sapphire laser with a center wavelength of
814 nm and a repetition rate of 90.8 MHz are dispersed in
time using two pairs of prisms and a dispersive fiber. Pulses
are deflected in the vertical direction using an acousto-optic
deflector (AOD) at 105.4 kHz. Subsequently, the spectrum
of each pulse is mapped onto a horizontal line using a
pair of diffraction gratings. A combination of the vertical
deflection and horizontal mapping leads to a 2D raster scan
on the target. The pulse reflection off the surface of the
List of Figures xv

target is converted via an optical circulator to an electrical

signal using a single-pixel high-speed photodetector. This
is possible due to the prior wavelength-to-time mapping, so
that each wavelength component reaches the photodetector
at a unique time, and the information of different points
of the target are not overlapped. A 50 GS/s digitizer
acquires the electrical signal from the photodetector, which
corresponds to the spectrum of the optical pulses. After
correction for the background envelope, the spectrum of
each pulse reveals one horizontal line scan image of the
target. Stacking up many of these line scans in accordance
with the AOD scan frequency leads to a 2D raster scan of
the target . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 24
Fig. 4.3 Wavelength-to-time mapping with dispersive Fourier
transformation. (a) Optical pulses reflected off the
target corresponding to horizontal line scans at different
deflection angles of the AOD are measured by a high-speed
photodetector. The period of the horizontal scans is about
11 ns, which corresponds to the mode-locked laser’s pulse
repetition rate (90.8 MHz). (b) Good agreement between
the amplitude of the photodetector signal measured with
the digitizer (shown in blue) and the power spectrum
measured with a conventional optical spectrum analyzer
indicates the demonstration of the wavelength-to-time
mapping using the prism pairs and dispersive fiber in the
800 nm band . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 25
Fig. 4.4 Imaging with the HDLS. (a) Image of the word “UCLA”
engraved on the surface of a reflective substrate captured
by a CCD camera. (b) Image of the same sample captured
by the HDLS. The word “UCLA” is clearly shown . . . . . . . . . . . . . . . 25
Fig. 4.5 3D surface profilometry or 2D surface vibrometry with
the HDLS. 2D raster scans by the HDLS are used in
conjunction with a Michelson interferometer to perform
2D surface vibrometry. A beamsplitter splits scan pulses
into two arms. Optical pulses in one arm hit the target,
and the light in the other arm (reference arm) is reflected
intactly by a mirror. Reflected pulses from both arms are
combined at the beamsplitter and form an interference
pattern. If the reflectivity of the vibrating target is not
changing rapidly, Hilbert transformation can be used
xvi List of Figures

to extract the relative optical phase of each wavelength

component. Therefore, variations of the optical path length
at each wavelength component are measured and used to
form 3D surface profiles of the vibrating sample at a scan
rate of 105.4 kHz with 0.4 nm axial resolution .. . . . . . . . . . . . . . . . . . . . 26
Fig. 4.6 Surface vibration captured by the HDLS. Frames from 3D
scans of a vibrating diaphragm by the HDLS show a period
of nanomechanical vibrations at 1 kHz. Only one of every
ten scans is shown here . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 27
Fig. 4.7 Comparison of conventional and HDLS flow cytometers.
(a) In regular flow cytometry, a single interrogation beam
covers the desired field of view in the channel. Therefore,
it does not efficiently differentiate multiple cells such as
doublets. In HDLS flow-cytometer, diffraction limited
wavelength components of the interrogation beam cover
the required field of view, and extract high-resolution
spatial information of the sample. HDLS data can be used
to identify abnormalities, e.g., multiple cells and result
in a lower statistical error. (b) In our demonstration of
HDLS flow cytometer, an inertial focusing microfluidic
channel with a dielectric mirror substrate is used to order
randomly distributed cells into a single stream. The
microfluidic device was fabricated using standard replica
molding methods in thermoset polyester (TPE) to ensure
stability. HDLS scan pulses are focused on the stream.
Forward-scattered light from the cells is reflected by
substrate mirror and collected by an objective lens . . . . . . . . . . . . . . . 28
Fig. 4.8 Experimental results of HDLS flow cytometer. (a) Identical
samples of white blood cells and MCF7 breast cancer cells
are measured separately with conventional and HDLS flow
cytometers. There is a considerable overlap in forward
scattering range of these cell types for a conventional flow
cytometer. However, this overlap decreases significantly for
HDLS flow cytometer measurements because white blood
cell multiples are not identified as MCF7 cancer cells. (b)
Receiver operating characteristic (ROC) curves based on
identification of white blood cells and MCF7 breast cancer
List of Figures xvii

cells show that without sacrificing throughput, HDLS flow

cytometer achieves higher specificity and sensitivity than a
conventional flow cytometer . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 29
Fig. 5.1 Optical setup of Coherent-STEAM. A Coherent-STEAM
setup is formed by combination of STEAM and a
Michelson interferometer. A pair of diffraction gratings
generates a 1D rainbow with different wavelength
components imaging different points on the cells flowing
in a microfluidic channel. A pellicle beam-splitter and
two identical long working-distance objective lenses are
used to form the interferometer for phase measurement.
Back apertures of objective lenses are fully illuminated
with each wavelength component of the broadband
mode-locked laser pulses to ensure diffraction-limited
resolution. An amplified time stretch system chirps,
stretches, and amplifies each pulse, so that different
wavelength components reach the photodetector serially.
A very shallow microfluidic channel with hydrodynamic
focusing is designed and fabricated to align cells within the
focal depth of the system . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 36
Fig. 5.2 Digital signal processing of Coherent-STEAM. (a) The
photodetector output signal is digitized and recorded by an
ADC. This signal shows sequential laser pulses. (b) Each
pulse is saved separately as a frame for further processing.
(c) The analytic form of high-frequency components of
each pulse is generated using Hilbert transformation, and
the phase component of this analytic form is extracted. (d)
An unwrapping algorithm is used to fix unrealistic phase
jumps, and the result shows an approximately linear phase
increase. (e) If the phase component of the interferometer
fringe frequency is removed, the phase induced by cells in
optical pulse can be seen. (f) Many of these line images
generated from subsequent frames are used to form a
spatial map of optical path difference in two dimensions,
which is used for cell characterization . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 37
xviii List of Figures

Fig. 5.3 Calibration with NIST traceable beads. Polystyrene

beads with a NIST traceable diameter of 5 m are used
to calibrate the image processing algorithm for size
measurements. (a) A custom designed image processing
algorithm in CellProfiler software is used to find the beads
in spatial map of optical path difference and measure the
diameter. (b) Histogram of bead diameters demonstrates
the measured size distribution has an expected mean of
5 m and a standard deviation within the range of optical
resolution limit. (c) Since all the beads are made out of the
same material, the coefficient of variation for refractive
indices (0:014=1:57 D 0:89%) is much smaller than that of
diameters (0:405=5:06 D 8:00%) . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 39
Fig. 5.4 Cell classification based on size and protein concentration
measurement by Coherent-STEAM; images of (a) SW480
and (b) OTII cells taken by Coherent STEAM setup show
that they are spherical. (c) Scattering plot of cell protein
concentration versus diameter is shown for OTII (blue)
and SW480 (green) cells. (d) Comparison of the ROC
curves of size measurement only (purple line) to that of
simultaneous size and protein concentration measurement
(orange line) shows significant improvement in sensitivity . . . . . . 40
Fig. 6.1 Time stretch quantitative phase imaging (TS-QPI) and
analytics system. A mode-locked laser followed by a
nonlinear fiber, an erbium doped fiber amplifier (EDFA),
and a wavelength-division multiplexing (WDM) filter
generate and shape a train of broadband optical pulses.
Box 1: The phase shift and intensity loss at each location
within the field of view are embedded into the spectral
interference patterns using a Michelson interferometer.
Box 2: Time stretch dispersive Fourier transform.
Box 3: Big data analytics pipeline . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 44
Fig. 6.2 Comparison of the interferograms measured by optical
spectrum analyzer and time stretch dispersive Fourier
Transform; (a) Optical spectrum of the signal after
quantitative phase imaging (box 1 in Fig. 6.1) and before it
enters the amplified time stretch system (box 2 in Fig. 6.1).
The interference pattern in spectral domain is measured by
List of Figures xix

