Artificial Intelligence in Digital Holographic Imaging E-Book

Inkyu Moon

• Herausgeber: John Wiley & Sons
• Kategorie: Wissenschaft und neue Technologien
• Serie: Wiley Series in Biomedical Engineering
• Sprache: Englisch

Seitenzahl: 439

Veröffentlichungsjahr: 2022

Beschreibung

Artificial Intelligence in Digital Holographic Imaging Technical Basis and Biomedical Applications An eye-opening discussion of 3D optical sensing, imaging, analysis, and pattern recognition Artificial intelligence (AI) has made great progress in recent years. Digital holographic imaging has recently emerged as a powerful new technique well suited to explore cell structure and dynamics with a nanometric axial sensitivity and the ability to identify new cellular biomarkers. By combining digital holography with AI technology, including recent deep learning approaches, this system can achieve a record-high accuracy in non-invasive, label-free cellular phenotypic screening. It opens up a new path to data-driven diagnosis. Artificial Intelligence in Digital Holographic Imaging introduces key concepts and algorithms of AI to show how to build intelligent holographic imaging systems drawing on techniques from artificial neural networks, convolutional neural networks, and generative adversarial network. Readers will be able to gain an understanding of the basics for implementing AI in holographic imaging system designs and connecting practical biomedical questions that arise from the use of digital holography with various AI algorithms in intelligence models. What's Inside * Introductory background on digital holography * Key concepts of digital holographic imaging * Deep-learning techniques for holographic imaging * AI techniques in holographic image analysis * Holographic image-classification models * Automated phenotypic analysis of live cells For readers with various backgrounds, this book provides a detailed discussion of the use of intelligent holographic imaging system in biomedical fields with great potential for biomedical application.

Leseprobe

Artificial Intelligence in Digital Holographic Imaging

Technical Basis and Biomedical Applications

Inkyu Moon

Department of Robotics and Mechatronics Engineering

Daegu Gyeongbuk Institute of Science & Technology (DGIST)

Daegu, South Korea

This edition first published 2023© 2023 John Wiley & Sons, Inc.

All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, except as permitted by law. Advice on how to obtain permission to reuse material from this title is available at http://www.wiley.com/go/permissions.

The right of Inkyu Moon to be identified as the author of this work has been asserted in accordance with law.

Registered OfficeJohn Wiley & Sons, Inc., 111 River Street, Hoboken, NJ 07030, USA

Editorial Office111 River Street, Hoboken, NJ 07030, USA

For details of our global editorial offices, customer services, and more information about Wiley products visit us at www.wiley.com.

Wiley also publishes its books in a variety of electronic formats and by print‐on‐demand. Some content that appears in standard print versions of this book may not be available in other formats.

Limit of Liability/Disclaimer of WarrantyIn view of ongoing research, equipment modifications, changes in governmental regulations, and the constant flow of information relating to the use of experimental reagents, equipment, and devices, the reader is urged to review and evaluate the information provided in the package insert or instructions for each chemical, piece of equipment, reagent, or device for, among other things, any changes in the instructions or indication of usage and for added warnings and precautions. While the publisher and authors have used their best efforts in preparing this work, they make no representations or warranties with respect to the accuracy or completeness of the contents of this work and specifically disclaim all warranties, including without limitation any implied warranties of merchantability or fitness for a particular purpose. No warranty may be created or extended by sales representatives, written sales materials or promotional statements for this work. The fact that an organization, website, or product is referred to in this work as a citation and/or potential source of further information does not mean that the publisher and authors endorse the information or services the organization, website, or product may provide or recommendations it may make. This work is sold with the understanding that the publisher is not engaged in rendering professional services. The advice and strategies contained herein may not be suitable for your situation. You should consult with a specialist where appropriate. Further, readers should be aware that websites listed in this work may have changed or disappeared between when this work was written and when it is read. Neither the publisher nor authors shall be liable for any loss of profit or any other commercial damages, including but not limited to special, incidental, consequential, or other damages.

