IMAT-ITCR Collaboration: Combining FIBI and topological data analysis: Synergistic approaches for tumor structural microenvironment exploration

点击翻译按钮获取中文摘要

SUMMARY: In response to PA-20-272, we propose a collaborative pilot study between teams at Stony Brook University (ITCR - MPIs Prasanna and Chen) and UC Davis (IMAT - PI Levenson). By leveraging novel mathematical topology tools developed by the ITCR team, we aim to understand and quantify the similarities and differences in the structural microenvi- ronment across Fluorescence Imitating Brightfield Imaging (FIBI) and conventional hematoxylin and eosin (H&E) images. FIBI, an inexpensive slide-free tissue imaging technique developed by the IMAT team, provides immediate high-quality im- ages from fresh or fixed tissues that resemble those generated after time-consuming methods used to prepare traditional H&E slides. A key advantage of FIBI is that it can detect continuous linear structures that are interrupted or poorly visualized on standard slides, a feature with significance to the evaluation of tumor micro-structural environment and clinical diagnostics. However, to advance FIBI as a diagnostic imaging modality, it will be important to gain a deeper understanding of such intricate structural features. While a pilot validation report has shown that FIBI images retain diagnostic power compared to H&E images, a comprehensive quantification of the differences between these two imaging modalities is lacking; further- more, there are no FIBI-specific quantitative histomorphometry tools that can help characterize and quantitatively evaluate different structures that are more salient on thickly vs. thinly cut tissue sections. The ITCR team will adapt topological data analysis (TDA) tools proposed in their R21 project to study the 3D structures in FIBI images generated by the IMAT team in their R33 project. This analysis will focus on fine-scale structures with connectivity and surface-profile features easily appreciable in FIBI images such as collagen bundles and blood vessels. The extracted features will be color coded and mapped onto the FIBI images for interpretable visualization to establish comprehensive taxonomies for the discovered topology profiles. Analysis will include at least 100 FIBI samples, each containing FIBI and H&E images, with comparisons between the topological features extracted from images to quantify the differences in describing the structural microenvironment. Expert segmentations of specific structures of interest within FIBI images will be obtained, followed by the extraction of descriptors to characterize topology. The relationships between structures and regions of interest, such as cancer and normal regions, and different subtypes of cancer, will be investigated using statistical techniques and predictive models. The study will enhance the understanding of FIBI (IMAT team), as a diagnostic imaging modality and refine topological analysis methodology (ITCR team). Deliverables will include a set of tools and techniques for holistic characterization of the structural environment as observed on the FIBI scans. The effort will leverage the teams’ expertise in mathematical tools and image analysis, building upon previous collaborations on breast cancer image analysis. Successful identification of meaningful phenotype-feature associations will demonstrate the clinical utility of the FIBI technique, particularly as a diagnostic tool for guiding treatment decisions.