Pittsburgh Computational Pathology Interest Group and Lecture Series aims to bring focus on developing basic computer vision and machine learning algorithms for transmitted light and fluorescence microscopy imaging data from histopathology and on highlighting the translation role of computational pathology research in interfacing with genomics, personalized medicine and microbiomics.
Our goal is to promote a vibrant community of computational pathology research and education practices here in Pittsburgh by engaging researchers and clinicians across Pitt, CMU and Duquesne campuses, and the local industrial partners.
"Nanoscale nuclear architecture mapping (nanoNAM) for cancer-risk stratification and prediction"
The current standard in patient care for populations at risk of developing cancer requires screening of the entire at-risk cohort. In certain populations - for example, patients with chronic inflammatory bowel disease (IBD) colitis - although there is an increased risk of colorectal cancer, the majority of patients will not develop cancer. For such patients, the clinical gold standard of microscopic histological examination of their tissue biopsy may appear histologically non-dysplastic because it does not have enough sensitivity to identify dysplasia in early-stage carcinogenesis. The resulting inability in identifying the subset of patients that will go on to develop cancer has negative impact on patient surveillance and preventive care. In this talk I will present nanoNAM-based optical approach we have developed that is able to overcome this sensitivity barrier by mapping the nuclear architecture with nanometer-scale sensitivity. I will discuss its underlying mathematical theory and the computational approaches that make the nanoscale characterization possible. I will present results showing how such characterization allows nanoNAM to track malignant transformation in murine model of colon carcinogenesis, and identify cancer progression in IBD patients with ulcerative colitis. I will also present preliminary results that identify low-dimensional manifolds associated with this differential progression. Finally, I will present a classifier that allows nanoNAM to predict cancer progression risk.