Library of Congress Cataloging‐in‐Publication Data applied forHardback ISBN: 9780470647509

Cover Design: WileyCover Images: © KATERYNA KON/SCIENCE PHOTO LIBRARY/Getty Images

Preface

Quantitative label‐free optical imaging technique represents a new, highly promising approach to identify cellular biomarkers, particularly when it is combined with artificial intelligence (AI) technologies for scientific, industrial, and biomedical applications. Among several new optical quantitative imaging techniques, digital holographic microscopy (DHM) has recently emerged as a powerful new technique well suited to non‐invasively explore cell structure and dynamics with a nanometric axial sensitivity and the ability to identify new cellular biomarkers. This book provides detailed explanations for using DHM to perform label‐free phenotypic cellular assays, thus allowing the non‐invasive isolation of different specific cellular phenotypes. Practically, phenotypes related to the monitoring of cell responses and cytotoxicity profiling upon interaction with drugs are presented. Thus, promising theragnostic cellular biomarkers can be successfully explored. This book further provides explanations of AI and deep learning pipelines for the development of an intelligent DHM that can perform optical phase measurement, phase image processing, feature extraction, and classification. Multiple biophysical single‐cell features such as morphological parameters, optical loss characteristics, and protein concentration are automatically measured in individual biological cells. These biophysical measurements form a hyper‐dimensional feature space in which supervised learning can be performed for cell analysis. This technology is undergoing clinical testing for blood screening and live cardiomyocytes analysis as well as for studying neuronal activities in mental diseases including psychiatry disorders. Furthermore, combining DHM with stem‐cell technology including induced pluripotent stem cell (iPSC) approaches paves the way to develop personal medicine, considering that iPSCs derived from a patient can be differentiated a priori into any types of cells, including cardiac cells, neural cells, and so on. However, the development of systems and methods for performing high‐throughput analysis of holographic images in a large volume are indispensable to have a DHM‐based automated high‐content screening approach aiming at identifying theragnostic cellular biomarkers. By combining DHM with AI technology, including recent deep learning approaches, this system can achieve a record‐high accuracy in non‐invasive, label‐free cellular phenotypic screening. It opens up a new path to data‐driven diagnosis. Specifically, AI is one of the most rapidly evolving subjects in computing and engineering fields, with a special emphasis on creating intelligent automated systems or applications. These AI algorithms are becoming essential for developing intelligent systems. The main goal of this book is to explain key concepts and algorithms of AI to show how to build intelligent DHM systems drawing on techniques from artificial neural networks (ANN), convolutional neural networks (CNN), and generative adversarial network (GAN). Principles behind these techniques are explained by showing how various techniques can be implemented for intelligent DHM systems design. Depending on problems to be solved, AI algorithms can be applied to recognition, classification, regression, and prediction problems. This book describes representative algorithms for each problem with some good examples and how to implement intelligent DHM systems with ANNs, CNNs, and GANs on the computer. Furthermore, this book gives details of the deep learning CNN to automatically reconstruct the best focused images in DHM. It describes GAN models to eliminate superimposed twin‐image noise in the phase image of Gabor holography. It also introduces a deep learning model to compute an unwrapped phase solution in DHM. This book brings together the literature addressing biomedical applications of DHM combined with AI algorithms (e.g. drug safety testing and compounds selection as a new paradigm for drug toxicity screening) to present recent achievements in this interesting field. For readers with various backgrounds, this book provides a detailed discussion of the use of intelligent DHM in biomedical fields with great potential for biomedical application. This book provides two representative examples of applying intelligent DHM in biomedical fields. The first example describes how instant phenotypic assessment of red blood cells (RBCs) storage lesion can be automatically performed by AI based‐DHM, which has the potential to lead to new efficient tools for safe transfusions as well as measurement of stored RBC quality. The second example demonstrates that relevant dynamic parameters of cardiomyocytes can be obtained by DHM phase signal analysis based on AI algorithms to characterize the physiological state of live cardiomyocytes. This finding opens the possibility of automated quantitative analysis of cardiomyocytes suitable for further monitoring some specific drug mediated effects on the dynamics of cardiomyocytes, which represents a promising label‐free approach for drug discovery. Therefore, my intention in writing this book, is to introduce AI‐based DHM background with detailed description of these two examples in biomedical fields to make it easier for readers with diverse backgrounds to read this book. I hope that readers will be able to gain an understanding of the basics for implementing AI in DHM designs and connecting practical biomedical questions that arise from the use of DHM with various AI algorithms in intelligence models.

