NOVEL TECHNIQUES IN FOURIER PTYCHOGRAPHIC MICROSCOPY FOR MEDICAL IMAGING

Abstract

The emerging Fourier ptychographic microscopy (FPM) offers a highly efficient approach to enhance the throughput of conventional microscopy. By illuminating the samples with a series of incident angles, FPM computationally reconstructs the high-resolution sample images from the acquired low-resolution measurements. Thus, it overcomes the limitations of traditional optical imaging systems to simultaneously achieve large field of view and adequate spatial resolution. However, to maximize its potential, it is necessary to continuously optimize the performance and verify the clinical utility of FPM system. For this purpose, this dissertation is majorly composed of four different studies. In the first study, we initially explore the feasibility of implementing FPM to reconstruct metaphase chromosomes, aiming to enhance the imaging efficiency by mitigating the trade-off between field of view (FOV) and resolution in conventional microscopic systems. The second study focuses on improving the efficiency of data acquisition of FPM. By employing symmetric illumination and a color detector to accelerate the process, we can potentially increase the data acquisition speed up to 12 times. In the third study, we develop and evaluate a doublet based FPM prototype to further advance the design of more affordable FPM configurations. This prototype replaces the standard 4× objective lens with a commercial achromatic doublet lens of 60mm focal length. Using a 150mm tube lens and a 15×15 LED array, it achieves a theoretical equivalent NA of 0.4. The fourth study is designed to measure the DOF of FPM systems, providing essential insights for the future development of FPM based digital pathology scanners. The measurements follow the principle that DOF is the range along optical axis where the contrast value remains at or above 80% of the maximum as the focus is altered. The corresponding contrast value for each focus position is estimated based on the specific bar pattern where the contrast value of the in-focus MTF curve drops to 0.5. The four studies on FPM investigate its applications, technological advancements, and evaluations, providing a comprehensive exploration of this new technology's potential and supporting its translation toward practical implementation.In addition to the FPM studies, the appendix presents two projects dedicated to computer-aided diagnosis (CAD) scheme developments. These studies utilize radiomics and deep learning techniques to enhance the prediction of chemotherapy outcomes in ovarian cancer treatment with CT scans. The investigations underscore the value of integrating analytical approaches with medical imaging modalities, illustrating how CAD contributes to personalized healthcare through improved diagnostic and prognostic capabilities.