an optical spectrum analyzer. (b) With time stretch, the

interference pattern in spectral domain is linearly mapped
into time. The baseband intensity envelope is slightly
modified by the wavelength-dependent gain profile of the
Raman amplifier. The inserts in panels a and b show the
zoomed-in spectrum and waveform in the dashed black
boxes, respectively. Clearly, the single-shot interferogram
measured by Raman-amplified time stretch dispersive
Fourier Transform has a higher signal-to-noise ratio
compared to that captured by optical spectrum analyzer . . . . . . . . . . 48
Fig. 6.3 PDMS microfluidic channel mounted on a highly
reflective surface with near-infrared dielectric coating; The
microfluidic device consists of a hydrodynamic focusing
region and an imaging region targeted by the interrogation
rainbow flashes in TS-QPI system. (a) Sample solution
with suspended cells is fed into the channel through the
sample inlet, and deionized water as the sheath fluid is
injected through the sheath inlet. At the hydrodynamic
focusing region, the sheath pressure focused the sample
at the center of the channel by narrowing its flow width
from 200 µm to about 40 µm with a sheath to sample
volume ratio of 3:1. (b) The pattern of the mask used to
imprint microfluidic channel design on silicon wafer with
photoresist. The circles are inlet and outlet reservoirs .. . . . . . . . . . . . 51
Fig. 6.4 Left: screenshots of the video of OT-II hybridoma
T-lymphocytes flowing in a microfluidic channel; The cells
are aligned at the center of the channel by hydrodynamic
focusing. Optical path difference measured at four of the
interrogation points on the rainbow flash is shown as a
function of time in the right panels. Right: screenshots of
the video of SW-480 colon cancer epithelial cells flowing
in a microfluidic channel; The cells are aligned at the
center of the channel by hydrodynamic focusing. Optical
path difference measured at four of the interrogation points
on the rainbow flash is shown as a function of time in the
right panels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 58
xx List of Figures

Fig. 6.5 Quantitative optical phase and loss images of OT-II (blue)
and SW-480 (green box) cells; The optical loss images of
the cells are affected by the attenuation of multiplexed
wavelength components passing through the cells. The
attenuation itself is governed by the absorption of the light
in cells as well as the scattering from the surface of the
cells and from the internal cell organelles. The optical loss
image is derived from the low frequency component of the
pulse interferograms. The optical phase image is extracted
from the analytic form of the high frequency component
of the pulse interferograms using Hilbert Transformation,
followed by a phase unwrapping algorithm. Also,
supplementary Videos 1 and 2 show measurements of
cell-induced optical path length difference by TS-QPI
at four different points along the rainbow for OT-II and
SW-480, respectively . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 59
Fig. 6.6 (a) Pairwise correlation matrix visualized as a heat map.
The map depicts the correlation between all major 16
features extracted from the quantitative images. Diagonal
elements of the matrix represent correlation of each
parameter with itself, i.e., the autocorrelation. The subsets
in box 1, box 2, and box 3 show high correlation because
they are mainly related to morphological, optical phase,
and optical loss feature categories, respectively. (b)
Ranking of biophysical features based on their AUCs in
single-feature classification. Blue bars show performance
of the morphological parameters, which includes
diameter along the interrogation rainbow, diameter
along the flow direction, tight cell area, loose cell area,
perimeter, circularity, major axis length, orientation, and
median radius. As expected, morphology contains most
information, but other biophysical features can contribute
to improved performance of label-free cell classification.
Orange bars show optical phase shift features, i.e., optical
path length differences and refractive index difference.
List of Figures xxi

Green bars show optical loss features representing

scattering and absorption by the cell. The best performed
features in these three categories are marked in red . . . . . . . . . . . . . . . 62
Fig. 7.1 Spectral components of Coherent-STEAM signal. For a
Coherent-STEAM setup with long enough arms’ length
mismatch the spectrum of the output signal shows two
separate spectral bands. The low frequency components
correspond to the intensity of the sample, while the high
frequency components contain the phase information in
addition to the intensity information of the cells . . . . . . . . . . . . . . . . . . 68
Fig. 7.2 Analog preprocessing of Coherent-STEAM signal. The
analog signal processing system for reducing the data
rate of Coherent-STEAM is essentially a quadrature
down-conversion unit. I and Q outputs and their
corresponding spectra show that the down-conversion
is effective in reducing the bandwidth and the required
sampling rate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 69
Fig. 7.3 Digital signal processing system for acquisition of analog
preprocessing unit outputs. This system is built with simple
blocks such as argument calculator, unwrapper, and first in,
first outs (FIFOs), which can be performed in real-time . . . . . . . . . . 69
Fig. 7.4 Sample images acquired by the analog preprocessing
system. Both phase and intensity images for two different
sets of OT-II hybridoma T cells in flow are shown . . . . . . . . . . . . . . . . 70
Fig. 8.1 Machine learning pipeline. Information of quantitative
optical phase and loss images are fused to extract
multivariate biophysical features of each cell, which are
fed into a fully connected neural network. The neural
network maps input features by a chain of weighted sum
and nonlinear activation functions into learned feature
space, convenient for classification. This deep neural
network is globally trained via area under the curve (AUC)
of the receiver operating characteristics (ROC). Each ROC
curve corresponds to a set of weights for connections to
an output node, generated by scanning the weight of the
bias node. The training process maximizes AUC, pushing
the ROC curve toward the upper left corner, which means
improved sensitivity and specificity in classification . . . . . . . . . . . . . . 74
xxii List of Figures

Fig. 8.2 Classification of white blood cells (OT-II) and cancer

cells (SW-480) by TS-QPI label-free features; (a) Training
process of the neural network leads to improvement
of classification accuracy over generations of genetic
algorithm. In addition to multivariate analysis using all 16
biophysical features extracted from the TS-QPI quantitative
images (blue curves), we also show training process
by three single features. Red, green, and orange curves
represent the best biophysical feature in each category,
morphology, optical phase, and optical loss, respectively.
The values represent average balanced accuracy among
training datasets at the end of optimization. Clearly, the
final achievable accuracy by multivariate classification
is considerably higher than that of single features. (b)
For each case, we show 5 ROC curves for different test
datasets. The gray diagonal line shows results of random
guess classification. Multivariate analysis based on TS-QPI
images (blue curves) shows significant improvement
in classification sensitivity and specificity. The fact
that the classifiers remain almost unchanged during the
five iterations of cross validation shows consistency
and robustness of the classifiers. (c) To visualize the
multivariate classification results, data points are depicted
in the space of the first two PCA components .. . . . . . . . . . . . . . . . . . . . 77
Fig. 8.3 Three-dimensional scatter plot based on size, protein
concentration, and attenuation of OT-II and SW-480
cells measured by TS-QPI. The green and blue dots are
two-dimensional (2-D) projections of cell data points on
the planes containing only two of the biophysical features.
The cell protein concentration corresponds to the mean
refractive index difference of the cell (Refractive index
feature in Table 6.2). The attenuation is a feature describing
the optical intensity loss caused by cell absorption
(Absorption-1 feature in Table 6.2). Comparison of 2-D
scatter plots reveals that additional biophysical features (in
this case mainly protein concentration) serve to classify the
cell types more accurately . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 78
List of Figures xxiii

Fig. 8.4 Classification of algal cells (Chlamydomonas

reinhardtii) based on their lipid content by TS-QPI. (a)
Three-dimensional scatter plot based on size, protein
concentration, and attenuation of the cells measured by
TS-QPI, with 2D projections for every combination of two
features. Inset: Conventional label-free flow cytometry
using forward scattering and side scattering is not enough
to distinguish the difference between high-lipid content
and low-lipid content algal cells. TS-QPI is much more
effective in separating the two algae populations. (b) ROC
curves for binary classification of normal and lipid-rich
algae species using ten-fold cross validation; blue curves
show the classifier performance using all 16 biophysical
features extracted from the TS-QPI quantitative images.
Red, green, and orange curves show the classifier decision
performance using only the best biophysical feature in each
category: morphology (Diameter-RB in Table 6.2), optical
phase (OPD-1 in Table 6.2), and optical loss (Absorption-2
in Table 6.2). The label-free selection of algal strains
improves as more biophysical features are employed . . . . . . . . . . . . . 79
Fig. 8.5 (a) The learning curves of the training and test datasets
in the tumor cell detection. Larger number of training
data points decreases the cross entropy of the test dataset,
which means the classifier is performing more accurately.
However, the trend is opposite for the training dataset
because with a larger number of training data points fitting
error accumulates. The discrepancy of the training and test
errors, i.e., generalization error, decreases up to N Š 850,
which is the necessary training data size for achieving
final performance in our TS-QPI demonstration with deep
learning neural network. (b) Comparison of multiple
machine learning classification techniques based on the
biophysical features extracted from the label-free cell
images captured by TS-QPI. Our AUC-based deep learning
model (DNN C AUC) has both the highest accuracy and
consistency against support vector machine (SVM) with
Gaussian kernel, logistic regression (LR), naive Bayes, and
conventional deep neural network trained by cross entropy
and backpropagation (DNN) .. . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 81
xxiv List of Figures