Part IDigital Holographic Imaging

1Introduction

Biomedical imaging technologies promise opportunities for effective diagnostics and treatments as well as new economic perspectives for the medical technology industry. Increasing demand for accurate early diagnosis and the growth of an aging population are major forces driving the biomedical imaging market. An increase in the number of patients suffering from chronic diseases, such as cancer, and a demand for early diagnosis and treatment are also major factors fueling the growth of medical diagnostics and medical therapeutics.

Cellular imaging remains one of the most important techniques to solve major challenges in life sciences and medicine. Optical microscopy is one of the most productive scientific tools in cell imaging. The identification of microorganisms and cells was explored for the first time in the nineteenth century, which could be considered the beginning of the emergence of modern biology and medicine. However, optical microscopy still faces two major problems. One is resolution limitation due to Abbe's law. Another is the lack of quantitative information due to the inherent limits of conventional optical microscopes. Conventional intensity‐based imaging techniques are not robust enough to provide detailed quantitative information for cell morphology. They provide low‐contrast images, especially when investigating cells with transparent or semi‐transparent features, which makes it difficult to analyze cells.

Consequently, several optical imaging modalities based on contrast mechanisms were developed to overcome these limitations. Among many contrast‐generating modes, the Zernike phase contrast (PhC) mode and Nomarski differential interference contrast (DIC) are widely used for live‐cell imaging. Unlike fluorescence‐based imaging techniques, PhC and DIC can visualize transparent specimens, particularly subcellular structures of living cells, without using a specific staining contrast agent. However, these two non‐invasive modes cannot provide a direct or quantitative measure of phase shift or the optical path length over a cell area. Since PhC or DIC signals contain only qualitative information, it is difficult to further perform quantitative analysis of the biophysical properties of live cells with reconstructed cell images (see Figure 1.1).

Figure 1.1 Quantification comparison of digital holographic microscopy and other optical imaging modalities.

On the other hand, interference microscopy can provide a direct quantitative measurement of the optical path length based on interference between the reference wave and the object wave that passes through the specimen. Interference microscopy was introduced in the 1950s. Gabor proposed the concept of holography in 1948, which enabled lens‐less imaging by reproducing the exact wavefront emerging from the observed specimen. However, due to the non‐availability of coherence light sources such as lasers and the high cost of opto‐mechanical designs, only a few studies of imaging of live cells are reported in the literature.

In recent years, fluorescence microscopy in a confocal configuration, and its extension into multi‐photon fluorescent excitation, have been widely used for cell imaging in biology among various contrast‐generating modes. However, they cannot provide any information about dielectric properties in terms of the underlying biological functions of live cells. Fortunately, these dielectric properties can be measured using new digital holographic imaging approaches that have recently emerged as promising techniques for the accurate, quantitative visualization of cell structure and dynamics in a non‐invasive manner. To accurately observe live cells without disturbing them is tremendously important. For example, when the aim is to assess drug‐mediated cellular effects.

The rapid development of computing technologies and scientific advances in light sources has opened up a new opportunity in the field of holography and interferometry. Integrating techniques in holography with numerical processing has led to the development of digital holographic microscopy (DHM) with a nanometric axial sensitivity that provides a reliable and quantitative phase signal observed in live cells [1–5]. Therefore, the fusion of DHM and information technologies offers an automatic, low‐cost, and reliable tool to identify various cell types, including protozoa, bacteria, plant cells, blood cells, nerve cells, stem cells, and cardiomyocytes. DHM allows scientists [2] to observe the growing process of the cell as close to natural conditions as possible, whether in a cell‐culture flask or the tissue environment. Moreover, by numerical reconstruction, DHM offers unique possibilities including an extended depth of focus and a posteriori numerical refocusing, which allow quantitative and non‐invasive analysis of cell structure, contents, and dynamics with different time scales that vary from a few milliseconds to several days. Another advantage of DHM is that images of both single cells and populations can be obtained. However, application of DHM to the field of cell biology is still in its beginning stages. There are few studies on the automated quantitative analysis of large‐scale holographic cell datasets. Therefore, systems and methods for performing intelligent analyses of large‐scale holographic images are becoming more important as digital holographic information rapidly increases with the development of DHM technology. The development of automated procedures to study various cell types using DHM can significantly benefit cell biology studies. Until now, most experiments using DHM in its single mode were performed to prove that the technique is useful. Practically, DHM can measure relevant biophysical cell parameters including absolute volume, dry mass density, nanoscale membrane fluctuations, and biomechanical properties. Furthermore, the development of DHM in a multimodal platform with the fusion of DHM and other cell imaging methods such as fluorescence confocal imaging opens up the possibility to simultaneously measure a large number of relevant and specific cell parameters, which can help scientists understand cellular processes including cell differentiation, cell cycle, apoptosis, and cell migration. For example, DHM systems integrated with fluorescence microscopy can add valuable information about cell morphology and motility with a broad variety of fluorescence labeling tools to study cell function. Measurements of multiple cell parameters resulting from multimodal imaging could even enable high throughput cellular screening assays to identify cell biomarkers for various diseases. The time has come to develop intelligent DHM in a multimodal platform and apply multimodal holographic cell imaging informatics to medical and biological research.

