High-Throughput Microscopy via Ptychographic Imaging

Abstract

The core ideas of ptychographic imaging grew out of coherent diffraction imaging (CDI),which was originally developed for measuring periodic crystalline structures in the last century. As interest shifted toward recovering phase changes introduced by a sample under coherent illumination, researchers began to reconstruct the complex-valued object (amplitude and phase) from measured diffraction intensities. These concepts were soon extended beyond periodic crystals to a much wider range of applications, including biomedical imaging, complex materials, and optical metrology. The term ptychography comes from the Greek word ptych¯e, meaning “fold,” reflecting the use of overlapping measurements. The central idea of ptychography is to exploit redundancy in a series of overlapping intensity measurements to recover information that conventional optical systems cannot directly provide, such as quantitative phase and resolution beyond the native limit of the imaging optics. The demand for high-throughput microscopy spans many fields, from biomedical imaging to industrial inspection, and ptychography has emerged as a robust approach for extending imaging performance beyond conventional optical microscopes. This dissertation aims to demonstrate ptychographic imaging as a unifying framework for high-throughput microscopy across a broad range of applications. By recording only intensity measurements and computationally recovering the complex object, ptychographic methods can overcome limitations of traditional optical systems in biomedical imaging, metrology, materials characterization, and large-scale circuit inspection. The dissertation begins with an introduction and background on ptychographic imaging. Chapter 1 presents the historical development of ptychography from CDI, reviews prior research and its evolution into many specialized variants tailored to different applications, and summarizes key advances relevant to high-throughput microscopy. The chapter concludes by positioning ptychographic imaging within the broader landscape of computational imaging methods, highlighting related techniques and discussing their respective advantages and trade-offs for different imaging tasks. Chapter 2 presents a self-calibrating FPM framework that addresses one of the most critical challenges in practical FPM, namely model mismatch due to system misalignment and other experimental imperfections. We introduce an algorithmic framework based on automatic differentiation (AD) to jointly correct illumination misalignment and other forward-model discrepancies, thereby greatly enhancing the robustness of FPM in real-world settings and opening a more viable path toward routine deployment. The capability of the proposed selfcalibrating FPM framework is demonstrated through both numerical simulations, where it is compared against conventional FPM reconstruction methods, and experimental validation on cervical cell samples. In particular, we show deep ultraviolet (DUV) label-free imaging of cervical cells over an extremely large FOV, illustrating the potential of the proposed self-calibrating FPM approach for high-throughput biomedical imaging applications. Chapter 3 extends the self-calibrating FPM framework to a compact reflective RFPM architecture operating in the DUV regime. The newly designed reflective configuration enables the acquisition of both bright-field and dark-field images with a significantly simpler experimental layout than previous RFPM implementations, reducing the number of optical components and alignment degrees of freedom. This compact geometry makes it easier to integrate deep-ultraviolet Fourier ptychographic microscopy (DUV-FPM) into different platforms for surface metrology, while maintaining nanometer-scale height sensitivity over an extremely large FOV. We demonstrate the compact DUV-FPM system on semiconductor chip standards, highlighting its capability for defect detection in potential large-area integrated circuit inspection. In addition, we characterize mirror surfaces, including surface roughness and machining traces, illustrating the potential of the proposed RFPM configuration for high-throughput, quantitative optical metrology. In Chapter 4, we focus on coded ptychography (CP) for high-throughput lensless imaging, a relatively recent development in the ptychographic imaging family. We first introduce the general forward model, the initial experimental setup, and several potential application scenarios. We also compare FPM with CP to emphasize the conceptual and practical differences between these two approaches, and to clarify how they are related. Finally, we present experimental results demonstrating large–FOV biomedical imaging, including highthroughput screening of red blood cells (RBCs) and cervical cells, highlighting the strengths and limitations of CP in realistic biomedical settings. In Chapter 5, we introduce a recently developed CP approach. We propose and demonstrate a cost-effective, motionless lensless imaging system based on CP for high-throughput microscopy. A new reconstruction algorithm is developed that is specifically tailored for motionless CP using a 3D-printed phase mask. The design, fabrication, and integration of the customized phase mask into the imaging device are described in detail. The resulting lensless architecture requires minimal calibration, making it well-suited for portable and adaptable imaging platforms in diverse application scenarios. This approach enables imaging beyond conventional limits, achieving millimeter-scale FOV with micron-scale resolution without using lenses or any mechanical scanning. The motionless CP system can also capture full red-green-blue (RGB) color reconstructions and phase imaging simultaneously. Experimentally, we demonstrate the proposed setup on murine organ histology slides and diffractive optical elements (diffractive optical elements (DOEs)), illustrating the broad applicability of the system to both biomedical imaging and structured optical components. Relevant prior work in coded and lensless computational imaging is systematically compared and summarized in a table, providing context for the proposed method and outlining directions for future research. Finally, Chapter 6 discusses future directions for advancing ptychographic imaging, with a focus on making the methods developed in this dissertation more practical and broadly applicable. Building on the self-calibrating FPM, compact DUV-FPM, and motionless CP frameworks presented in the previous chapters, we outline strategies to further reduce the reliance on strict prior knowledge and precise calibration of the imaging system. These directions are essential for translating ptychographic methods into robust, user-friendly tools for high-throughput microscopy in real-world settings. Chapter 5 also summarizes the main remaining technical challenges and practical obstacles, and discusses potential avenues to overcome them in future work.