Fig. 8.6 Principal component analysis (PCA) on the multivariate

data set produced by time stretch quantitative phase
imaging. (a) Upper bar chart shows accuracy of
classification by each individual principal component, and
lower bar chart shows the percentage of the total variance
explained by each principal component, accounting for the
variability expressed in the data. As expected, principal
components with larger variability do not necessarily give
high accuracy in classification. (b) Cumulative accuracy.
The value at each data point corresponds to the number of
PCA components retained in order to achieve that total
explained variance. In order to reduce the number of
input features and decrease computation time, a subset
of the PCA components can be used for classification.
The classification accuracy improves as the total variance
retained in the subset of PCA components goes up.. . . . . . . . . . . . . . . 82
Fig. 8.7 (a) The implementation of the k-fold cross-validation here
splits data points into training, validation, and test subsets.
In each iteration, one fold is used for fine tuning the
learning model (validation dataset) and another fold is used
for evaluation of the final results (test dataset), while rest of
the data points are used for training (training dataset). The
final reported results are aggregate of the outcomes from
the test datasets. (b) A suitable regularization parameter, ,
balances the trade-off between overfitting (variance) and
underfitting (bias) and minimizes the cross entropy of the
validation dataset . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 83
Fig. 9.1 Illustration of warped stretch transform in imaging. (a) The
field of view consists of a cell against the background such
as a flow channel or a microscope slide. Illumination by
an optical pulse that is diffracted into a one-dimensional
rainbow maps one dimension of the space into the optical
spectrum. The other dimension is scanned by the cell
flow through the rainbow. In the conventional time stretch
imaging (STEAM), the spectrum is linearly mapped into
time using a dispersive optical fiber with a linear group
delay. The temporal waveform is then sampled by a
List of Figures xxv

digitizer with fixed sampling rate resulting in uniform

spatial sampling. But uniform spatial sampling generates
superfluous data by oversampling the sparse peripheral
sections of the field of view. (b) Similar functionality can
be achieved in STEAM by using a nonlinear group delay
profile in the spectrum-to-time mapping process resulting
in a nonuniform sampling of the line image, assigning
more pixels to the information-rich central part of the field
of view and less to the low-entropy peripherals.. . . . . . . . . . . . . . . . . . . 90
Fig. 9.2 Warped stretch transform in imaging inspired by biology
and art. (a) The human vision is a form of warped imaging
system where high sampling resolution is needed in the
central vision while coarse resolution can be tolerated in
the peripheral vision. (b) The reconstruction is similar to
anamorphic art where the drawn shape is a stretched and
warped version of the true object yet the viewer sees the
true object upon reflection from a curved mirror. In our
system, this unwarping operation is a nonlinear mapping
using the inverse space-to-frequency-to-time mapping
transfer function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 92
Fig. 9.3 Linear and warped (anamorphic) stretch transforms.
The linear group delay profile results in uniform
frequency-to-time mapping (orange plots), whereas
the warped group delay profile results in nonuniform
mapping (blue plots). (a) A nonlinear group delay with
the same dispersion (slope) at the center of the spectrum
as linear case, but shorter total group delay, leads to high
sampling resolution in the center of the spectrum and lower
resolution at the wings. This keeps the image quality at
the central part of the field of view intact, while reducing
the quality at the sparse peripheral regions where uniform
stretch would have produced redundant samples. (b)
A nonlinear group delay profile with higher dispersion
(slope) at the center of the spectrum than the linear case,
but same total group delay over the bandwidth, leads to a
higher spectral resolution in the center of the spectrum and
lower resolution at the wings (compare the spectrums). The
gray curves show the analog waveforms before sampling
for the purpose of comparison . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 93
xxvi List of Figures

Fig. 9.4 Simulations illustrate the effect of warped stretch transform

on a two-dimensional image. The analog reshaping of the
image performed in the optical domain by the warped
stretch transform is emulated here in the digital domain.
(a) The transformation consists of nonuniform stretch
in the horizontal direction with the warp stretch profile
shown in Fig. 9.3. (b) A sample image with 28,001,672
pixels and 4.46 MB file size is used as the input. (c)
The image is stretched nonuniformly and down-sampled
with a compression ratio of 14. (d) A uniform stretch
with down-sampling can achieve the same file size
but the image quality is dramatically lower. (e) While
down-sampling is not an issue for the sparse peripherals, it
is problematic for the information-rich central part. (f) The
reconstruction of the nonuniformly stretched image. (g)
The information-rich region at the center is well preserved
while maintaining the same sampling rates . . .. . . . . . . . . . . . . . . . . . . . 94
Fig. 9.5 Experimental setup used in proof-of-concept demonstration
of optical data compression via warped time stretch
imaging. A train of broadband optical pulses was generated
at 1550 nm central wavelength with a repetition rate of
36.129 MHz and a pulse width slightly less than 100 fs.
The laser pulses were temporally stretched to about 1 ns
by a dispersion compensating fiber and amplified by an
erbium-doped fiber amplifier (EDFA). The bandwidth
over 1541–1561 nm was selected by a wavelength division
multiplexing (WDM) filter. The pulses passed through
an optical circulator and were coupled to free-space part
of STEAM setup with a fiber collimator. There, a pair of
diffraction gratings generates a one-dimensional rainbow
with each wavelength component imaging a different
location at the target. The spectrally encoded rainbows are
reflected and coupled back into the fiber, carrying the image
information. The nonuniform space-to-frequency-to-time
mapping is achieved with a warped chirped fiber Bragg
grating (CFBG). After optical image amplification by
another EDFA, different wavelength components are
detected serially by a single-pixel photodetector and
acquired by an analog-to-digital converter (ADC) .. . . . . . . . . . . . . . . 96
List of Figures xxvii

Fig. 9.6 Proof-of-concept experimental setup. (a) The test sample

reflected one-dimensional rainbow illumination pulses,
which are used to perform time stretch imaging at a scan
rate of 36 MHz. The field of view determined by the length
of the rainbow was 5 cm and covers the width of the
target. The vertical direction was scanned by mechanical
translation at 0.5 mm per step. (b) The warped stretch
transform leading to nonlinear spectrum-to-time mapping
is performed by a custom chirped fiber Bragg grating with
sublinear group delay (GD) profile. This profile gives
higher group delay dispersion at the center frequency and
reduced dispersion at the wings of the bandwidth . . . . . . . . . . . . . . . . 97
Fig. 9.7 Proof-of-concept experimental results. (a) If we use a
linear group delay profile with the same dispersion as
that of the warped stretch at the center frequency and a
single pulse per image line, the image data size would be
24.3 kB (55,345 measured pixels). (b) The single-pulse
reconstructed image based on the waveform nonlinearly
stretched by the chirped fiber Bragg grating has an obvious
warping effect at the center of the field of view (letter “S”)
(18,945 measured pixels). (c) The single-pulse unwarped
reconstructed image data size is 8.3 kB achieving about
three times optical image compression (18,945 measured
pixels). (d, e and f) When many pulses (722 pulses here)
are averaged to form each horizontal line image, the
images’ quality improve only slightly over Fig. 9.7a–c,
proving high signal-to-noise ratio of our camera even
in single-pulse capture mode. The number of measured
pixels used in the formation of Fig. 9.7d is 72,255,345, and
for Fig. 9.7e, f is 72,218,945. The temporal durations of
the waveforms corresponding to each horizontal line in
Fig. 9.7a–f are 27.7 ns, 9.5 ns, 9.5 ns, 20.0 s, 6.8 s, and
6.8 s, respectively . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 98
Fig. 10.1 Linear and warped time stretch dispersive Fourier
transforms. In linear time stretch (orange plots), a linear
group delay profile with significant group delay over the
signal bandwidth is used as the kernel to delay various
xxviii List of Figures

