Artificial intelligence in digital holographic microscopy : technical basis and biomedical applications / Inkyu Moon.

Includes bibliographical references and index.

Coherent optical imaging -- Lateral and depth resolutions -- Phase unwrapping  -- Off-axis digital holographic microscopy -- No-search focus prediction in DHM with deep learning -- Deep learning model -- Noise-free phase imaging in Gabor DHM with deep learning -- Red blood cells phase image segmentation -- Red blood cells phase image segmentation with deep learning -- Automated phenotypic classification of red blood cells -- Automated analysis of red blood cell storage lesions -- Automated red blood cells classification with deep learning -- High-throughput label-free cell counting with deep neural networks -- Automated tracking of temporal displacements of red blood cells -- Automated quantitative analysis of red blood cells dynamics -- Quantitative analysis of red blood cells during temperature elevation -- Automated measurement of cardiomyocytes dynamics with DHM -- Automated analysis of cardiomyocytes with deep learning -- Automatic quantification of drug-treated cardiomyocytes with DHM -- Analysis of cardiomyocytes with holographic image-based tracking -- Conclusion and future work.

"Real-time automated identification of pathogenic micro/nano biological organisms or other specimens has many potential applications in security and defense or health related applications. Developing reliable, automated, and low-cost methods for real-time sensing, monitoring, and identification of harmful pathogens or malignant cells is beneficial in combating catastrophic pandemics, providing disease detection and monitoring for emerging medical treatment procedures, food safety, environmental health and safety monitoring. Conventional methods used to inspect and identify bacteria and other biological species often involve labor-intensive and time-consuming biochemical and/or biomolecular processing. Optical imaging systems based on digital holography and integral imaging have been extensively investigated for 3D visualization and recognition of rigid, macro objects. However, biological organisms are typically non-rigid and exhibit dynamic behavior such as moving, dividing and growing. This makes it difficult to identify biological species based on their shape, size or morphology in conventional 2D imaging. Moreover, many unicellular biological species such as bacteria, yeast or protozoans appear essentially transparent under bright field microscopes unless the specimen is stained and/or fixed: a process in which the cells are killed and dynamics cannot be studied. Meanwhile, 2D intensity images of the microorganisms are usually insufficient for identification or visualization of transparent microorganism parts, e.g. sperm tails. Therefore, developing high-speed, low-cost and reliable system for three-dimensional (3D) analysis, visualization, identification and monitoring of harmful pathogens or biological cells are essential"--