Topology-based tumor analysis for medical images

点击翻译按钮获取中文摘要

Project Summary/Abstract Tumor shapes and patterns have been used as important markers for cancer diagnosis and treatment. Recent developments in medical imaging technology have enabled a more detailed description of tumor regions in high resolution. However, existing studies have described the shape of tumors in a limited scope. There is a scientific need to enhance understanding of tumor metastasis using medical images and provide a new insight for medical decision-making. This project aims to develop topological tumor shape representations for different types of medi- cal images and provide model-based approaches based on the topological image features. Our preliminary results suggest that topological features of medical images capture shapes and patterns of tumor regions and predict prog- nosis and survival after controlling key clinical parameters. The proposed project will further develop topology-based tumor analysis methods for pathology and radiographic images and provide tools to aid medical decision-making. The objective of the project will be accomplished by three aims: (1) develop methodologies to pair spatial and shape information of tumors and investigate relationships between genomic characteristics and topological features for three-dimensional gliomas of radiographic images; (2) develop topological tumor shape analysis methods that analyze shapes and interactions of multiple cell-type regions and extract image size-invariant shape representa- tions for two-dimensional lung adenocarcinoma pathology images; and (3) provide an accessible resource to the research community by offering user-friendly software and education. Our applications using lung adenocarcinoma pathological images and magnetic resonance imaging of primary gliomas images will provide prognostic information beyond standard clinical factors and serve as strong evidence for clinical usage of topological tumor shape analysis. This proposal will also allow students from various backgrounds at Southern Methodist University to experience a broad spectrum of medical research, including but not limited to genomic and cellular processes involved in tumors, survival modeling, topological data analysis, and deep learning in medical imaging while working with collaborative researchers from regional institutions.