spectral components differently, but with a constant group

delay dispersion. If the input signal is a pulse train, pulse
spectra are linearly mapped to the silent times in between
the pulses. In contrast, a nonlinear group delay profile
(blue plots) with varying group delay dispersion over
the bandwidth (different slopes) can stretch the signal
spectrum nonlinearly, in which parts of the spectrum are
stretched more than the others. This can be used, even
with a constant rate sampler, to increase the spectral
resolution at regions of the bandwidth where higher
resolution is required and to reduce the resolution where
the spectral features are sparse (see blue bars). In this way,
the spectrum is warp stretched into a waveform with the
same temporal duration as the linear profile, but smaller
bandwidth. In other words, for signals with spectral
sparsity, the envelope time bandwidth product can be reduced . . . 103
Fig. 10.2 Group delay design based on spectrotemporal sparsity at
the spectrum peripheries. (a) Envelope of the electric field
of an input optical signal. (b) The spectrum magnitude
of the input envelope. (c) Spectrogram of the spectrum
magnitude formed by short-term Fourier transform. (d) If
a temporally dispersive element with a linear group delay
profile over the optical bandwidth is used to stretch the
input optical field, (e) the spectrum maps uniformly to
temporal envelope of the electric field. (f) Spectrogram of
the envelope waveform amplitude resembles that of the
spectrum magnitude. (g) If a nonlinear group delay profile
with lower dispersion at the sides of the bandwidth is used
to stretch the optical pulse, (h) the spectrum is nonlinearly
mapped to the electric field envelope in time. (i) The
spectrogram of the electric field envelope amplitude after
the nonuniform dispersion shows that a shorter temporal
window is required to capture the waveform with the same
acquisition bandwidth. The green and blue dot-dash boxes
in Fig. 10.4f, i show the acquisition time and bandwidth,
and those in Fig. 10.4c show the effective bandwidth for
linear case and nonlinear kernel design, respectively .. . . . . . . . . . . . 105
List of Figures xxix

Fig. 10.3 Spectral sampling resolution of time stretch dispersive

Fourier transform tuned for the spectrum center. (a)
Spectral resolution limits for the linear group delay profile
and the acquisition system of Fig. 10.2d–f. The overall
spectral sampling resolution of the linear time stretch is
independent of the envelope optical frequency and limited
by the ambiguity in the frequency-to-time mapping of the
dispersive Fourier transform. (b) The spectral sampling
resolution of the linear group delay profile magnified ten
times (for visual clarity) and overlapped on the profile. (c)
Spectral resolution limits for the nonlinear group delay
profile and the acquisition system of Fig. 10.2g–i, unlike
the linear stretch, depend of the envelope optical frequency.
(d) Magnified spectral sampling resolution of the warped
time stretch overlapped on its group delay profile clearly
shows the ambiguity grows at the spectrum peripheries
DFT dispersive Fourier transform, PD photodetector, ADC
analog-to-digital converter, Total overall spectral sampling
resolution .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 108
Fig. 10.4 Group delay design based on spectrotemporal sparsity
at the spectrum center. (a) Envelope of the electric field
of an input optical signal. (b) The spectrum magnitude
of the input envelope. (c) Spectrogram of the spectrum
magnitude formed by short-term Fourier transform. (d) If
a temporally dispersive element with a linear group delay
profile over the optical bandwidth is used to stretch the
input optical field, (e) the spectrum maps uniformly to
temporal envelope of the electric field. (f) Spectrogram
of the envelope waveform amplitude resembles that of
the spectrum magnitude (shown in Fig. 10.4c). (g) If a
nonlinear group delay profile with lower dispersion at the
center of the bandwidth is used to stretch the optical pulse,
(h) the spectrum is nonuniformly mapped to the electric
field envelope in time. (i) The spectrogram of the electric
field envelope amplitude after the nonuniform dispersion.
The green and blue dot-dash boxes in Fig. 10.4f, i show the
acquisition time and bandwidth, and those in Fig. 10.4c
show the effective bandwidth for linear case and nonlinear
kernel design, respectively . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 109
xxx List of Figures

Fig. 10.5 Spectral sampling resolution of time stretch dispersive

Fourier transform tuned for the spectrum peripheries. (a)
Spectral resolution limits for the linear group delay profile
and the acquisition system of Fig. 10.4d–f. The overall
spectral sampling resolution of the linear time stretch
is independent of the envelope optical frequency and
limited by the Nyquist bandwidth of the analog-to-digital
converter. (b) The spectral sampling resolution of the linear
group delay profile magnified ten times (for visual clarity)
and overlapped on the profile. (c) Spectral resolution limits
for the nonlinear group delay profile and the acquisition
system of Fig. 10.4g–i. (d) Magnified spectral sampling
resolution of the warped time stretch overlapped on its
group delay profile clearly shows the ambiguity grows at
the spectrum center
DFT dispersive Fourier transform, PD photodetector, ADC
analog-to-digital converter, Total overall spectral sampling
resolution .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 110
Fig. 10.6 Design of an ideal group delay profile based on
spectrotemporal sparsity and input signal chirp. (a)
Envelope of the electric field of an input optical signal.
(b) The spectrum magnitude of the input envelope. (c)
Spectrogram of the spectrum magnitude formed by
short-term Fourier transform. (d) To perform uniform
frequency-to-time mapping, a linear group delay design
minus the input signal chirp should be used to stretch the
input optical field. (e) The spectrum maps uniformly to
temporal envelope of the electric field by the total group
delay profile. (f) Spectrogram of the envelope waveform
amplitude resembles that of the spectrum magnitude
(shown in Fig. 10.6c). (g) If a nonlinear group delay
profile is designed based on the blue dot-dash contour in
Fig. 10.6c and the input signal chirp, (h) the spectrum is
nonlinearly mapped to the electric field envelope in time.
(i) The spectrogram of the electric field envelope amplitude
after the nonuniform dispersion. The green and blue
dot-dash boxes in Fig. 10.6f, i show the acquisition time
and bandwidth, and those in Fig. 10.6c show the effective
bandwidth for linear case and nonlinear kernel design,
respectively . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 111
List of Figures xxxi

Fig. 10.7 Spectral sampling resolution of time stretch dispersive

Fourier transform designed for ideal exploitation of the
spectrotemporal sparsity. (a) Spectral resolution limits
for the chirp compensated linear group delay profile and
the acquisition system of Fig. 10.6d–f. (b) The spectral
sampling resolution of the linear group delay profile
minus chirp magnified ten times (for visual clarity) and
overlapped on the profile. One tenth of the overlay width
at each group delay corresponds to the set of the optical
frequencies that are captured at the same delay and are
indistinguishable in the temporal waveform. (c) Spectral
resolution limits for the nonlinear group delay profile and
the acquisition system of Fig. 10.6g–i, unlike the linear
stretch, depend of the envelope optical frequency. (d)
Magnified spectral sampling resolution of the warped time
stretch overlapped on its group delay profile
DFT dispersive Fourier transform, PD photodetector, ADC
analog-to-digital converter, Total overall spectral sampling
resolution .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 113
Fig. 10.8 Design of an ideal group delay profile for a signal with
asymmetric spectrum about the carrier frequency. (a)
Envelope of the electric field of an input optical signal.
(b) The spectrum magnitude of the input envelope is
asymmetric. (c) Spectrogram of the spectrum magnitude
formed by short-term Fourier transform. (d) To perform
uniform frequency-to-time mapping, a linear group delay
design minus the input signal chirp should be used to
stretch the input optical field. (e) The spectrum maps
uniformly to temporal envelope of the electric field by the
chirp-corrected linear group delay profile. (f) Spectrogram
of the envelope waveform amplitude resembles that of
the spectrum magnitude (shown in Fig. 10.8c). (g) If a
nonlinear group delay profile is designed based on the blue
dot-dash contour in Fig. 10.8c and the input signal chirp,
(h) the spectrum is nonuniformly mapped to the electric
field envelope in time. (i) The spectrogram of the electric
field envelope amplitude after the nonuniform dispersion.
The green and blue dot-dash boxes in Fig. 10.8f, i show the
acquisition time and bandwidth, and those in Fig. 10.8c
show the effective bandwidth for linear case and nonlinear
kernel design, respectively . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 114
Another random document with
no related content on Scribd:
 Here I pointed to two large lenses that, stripped of their paper, lay
on the stones. The agony of the customs-master was now complete.
He was afraid even to touch the various parcels. There they lay. He
wept.
Regardless of his entreaties, as he had been of those of various
Europeans whose bottled stout lay exploding in the sun, I smilingly
retired, telling him he would doubtless hear from the Governor.
He did. The Governor was furious; when the custom-house
master’s zeal touched him personally, he was really enraged; though
when we had appealed to him to get perishable things given up to
us, offering an indemnity, if it could be at any time proved that any
duty was due, he had told us he could do nothing.
The customs-master was heavily fined, and at any time during the
rest of my stay in Shiraz when I sent for my boxes, they were given
up at once, and when my servant, as directed, asked if it were
wished to examine them or not, the customs-master, pale with rage,
would reply:
“Go, son of a burnt father, no; I have opened his boxes once, I
never want to do so again.”
All this my man would gleefully narrate on triumphantly bringing
home my beer, or whatever had arrived.
I had had one other transaction with this customs-master. He had
a handsome colt rising three; I had long tried to buy it, but he would
never sell, or demanded a preposterous price. At last he sent over
one day to me saying, “What will you give me in cash for my grey
colt?”
I replied, “Ninety-five tomans” (about thirty-eight pounds). This is
really a very high price for a horse only rising three. To my
astonishment and delight the horse was sent over. I gave a cheque
for the money and tied my purchase up. The next day I was left in
peace to admire him; the third day came a letter politely written, the
pith of which was, “Return me my colt, I have repented.” I looked on
the affair as a joke, but no; the man had not cashed my cheque. Had
I paid him in specie the bargain would have been concluded; as it
was he was in the right, and I reluctantly gave back the horse I had
had my eye on for months.
It was the law, and by that one must abide.
A peculiarity of the Shirazi is his fondness for repeating words,
changing the initial of the second. Use is second nature, and a
curious instance of the habit is narrated of the late Kawam-u-Dowlet.
When in the presence of the Shah, the Kawam-u-Dowlet was asked
by his Majesty—
“Why is it, Kawam, that you Shirazis always talk of kabob-mabob,
and so on? you always add a nonsense word; is it for euphony?”
“Oh, ‘Asylum of the Universe,’ may I be your sacrifice; no
respectable person in Shiraz does so, only the lūti-pūti says it.”
Pūti is, of course, a nonsense (or meaningless) word, and lūti, as
here used, means a “blackguard!”
 CHAPTER XXII.
SHIRAZ AND FUSSA.

