Abstract #2384

The Effects of Ground Truth Variance on Radio-Pathomic Mapping in Prostate Cancer

Sean D McGarry¹, John D Bukowy², Kenneth A Iczkowski³, Wei Huang⁴, Tatjana Antic⁵, Gladell Paner⁵, Allison K Lowman², Tucker Keuter⁶, Anjishnu Banerjee⁶, Alex Barrington², Samuel Bobholz¹, Petar Duvnjak², Michael Griffin², Mark Hohenwalter², Kenneth Jacobsohn⁷, and Peter S LaViolette²

¹Biophysics, Medical College of Wisconsin, Wauwatosa, WI, United States, ²Radiology, Medical College of Wisconsin, Wawautosa, WI, United States, ³Pathology, Medical College of Wisconsin, Wawautosa, WI, United States, ⁴Pathology, University of Wisconsin Madison, Madison, WI, United States, ⁵Pathology, University of Chicago, Chicago, IL, United States, ⁶Biostatistics, Medical College of Wisconsin, Wawautosa, WI, United States, ⁷Urological Surgery, Medical College of Wisconsin, Wawautosa, WI, United States

achine learning provides a framework for non-invasively extracting more information from a clinical prostate scan by leveraging aligned post-surgical tissue samples with in-vivo imaging to create predictive models of histological characteristics. Many of these algorithms rely on a pathological diagnosis as the ground truth for the classification or regression task. This study aims to investigate the effects of varying the ground truth label in generating voxel-wise radio-pathomic maps of epithelium and lumen density in prostate cancer.

This abstract and the presentation materials are available to members only; a login is required.

Join Here