To summarize, the DHM enables label‐free, quantitative assessment of biological specimens. There is recent growth in the study of techniques and applications of DHM to address important biomedical questions that cannot be solved with conventional optical imaging techniques. This rapidly emerging field enables scientists to investigate cells and tissues in terms of their morphology and dynamics at a nanoscale resolution over temporal scales ranging from milliseconds to days. Quantitative measurements of intrinsic optical, chemical, and mechanical properties are likely to yield a new understanding of cell and tissue pathophysiology.

Before explaining key concepts of DHM and its success in biomedical imaging and applications, Chapter 2 provides an introductory background on DHM with a quick review and summary of scalar diffraction theory, coherent imaging, and diffraction‐limited imaging. In Chapter 3, fundamental definitions and concepts in lateral and depth resolutions in optical imaging are described. Phase information obtained by DHM provides principal values wrapped in a range of ‐π to π, which can cause 2π phase jumps due to phase periodicity of trigonometric functions. A phase unwrapping process must be conducted to remove 2π phase discontinuities in the image and obtain an estimate of the true continuous phase image. Chapter 4 provides a review of phase unwrapping algorithms to solve challenging problems such as phase discontinuities. Advanced unwrapping algorithms can be categorized into three types: global algorithms, region algorithms, and path‐following algorithms. This chapter further gives a detail explanation of the key concepts of quality‐guided path‐following algorithms. Chapter 5 introduces off‐axis DHM in an example transmission of the quantitative visualization of phase objects such as living cells. This chapter also presents a detailed description of the numerical reconstruction procedure in DHM. In addition, this chapter shows that the transverse resolution is equal to the diffraction limit of the imaging system. Chapter 6 introduces Gabor DHM with a simple optical setup as a promising tool for measuring the distribution(s) of particles in a liquid solution with large depths. It demonstrates that the Gabor DHM can resolve the locations of several thousand particles and measure their motions and trajectories with time‐lapse imaging.

References

1

Marquet, P., Rappaz, B., Magistretti, P. et al. (2005). Digital holographic microscopy: a noninvasive contrast imaging technique allowing quantitative visualization of living cells with subwavelength axial accuracy.

Opt. Lett.

30: 468–470.

2

Moon, I., Daneshpanah, M., Javidi, B., and Stern, A. (2009). Automated three‐dimensional identification and tracking of micro/nanobiological organisms by computational holographic microscopy.

Proc. IEEE

97: 990–1010.

3

Cuche, E., Marquet, P., and Depeursinge, C. (1999). Simultaneous amplitude and quantitative phase contrast microscopy by numerical reconstruction of Fresnel off‐axis holograms.

Appl. Opt.

38 (34): 6994–7001.

4

Anand, A., Moon, I., and Javidi, B. (2017). Automated disease identification with 3‐D optical imaging: a medical diagnostic tool.

Proc. IEEE

105: 924–946.

5

Moon, I., Daneshpanah, M., Anand, A., and Javidi, B. (2011). Cell identification computational 3‐D holographic microscopy.

Opt. & Photon. News

22: 18–23.