Cheapness of ice—Variety of ices—Their size—Mode of procuring ice—Water of

Shiraz: its impurity—Camel-fight—Mode of obtaining the combatants—Mode
of securing camels—Visit to Fussa—Mean-looking nag—His powers—See
the patient—State of the sick-room—Dinner sent away—A second one arrives
—A would-be room-fellow—I provide him with a bedroom—Progress of the
case—Fertility of Fussa—Salt Lake—End of the patient—Boat-building—Dog-
cart—Want of roads—Tarantulas—Suicide of scorpions—Varieties—
Experiment—Stings of scorpions—The nishan.

A great thing in such a warm place as Shiraz is the cheapness of

ice; for about fifteen shillings in dear years and five in cheap ones,
ice can be obtained, all through the warm weather, and in fact is
used from May to October, as no one would think of drinking
anything uncooled. A huge block is thrown down in one’s doorway
each morning by the ice-seller; it is supposed to weigh two Tabriz
maunds, or fourteen pounds.
The Persians well understand the art of making water-ices and
ice-creams, and various ices unknown to us are made by them, as
tamarind juice, pomegranate and cherry-water ices; iced “mast,” or
curdled milk, and various ices of pounded fruits, as apricots and
cherries, which are very good.
Ices, however, are served with them on a more lavish scale, and a
larger quantity eaten, than with us. When I accompanied Captain St.
J⸺ in a call he made upon the Muschir, four conical ices, the size
and shape of an ordinary sugar-loaf, were placed in handsome
Chinese porcelain basins before each of us. In fact the cheapness of
provisions generally causes among the well-to-do a lavishness and
profusion (not to call it waste) unknown in Europe.
 The Muschir has a “yakhjal,” or place for the making and collecting
and storing ice, in an open plain some six miles from the town at the
side of the Ispahan road. The earth is dug out to a depth of two feet;
with this earth a mud wall some twenty feet high is constructed of
sun-dried bricks. The excavation is some ten yards by one hundred,
and the wall is ten yards high by a hundred long; this wall runs in
such a way as to protect the open pond thus excavated, from the
hottest rays of the sun.
The delicious Ab-i-Rookhni (“stream of Rookhnabad”) is diverted
from its course during the first cold night. A few inches of still clear
water is collected in the pond, by morning it is frozen; at night the
water is again admitted, and another inch or two of ice made. When
three to six inches thick, the ice is broken and collected for storage in
a deep well on the spot: and so day by day the process goes on
during the short winter until the storehouses are full. Should the
supplies from these be exhausted by a very large demand, ice, or
rather blocks of snow, are brought from the mountains; but as these
are some distance, and as snow melts much faster than ice, the
weights being equal, the price rises.
An order is generally issued when the ice is running short that
each house is to be on half allowance—a wise measure, as it makes
the cooks careful, and so everybody gets some.
So common is the use of ice that the poorest are enabled to have
it, a big bit being sold for a farthing, and even the bowls of water for
gratuitous drinking at the shop doors are cooled by it. Ripe fruit is
generally also cooled prior to being eaten.
The water of Shiraz itself—unless that of Rookhni or Zangi,
another mountain spring, which has to be brought from a distance—
is almost poisonous, being much contaminated by surface drainage,
etc. The Mussulman world has a horrible idea that a certain body of
water, however great the quantity, or disgusting the nature, of the
filth poured into it, remains absolutely pure, and the result is that a
great deal of serious illness is produced; this is one of the reasons
why cholera is so very severe in the East, irrespective of the natural
action of the climate. I can only say that I was obliged, while in
Shiraz, never to drink water save when from the two springs, or in
the form of tea, when it was of course boiled; one glass of sherbet or
one tumbler of water making me ill.
My friend the British Agent, Mirza Hassan Ali Khan, C.S.I., used
always to send a servant to the Rookhni stream, but the fellow used,
I fancy, to fill his water-skins, which were carried on a mule,
elsewhere, for although Rookhni water tastes of the wild sage that
grows on its banks, and my friend the agent thus had as he fancied
an infallible means of knowing it, yet I think the wily water-bearer
was capable of having a stock of sage leaves with which he would
flavour the water got from the town stream, along the edges of which
sat all the laundresses of Shiraz washing (in it) their foul linen.
I met the Prince-Governor one day on horseback, and he asked
me if I had ever seen a camel fight. I replied that I had not; he told
me to ride on with him and he would show me sport, as he was
going to amuse himself with one. We rode to the back of the royal
garden, or “Bagh-i-No” (new garden); here are always encamped
hundreds of mules and camels. No sooner did the prince arrive than
the camel-men attempted to run away, but the farrashes with their
long sticks and a few horsemen soon brought them back.
The prince ordered them to bring two male camels (in a state of
must[21]). At this they wept and tore their hair, for they did not wish
their property to be destroyed or depreciated for the amusement of
the young shah-zadeh. However, there was no escape; the courtiers
soon pointed out two huge males secured apart from the rest of the
animals, and from their continued groanings and roarings evidently
in a state of must.
By the prince’s orders these were let loose; they “went for each
other” at once. At first they danced round one another in a lumbering
way; then what appeared like a huge bladder was projected from
their mouths: they then knelt before each other, and a sort of fencing
match took place; the ordinarily quiet, patient faces of the beasts
were changed into ones of savage fury; the mouths widely open, and
the retracted upper lips showed the white teeth; and from the open
mouths came quantities of foam. The long supple necks were
interlaced, and quickly darted from side to side, while now and then
the open savage mouths would be locked together. The object was
to seize either the throat or leg. The feints and meeting of the
mouths lasted some few minutes, accompanied by loud groans and
roarings of extraordinary fierceness. At length one beast, the paler
one, seized his adversary by the foot; while the other, a handsome,
long-haired animal, only got hold of his opponent’s ear. Blood flowed
freely, and the poor camel-men, who wanted to separate the
animals, were much beaten by the attendants. At last the dark long-
haired animal left go, and roared with agony; the victor commenced
to drag him about by the bitten foot. After some pressure the prince
allowed them to be separated. An awful wound was apparent on the
foot of the dark camel, and the efforts of some eight men with
bludgeons, ropes, and chains were required to restrain the victor
from pursuing his advantage, while the vanquished limped off with
his weeping master, roaring with mingled rage and pain. The prince,
ordering ten tomans to be given to the camel-men (poor fellows! I
doubt if anything ever reached them), rode off.
Camels are mostly not vicious, save when must. They can kill a
man with a kick; and when they bite, worry; and they generally bite
the piece out. They are usually tethered apart when in this state,
which soon passes off, and secured by bending one or even both
fore-feet, and tying the leg in this position, which renders them
powerless. This is the same principle as that adopted by Rarey the
horse-tamer.
About this time I was requested to go at once to Fussa—this place
is some ninety miles from Shiraz—where the Governor of the
province (that of Fussa) lay ill. I obtained leave of absence for a
week, and made an arrangement to go in one day and night, return
in another twenty-four hours, and have five days’ attendance on my
patient.
The Muschir, to whose daughter Mirza Ali Akbar the Khan Wakeel,
as he was called, was married, was anxious that I should start at
once, and gave me a handsome fee. I agreed that I was while at
Fussa to be the guest of my patient, and that he was to find me in
food and house-room; and this was an important stipulation, as in a
place where a European had possibly never been seen, it might
have been difficult to get a comfortable lodging or even food.
Two wild-looking horsemen and a servant of the British Agent
presented themselves, with a little bay pony of dolorous aspect in a
halter for my riding. I, knowing that no horses were laid out, asked
with scorn how the small pony could possibly carry me ninety miles
in twenty-four hours, which I had stipulated for; but all three men
assured me that I should certainly arrive, myself and saddle-bags, in
the time on the little animal. I confess that I disbelieved them, but we
started off at a smart canter, and we cantered for ten hours with
short breaks of ambling. We got to a village two hours after midnight,
slept for four hours, and arrived in Fussa at four, doing the journey
under the time; the small bay seemingly not at all distressed, and
prepared to go back to Shiraz at once at the same rate if desired. I
was then surprised; I know that any one of my own well-fed big
horses would have knocked up. But these little animals, used to the
severest continuous work, do it in the most extraordinary manner. I
am a big, heavy man; my saddle was a heavy English hussar
saddle, and my bags and bedding certainly weighed forty pounds or
more. Of course the beast I rode was a good one of his kind, and
probably a very good one, for when I left, the Governor of Fussa
declined to part with him even at a fancy price of three times his
apparent value.
We rode up to the door of the Governor, whom I found propped up
with pillows in a corner of the room, a huge, fat man about thirty-
eight, who was a general debauchee, opium-eater, wine and spirit
drinker, and bhang smoker. He was suffering from gout.
An aged Syud, with a long beard and blue turban, was in
attendance from Shiraz as his physician. The Governor himself was
a strikingly good-tempered, even jovial man, and between the
paroxysms of his gout, joked and talked. The village, or rather
district, magnates sat round him chiming in with all his observations,
and trying to soothe his pains. They were, master and retainers, the
fattest set of men I ever saw collected in one room. A long
description of the patient’s ills ensued, many pipes of peculiar
construction were smoked, and I was offered a tumblerful of strong
spirit as a matter of course, considerable surprise being expressed
at my refusing it.
Tea was continually handed round, which everybody, including the
patient, swallowed; a native bottle was frequently produced from
under his pillow from which he partook of copious draughts of pure
spirit, taken from a silver bowl holding half a pint: this was emptied
frequently. Every two hours my patient swallowed a bolus of opium.
Though we were in the middle of summer, some thirty to fifty people
were always in the room, and every window was shut and curtained;
thus a semi-darkness was produced. Smoke from innumerable pipes
filled the air, while the heat was rendered greater by a huge
samovar, in which water for tea-making bubbled. The temperature
was ninety degrees out of doors at five p.m. The chatter of
conversation was continuous, and four musicians strove to drive dull
care away by playing loudly in a corner. I found my patient had just
had an attack of delirium tremens, and was going the right way for
another.
At seven, after having prescribed for him, I escaped to my quarters
under the pretext of dining, and lay down to rest. At nine my servant
