Meeting Banner
Abstract #0706

Radio-pathomic mapping models trained with annotations from multiple pathologists reliably distinguish high-grade prostate cancer

Sean D McGarry1, John D Bukowy2, Kenneth Iczkowsk3, Allison K Lowman2, Michael Brehler2, Samuel Bobholz1, Alex Barrington2, Kenneth Jacobsohn4, Jackson Unteriner2, Petar Duvnjak2, Michael Griffin2, Mark Hohenwalter2, Tucker Keuter5, Wei Huang6, Tatjana Antic7, Gladell Paner7, Watchareepohn Palanghmonthip3,8, Anjishnu Banerjee5, and Peter S LaViolette2
1Biophysics, Medical College of Wisconsin, Milwaukee, WI, United States, 2Radiology, Medical College of Wisconsin, Milwaukee, WI, United States, 3Pathology, Medical College of Wisconsin, Milwaukee, WI, United States, 4Urological Surgery, Medical College of Wisconsin, Milwaukee, WI, United States, 5Biostatistics, Medical College of Wisconsin, Milwaukee, WI, United States, 6Pathology, University of Wisconsin Madison, Madison, WI, United States, 7Pathology, University of Chicago, Chicago, IL, United States, 8Pathology, Chiang Mai University, Chiang Mai, Thailand

This study demonstrates that radio-pathomic maps of epithelium density derived from annotations performed by different pathologists distinguish high-grade prostate cancer from G3 and benign atrophy. In a test set of 5 patients epithelium density maps consistently demonstrate an AUC greater than 0.9 independent of which pathologist’s annotations trained the model or which pathologist’s annotations the model is applied to. The results in a larger test set largely mirror the results in the small test set. We also showed that radio-pathomic maps of epithelium density out-performed ADC maps independent of which observer was used to train the model.

This abstract and the presentation materials are available to 2020 meeting attendees and eLibrary customers only; a login is required.

Join Here