PhD Thesis Defense

Over the past century, microscopy has evolved significantly through advances in hardware design. However, pushing the boundaries of imaging performance using purely optical and mechanical innovations has become increasingly challenging. Meanwhile, the rapid growth in computational power has transformed the way we process and analyze imaging data, enabling a paradigm shift in microscopy. By offloading complexity from hardware to algorithms, computational microscopy - a rising interdisciplinary field - offers a powerful approach to simplify system design, correct aberrations, and extract more information from data. In this thesis, I will explore how computational techniques, including artificial intelligence, can enhance modern imaging systems and image analysis for biomedical research.

This thesis develops physics-informed computational frameworks that extend the capabilities of optical imaging systems. By embedding physics models directly into reconstruction algorithms, I show how computational approaches can overcome traditional limitations in resolution and depth-of-field, through case studies in Fourier ptychographic microscopy and single-shot volumetric fluorescence imaging. 

In addition to reconstructions, this thesis advances analytic solutions with customized optical designs that leverage physical insight to enable robust, efficient, and automated microscopy. By identifying and exploiting principles in wave optics, analytic methods are developed for high-performance autofocusing and optimization-free volumetric refractive index imaging. These approaches improve robustness across imaging modalities.

Finally, this thesis demonstrates how artificial intelligence can be integrated with microscopic imaging to enable clinically relevant inference. Deep learning models are applied to digitized histopathology slides to predict progression risk in early-stage non-small-cell lung cancer patients, achieving performance that exceeds expert-level assessment. Beyond predictive accuracy, the models are systematically analyzed to identify the spatial feature scales that drive their predictions.