informed me that my dinner was about to be served; and a large
circular tray, having some six dishes on it, with bread and all et
cæteras, a huge bowl of iced sherbet and a bottle of wine, was
brought in. I was very hungry and anxious to fall to, and I felt a sense
of anguish, when, to my astonishment, my servant (whom I had
brought from Shiraz), assuming the part of the Governor of
Barataria’s physician, ordered the whole away in an indignant voice.
As soon as my dinner had disappeared, I demanded an
explanation of my man. It was this: “I know, sahib, that the dinner I
sent away was quite enough for the sahib, and a good dinner, but
here in Persia a man’s position is reckoned by the quantity of the
dinner sent him, and the number of plats. They have sent you six
plats. I have told them that you couldn’t think of dining on less than
eighteen; and if I allowed you to eat the dinner that was sent, good
as it was, you would be looked down on. Are you less than the
prince’s physician? Certainly not. They would send him, or rather he
would demand, at least twelve plats. I assure you I am acting in your
interest.”
I suppose the fellow was right. Dinner for at least twenty-four
persons was brought on three huge trays. I tasted some half-dozen
well-cooked dishes, and then my servant removed the rest, and I
observed him, with the master of the house and numerous hangers-
on, dining in the open air on the very copious dinner that remained.
The man was right. Such are some of the ways a Persian has of
keeping up his consequence.
About ten p.m. I went, a few doors off, to the house of my patient
the Governor. The same stifling room, the same hard drinking, only
now everybody was drinking. Dancing-boys and singers, shrieking
the noisy love-songs of Persia in chorus, were keeping up the spirits
of my patient. To the few who were not drinking wines or spirits, tea
was continually handed. Long conversations on the topic of the
patient’s illness took place, and on local politics. Wearily, at two a.m.,
on the ground that my patient must try to sleep, I succeeded in
getting away. I was accompanied home by a big man with a jet black
beard, a Khan who was one of the Muschir’s retainers, and a polite
fellow. What was my disgust to find his bedding spread in my little
room. I told my man to bring him my water-pipe to smoke, and then I
remonstrated. I said it was not a “Feringhi” custom to sleep in a room
with strangers. But the Khan said it was late, that there were no
other quarters, and commenced to disrobe. “If you must have a
separate room, take that,” said he, pointing to two heavy doors at the
end of the apartment.
I opened them, and found a small room, windowless, in which
apparently charcoal had been stored. Impossible for me to sleep
there; but, thought I, “these are really my quarters, why shouldn’t the
Khan sleep there? he is the intruder.” I pretended to collect my
bedding.
“You are never going in there!” said he.
“Why not?” I replied. “Look, look! the huge scorpion!” I shouted. He
jumped up, seized a stick, and ran into the charcoal cupboard, for it
was nothing more.
 “Where?” said he.
“There! there!” said I.
He was well inside—to slam and secure the heavy doors was the
work of a moment. He shouted and swore, kicking at the door, but it
was a very strong one, and nobody came to his assistance. He then
entreated, promising to go elsewhere; but I couldn’t trust him, and so
I composed myself to sleep, and soon dropped off. In the morning on
waking a melancholy voice entreated liberation; but I could not do it
then, as he might have taken vengeance. So I went off to the
Governor, and complained of the intrusion on my quarters; my man
then liberated the much-begrimed but now humbled Khan; and I got
another set of rooms, in which I was more comfortable.
I saw no more of the Khan: the laugh was too strong against him,
and he returned at once to Shiraz. Until my arrival the Governor was
not aware of the nature of his disorder; with great trouble I got him to
reduce his opium and cease his potations. I was happily able to give
him relief, and we parted mutually satisfied after I had been five days
at Fussa.
The place was much warmer than Shiraz; grain and cattle were
cheap indeed here. The soil, though sandy, is very fertile; and the
town, or rather collection of villages (for it is more a district than a
town), is interspersed with groves of date palms. Oranges are, of
course, abundant, and there is great plenty in the place. There had
need be, for the exactions of those in power are very great in Persia.
The people were a laughing, careless set, devoid of fanaticism,
having indeed very little religion. Nearly all drank wine to excess.
The women seldom veiled, and talked with me without any mauvaise
honte. They indeed seemed to do most of the work; for the field-work
was probably not heavy, save at harvest-time, the country being so
very fertile. The road from Fussa was a howling desert, except a
well-watered village about half-way.
We passed the edge of the big Salt Lake, some ten miles from
Shiraz, on which appears an island, or what looks like an island.
After skirting this lake, whose shores are bordered by an edge of
mud some fifty yards in width, we reached the village of Jaffir-a-bad,
and thence, passing small villages and gardens in every direction,
got to the plain of Shiraz. The pony brought me in as quickly as I had
gone out, and I had had a peep at country life in the south of Persia.
The prince’s hakim-bashi, the M.D. of Paris, replaced me, and he,
too, had a week’s leave. When he left, the old Syud told the patient
that he had gangrene, cut off the gouty toe, and being unable to
staunch the blood, the man died in forty-eight hours.
A year after this, one of the sergeants built a large boat for the
exploration of the Salt Lake. This boat-building was an amusement
for us, but the boat was found to be so heavy that it required fifteen
porters to carry her through the streets. She certainly held eight
people, but was very deep in the water, and more a barge than a
boat, but as she was flat-bottomed she would not turn over. While I
was in Ispahan, where I had gone on duty, she was placed on the
large tank of the Bagh-i-Takht, and after twenty-four hours left to the
mercy of the populace. I believe she is at the bottom of that tank
now.
Another of the sergeants, a really skilled workman, and fired by
the actual boat having been floated, in conjunction with the builder of
the first one, resolved to make a real wherry. This they did from
various drawings, and they succeeded in building a very handsome
boat, having curved planks, which were bent with great trouble with
hot water. This boat was also fastened with copper rivets, and really
handsomely finished; but though so light that three men could carry
it, it held two comfortably, and would probably have been speedy, but
it was so terribly crank that no one would venture in it; and though it
was ballasted till the gunwale was almost level with the water, it
turned over on the slightest movement. This, the second and last
boat built by the English in Shiraz, is also at the bottom of one of the
tanks. At Ispahan an extraordinary barge was made afterwards by
two of the staff for the Zil-es-Sultan, and this could be rowed; but it
was a barge, and had no pretensions to be called a boat.
I now started a dog-cart, which I received from Kurrachee,
intending it for Ispahan, where there are good roads; as fate would
have it, when the thing arrived I had been stationed at Shiraz, where
there were next to no roads. I put my trap together with some
difficulty, for the wood had warped in the long land journey, and
found it to be a big dog-cart of the largest, heaviest, and (luckily)
strongest type. Its weight appalled us all.
Followed by a crowd, our servants dragged it outside the city
gates, and I put my chestnut horse in. Of course, he had never been
in the shafts before. On attempting to urge him forward he sat down,
as a dog sits, and declined to stir; this manœuvre he constantly
repeated. I now in despair tried an old and valueless grey horse. He
walked off with the machine at once, and, barring the want of roads, I
had no difficulty. Luckily, the trap was built of a solidity I have never
seen, save in railway carriages, and so, regardless of roads, I was
able to go about. The stony bed of the river, with an occasional bit of
hard road, and thence to the sandy plain of Jaffir-a-bad, was about
the only drive.
Here, and here only in the neighbourhood of Shiraz, one sees
enormous tarantulas. These beasts, some with bodies as large as a
pigeon’s egg and legs in proportion, are very brave; when attacked
with a stick, instead of taking flight, they advance threateningly at the
person who molests them, and attempt to bite the stick; they are
really formidable-looking brutes, covered with brown hair.
A story was told me by the late Dr. Fagergren, a Swede in Persian
employ, who had been twenty-five years in Shiraz, to the effect that
scorpions, when they see no chance of escape, commit suicide; and
he told me, that when one was surrounded by a circle of live coals, it
ran round three times and then stung itself to death. I did not credit
this, supposing that the insect was probably scorched, and so died. I
happened one day to catch an enormous scorpion of the black
variety. In Persia they are of two kinds: black; and light green, or
greenish yellow; the black variety being supposed to be much the
more venomous. The full-grown scorpions generally are from two to
three inches long; I have seen one five inches when extended from
the tip of the claws to the sting, but he was phenomenal. The one I
caught was very large, and to try the accuracy of what I supposed to
be a popular superstition, I prepared in my courtyard a circle of live
charcoal a yard in diameter. I cooled the bricks with water, so that
the scorpion could not be scorched, and tilted him from the finger-
glass in which he was imprisoned unhurt into the centre of the open
space; he stood still for a moment, then, to my astonishment, ran
rapidly round the circle three times, came back to the centre, turned
up his tail (where the sting is), and deliberately by three blows
stabbed or stung himself in the head; he was dead in an instant. Of
this curious scene I was an eye-witness, and I have seen it repeated
by a friend in exactly the same way since, on my telling the thing,
and with exactly the same result. For the truth of this statement I am
prepared to vouch.
Of the effects of the sting of the scorpion (generally only the lower
orders are bitten, as they are barefooted, and their work may take
them to cool and damp places where the insects love to lie) I have
had much experience: I consider it is never fatal, save in the case of
infants stung in the throat, but it is very painful, the only remedy
being liquid ammonia to the wound, which gives speedy relief.
I have never seen a case of serpent-bite in Persia, and
hydrophobia is very uncommon, though it is said to exist. The only
case of rabies I have seen is that of Pierson’s dogs, narrated
previously. Hornets and wasps sting badly, and frequently I have
known death occur in a child much stung. The sting is worse than
that of the British wasp. As to the Persian tarantula, it merely bites.
A curious custom in Persia is the “nishan,” or token. The token is
some secret conveyed by a third party, as a token or sign of the
consent of the giver of the token to the request. Thus a man will say,
if away from home, and one wishes to borrow his horse, “Tell my
steward to give you my horse, by the ‘nishan’ (or token), that I gave
him a present this morning.” As the steward knows that the giving of
the present was only known to his master and himself, he hands
over the horse at once.
 CHAPTER XXIII.
SHIRAZ.—THE FAMINE.

