Spatial and Material Resolution in Coded-Aperture Volumetric X-ray Diffraction Imaging for Biomedical Applications

Abstract

The histological analysis of surgically resected tissue is constrained by the current tools and standards of clinical practice. The labor-intensive process of converting tissue samples into stained glass slides reduces the inherently three-dimensional structure of the tissue to a two-dimensional representation. Furthermore, the significant resource demands limit the amount of tissue that can be processed, leaving much of the sample entirely unexamined. Both aspects risk discarding critical information, thereby reducing diagnostic precision. These limitations are addressed by developing the XR4D, a novel volumetric X-ray diffraction imaging (XRDI) system designed for high-resolution, material-specific, and nondestructive imaging of biomedical tissue samples. This dissertation presents experimental and computational groundwork underpinning the XR4D's development and establishes methods and standards for evaluating its performance, offering a framework for characterizing this innovative imaging technology, and progressing towards a clinical solution for challenges hindering surgical pathology.Many design choices of the XR4D result from experimental work with a prior system, named the XR3D. After assessing and contributing to relevant resolution theories, experiments were conducted using custom-built 3D-printed phantoms to analyze the XR3D imaging capabilities. The first study quantified spatial resolution as correlated to geometric changes in system configuration and highlighted the volumetric imaging capability of a system initially designed for 2D projection applications. The second study builds on these results by reconstructing a multi-material 3D-phantom and using machine-learning classification techniques to describe each materials spatial distribution. These studies also identified spectral-spatial resolution tradeoffs and resolution limitations of detector-side coded-aperture architectures, which motivated the structured illumination (SI) design that was implemented in XR4D. MATLAB numerical simulations confirmed imaging performance tradeoffs in SI architectures, and supported design choices that strived to combine high resolution and imaging efficiency. The XR4D was subsequently designed and developed by Quadridox Inc. with the final task of the work being its characterization. Key results demonstrated axial resolution at 1 mm, and lateral resolution down to at least 0.5 mm, with the caveat that fabrication limitations set an experimental floor. Additionally, iterative or aliasing complications resulted in substantial signal fluctuations at a singular pixel level, giving a lateral error of ± 0.25 mm. Spectral analysis revealed a robust form factor reconstruction, maintaining a correlation coefficient of about 0.95 over temporally and spatially dispersed measurements. Two different 3D-printer base plastics were classified in combined datasets at about 80% accuracy, using only a mean reference spectrum, and the two having their own correlation coefficient of 0.93, implying their high degree of ground truth similarity. To the extent they were measured, the XR4D performance meets goals initially set forth, providing a convincing first step towards its potential future applications, of high spatial and spectral resolution volumetric datasets of biological specimen.