NephCure Funded Research 2017

Dr. Madabhushi will study if computational annotation of primitives, attained with this pilot study, will enable discovery of visual and sub-visual features that can be used in future studies for discovery of molecular mechanisms and therapeutic targets, and clinically relevant categorization of NS. Consequently, we propose to develop and test foundational Apps (tool-box) of the NEPTUNE Digital Pathology-based Analytic hiSto-omicS intErrogation plaTform (NDP-ASSET) for segmentation of normal and pathologic (primary and secondary) classes of structural primitives. Convolutional neural network-based, deep learning (DL) algorithms will be trained for machine-based annotation of primitives in WSI of normal and diseased kidneys, and subsequently validated in an independent replication set. To demonstrate utility of the deeply annotated structural data, we will develop and apply tools for sub-visual feature extraction to test whether the sub-visual features can predict APOL1 risk genotypes in African American NEPTUNE patients.