Approach of famine—Closing of shops—Rise in mule-hire—Laying in of stores—

Seizures of grain—Sale of goods by poor—Immigrations of villagers to the
towns—Desertions of children—Increase of crime—Arrival of money from
England—Orphanage—Labour question—Koomishah—Village ruffian—His
punishment—Prince’s accident—The kalāat—Mode of bringing it—Invitation
to the ceremony—Procession—Gala dress of the prince—The arrival of the
firman—Assemblage of grandees—The kalāat—The Kawam’s kalāat—Return
to town—Sacrifice of an ox.

The famine was now setting in in Persia seriously—for two years

not a drop of rain had fallen; the crowds of professional beggars
were reinforced by really hungry people, thefts from shops became
common, as did burglaries, and the roads were now very unsafe. In
the corn-chandlers’ shops very small supplies of grain were seen,
and these much adulterated by the addition of dust, stones, etc. The
bakers baked as little bread as they could, mixing their dough in as
small quantities and as slowly as possible; the loaves became
gradually worse and worse, though the price remained nominally the
same. The coarse barley-bread ceased to be baked altogether, and
at last the bakers refused to sell to the crowds which formed at their
shop doors unless they were their regular customers, and then only
for ready money, and one small loaf to each person, selling by
weight being discontinued altogether. All who had enough ready
money laid in a store of grain and flour.
 FAMINE GROUP.

