Quantitative Phase Imaging of Biological Samples using Optical Coherence Tomography

Abstract

Advancement in the diagnostic techniques is required for early detection of the disease which will avoid the many risks to patients. This thesis research work describes a novel imaging technique for quantitative phase imaging of the biological samples. We developed a full-field optical spatial coherence microscopy (FF-OSCM) system based on monochromatic laser. The system is characterized in terms of axial resolution, lateral resolution, phase sensitivity and phase stability. The developed system exploits the property of spatial coherence and its performance is comparable to the conventional optical coherence microscopy based on temporal coherence. The system is used for the quantification of different stages of malaria infected red blood cells (RBCs) through a fully-automated computer-aided system. The system further modified to study the different stages, especially early and late trophozoite of malaria with limited labelled data size using the customized convolutional neural networks (CNNs). The results were also compared with commonly known CNNs and shows that our automated system has a comparable performance with less computational time. We also develop an automated algorithms for the classifications of the human burnt skin injuries in vivo, and margin assessment of the breast cancer tissues using optical coherence tomography (OCT) images. Our proposed automated procedure entails building a machine learning based classifier by extracting quantitative features of normal and burn tissue images recorded by OCT and obtained good sensitivity and specificity. Our results show the capability of a computer-aided technique for accurately and automatically identifying burn tissue resection margins during surgical treatment. Furthermore, the study was performed in the classification of the human breast cancer tissues using OCT images. We developed an automated algorithm based on a pretrained CNN (Inceptionand#8208;v3) architects with reverse active learning for the classification of healthy and malignancy breast t