Meeting Banner
Abstract #1048

Unsupervised multi-characterstic framework for DW-MRI prostate cancer localization

Raisa Z Freidlin 1 , Harsh K Agarwal 2 , Sandeep Sankineni 3 , Anna M Brown 3 , Marcelino Bernardo 3,4 , Peter A Pinto 3 , Bradford J Wood 3 , Deborah E Citrin 3 , Peter L Choyke 3 , and Baris Turkbey 3

1 NIH/CIT, Bethesda, Maryland, United States, 2 Philips Research, New York, United States, 3 NIH/NCI, Maryland, United States, 4 Leidos, Maryland, United States

Existing studies using diffusion MRI models for prostate cancer (PCa) detection have not used an unsupervised approach. Our proof-of-concept study introduces a novel unsupervised multi-characteristic framework for localizing PCa. Our framework calculates voxel-based parameters from the IVIM and kurtosis models and identifies tumor and tumor suspicious voxels using patient-specific thresholds. Ten patients with moderate-high clinicopathological risk for PCa underwent 3T prostate MRI and subsequent biopsy. The index lesion was identified in all patients (100% patient-based detection rate). Of the 25 framework-identified lesions, 14 were true positives (56% lesion-based detection rate). This novel framework shows promise for identifying index PCa lesions.

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

Join Here