Those who had cheap horses and donkeys sold them, and the
price of the cheaper class of horse fell very low, till at last beasts
were turned out as worthless, and killed and eaten by the poor.
Meat fell in price, but this did not much help those who had no
money to buy even bread. Large establishments were suddenly
much reduced; the armies of hangers-on, who live on the leavings of
the rich and their attendants, were now thrown upon the streets.
Many of the bakers and butchers closed their shops and fled.
Mule-hire rose to an almost prohibitive price, and it must be
remembered that this, in a country where all transport is by mule and
camel, meant the paralysis of trade. All the animals, save of the very
rich, presented a half-starved appearance. In the waste grounds
near the towns, and by the sides of the high-roads, lay the bodies of
dead and dying mules and horses.
Flour became adulterated, and was ground at home by the
consumers. Grandees, and merchants began to lay in stores of grain
from their villages, disposing of none, although an enormous profit
could be obtained on the contents of their granaries. The Governors
of the towns seized grain, or paid for it at a nominal price, and sealed
it up in the public or Government grain stores. Provender on the
high-roads became unattainable.
Prices, though steadily rising for all descriptions of cereals,
suddenly dropped on the hope of rain, only to rise in a few hours to a
still more serious figure. The lower classes began to pledge and sell
their copper-ware, tools, arms, and clothing. In the post-houses,
where from six to ten horses were generally kept, only two, and at
times none, were seen.
Villagers in quest of food began now to pour into the towns, and
remained herding in starving crowds in the mosques, having neither
the means nor the strength to return to their homes. The charity of
the Persians themselves was nearly exhausted, for each rich man
had to feed his crew of hungry servants and their families. The few
unorganised attempts to feed the poor, resulted in the crushing to
death of several, and the one loaf of bread doled out to each person
on these rare occasions only served to prolong their sufferings.
Children now began to be deserted in the streets, the dead and
dying to be seen frequently, the greater portion of the bazaar to be
closed, typhoid to be rife, and crimes of violence to be frequent.
And now came the first funds from England from the Persian
Relief Committee. In each town the money was husbanded and relief
given in the way most efficient and economical. Money was found to
be the most safe plan, at all events in Shiraz, of which I speak from
experience, for any attempt to buy bread in quantities failed, and
caused an immediate rise in price. Very many applicants were sent
away; relief in the shape of a numbered ticket, entitling the bearer,
whose person was described in a book kept for the purpose, to
weekly relief in money, was given to the utterly destitute. The
difficulty of deciding on the claims of the various applicants was
great, and in many cases which had to be denied permanent relief,
temporary alms were given.
A large house was rented, and in it were placed all the deserted
orphans found in the streets; these were mostly the children of
villagers, though some were those of townspeople. These children
were plainly but comfortably clothed in the ordinary dress of well-to-
do Persian villagers, and well and regularly fed. They were placed
under the care of an intelligent and humane Persian, who really did
his duty to them, and were regularly inspected by the members of
the Relief Committee; also they were frequently seen at unexpected
times. The poor emaciated bundles of rags soon developed into
strong, healthy children, and the regular food, comfortable quarters,
and good clothes did wonders. Most of the staff took one or two into
their service.
Seven years after, one of my two, who were taken as stable-helps,
was getting pay from me at the rate of thirty shillings a month, and
was my head groom, and would anywhere obtain that pay. Two were
taken as markers in the billiard-room, and are now respectable
servants. As the famine ceased, the unclaimed orphans were
apprenticed to good trades, or placed in the houses of wealthy
Persians as servants. No attempt at proselytism was made, but a
Persian priest was engaged to teach the usual rudiments of reading,
writing, and the Koran.
Many villagers came in and claimed their children, and these were
often loath to leave their clean quarters and good food, to return to
hard drudgery and rags in their native villages.
It may be safely said that no deaths from starvation took place in
Shiraz after the arrival of the first instalment of relief money from
England. Of course, the application of the funds was carried out
irrespective of the religion of the applicants; and this application was
easier in Shiraz than in Ispahan. The Armenian community in Shiraz
were very few, and only some four families needed relief; while, on
the contrary, the Jews were many and terribly poor.
 As to the labour question, a few of the more able-bodied were set
to the nominal work of picking the stones off the high-road, but no
heavy labour was insisted on. In the winter, too, the snow having
blocked the streets, the poor were employed in removing it for the
general good.
I happened to go to Ispahan, and also assisted in the distribution
there. The Ispahanis are much more provident than the people of
Shiraz, and I do not think the distress would have been so great but
for the influx of villagers. At Koomishah, the third stage from Ispahan
towards Shiraz, the effects of the famine were very severe, and I
was glad to be able to distribute some four hundred kerans of the
Poor Fund, both going and coming, there. Of course this amount did
not go far, and I was besieged in the post-house by the hungry
crowd of women and children; the sum was too small to permit of
giving anything to the men. First we admitted all the aged women,
and gave them a keran and a half each; then each child was given a
keran, and, when they had secreted it, the whole number were
passed out and the gates closed. From the roof of the post-house I
perceived a big burly villager, who was employed in robbing the
children, as they went out, of their slender store, even throwing them
on the ground and taking the coin from their mouths. The other
villagers, of whom there was a large mob, merely laughed, but did
not interfere. But getting down from the back wall of the post-house
by means of horse-ropes, the postmaster, my groom and I
succeeded in catching the fellow, and dragging him into the post-
house, and then the post-boys gave him a good hiding by my order,
and we took the money away. He, of course, complained to the local
Governor, who requested an explanation. I called on him and told
him of the fellow’s misdeeds, and, much to his astonishment, the
man in power gave the ruffian a liberal bastinado.
Terrible stories are extant of what happened in certain places, and
there is no doubt of the truth of many of these. That the people ate
grass and the carrion, that they lived on the blood at the public
slaughter-houses, that they, having sold all, also sold their children,
is within my personal knowledge. Cannibalism, too, was proved. In
fine, had it not been for the exertions of the Persian Relief
Committee in London, the ravages of the famine would have only
ended in the temporary depopulation of the south and centre of
Persia.
Each great personage in Ispahan and Shiraz did his best to
preserve his own dependents from starvation; but there being no
kind of organisation among the Persians, and transit-rates being
prohibitive, and the roads unsafe, small local famines were frequent,
and the ravages of typhoid and diphtheria—the latter previously
unknown in the country—were very great.
Just now an accident to the Prince Zil-es-Sultan took place. He
was out shooting near Shiraz, and having charged one barrel of his
gun twice, the weapon burst, tearing the palm of his hand and the
ball of the thumb. I was called in to attend him, and was fortunate
enough to preserve the hand. For this his Royal Highness was very
grateful, and during the whole of my time in Persia showed me many
kindnesses, besides giving me an extremely liberal fee, even for a
king’s son: he compelled his vizier also to give me one. He even
insisted on decorating me with the star of the Lion and Sun; but as
Englishmen in Government employ are not allowed to accept the
decorations of foreign Governments without special permission, the
honour, much coveted among the Persians, was not of much benefit
to me. I got it in a very public and sudden manner, and as the
occasion of giving it was sufficiently curious, I may as well describe
it.
It is the custom in Persia to send to all governors, royal
personages, and ministers, a yearly present from the king, to show
the royal satisfaction. These presents are all termed kalāats (or
dresses of honour), even though the gift may be in jewellery, or even
specie; a dress or robe of greater or less value, or a jewelled
weapon, being the general kalāat. The withholding of the yearly robe
of honour to a provincial governor is generally the sign of the royal
disfavour, and the despatching of it often the token of the recipient’s
confirmation in office, though at times it is what gilds the bitter pill of
his recall. The kalāat is usually sent from the capital by the hands of
some person of consequence, generally some favoured servant of
the Shah, and this man is sent down that he may receive a present,
generally large in amount, from the recipient, and may bring back the
usual bribe to the Prime Minister for retention in power, or even the
same thing to the king himself.
The New Year’s festival is generally the time of the despatch of the
official dresses of honour from the capital. The bearer, and his two or
more attendants, generally come on post-horses, and the etiquette is
that the recipient goes out to meet the royal gift. The bringer, on
arrival at the last stage, is met by the servants or friends of the
recipient, who send off to announce the arrival. He now takes off his
travel-stained garments, puts on his finery, and starts on horses sent
out for him, bearing the royal bounty at his saddle-bow wrapped in a
Cashmere shawl. The recipient, accompanied by all his friends and
the greater portion of the populace—for the bazaar is closed by
order, and a general illumination commanded for the evening; all the
shops are visited, and severe fines inflicted on any one disobeying—
proceeds to meet the present, and await its arrival. The distance that
is gone is regulated by the position of the recipient—the greater the
personage, the less distance he goes.
One morning the prince sent for me and told me that a kalāat from
the Shah would arrive for him the next morning, and that he wished
his hakim-bashi and myself to ride out with the magnates of the
place, who would accompany him, to meet it. I of course expressed
my readiness to attend his Royal Highness, and I was told by the
hakim-bashi, who was very jubilant, that probably a decoration would
be given to each of us. To have declined would have been to give
mortal offence, and to have lost the favour of the Governor of the
province, whose partiality secured me against annoyance from the
natives of any kind. So the next day I presented myself at nearly
noon and found the prince in great feather, the head astrologer
having appointed two in the afternoon for the enduing of the dress of
honour. Every one was in gala dress, the streets were thronged by a
holiday mob in high good humour. And out we all rode. First came
four yessaouls, or outriders, with silver maces, showing off their
horses by capering in circles; then six running footmen, each with his
silver-headed staff and clad in the royal scarlet, in the ancient
costume of Persia, and with the strange head-dress somewhat like a