Meeting Banner
Abstract #3405

Prostate Cancer Localization Using Multi-Parametric MRI and a Maximum Likelihood Classification Algorithm

Sharon Clarke1, Bruce Daniel2, Jesse McKenney3, Manojkumar Saranathan2, Brian Andrew Hargreaves2, James Brooks4, Harachan Gill4, Mark Gonzalgo4, Benjamin Chung4, Emine U. Saritas5, Ajit Shankaranarayanan6, Graham Sommer2

1Diagnostic Radiology, Dalhousie University, Halifax, Nova Scotia, Canada; 2Diagnostic Radiology, Stanford University, Stanford, CA, United States; 3Anatomic Pathology, Cleveland Clinic, Cleveland, OH, United States; 4Urology, Stanford University, Stanford, CA, United States; 5Bioengineering, University of California Berkeley, Berkeley, CA, United States; 6GE Healthcare, Menlo Park, CA, United States

Segmentation of multi-parametric MR images of the prostate gland using a maximum likelihood classification algorithm correlate well with histopathology. These results show promise for identification of clinically relevant prostate cancer for either MR-guided biopsy or focal therapy.

Abstract PDF Presentation Video

Keywords

acceleration accuracy acquisition alongside among anatomic apparent approved arterial assessed automated axial basis benign bioengineering biopsy blinded board boulder boundaries brooks cancer cause channel classification classified classifier classify clinic clinically coefficient coil consent contrast correlates create death density diagnosed diagnostic diffusion disco drawn drew dual dynamic early either employing enhanced ethics examined expert focal frequently general genitourinary gill gland graham guided identification identify illustrates included incorporated informed input institution journal kappa keyhole leading likelihood manually maps mark matrix nova operative outlines overall parametric park pathologist pathology pixel pixels potential promise prospective prostate proven pseudo pulse quantitative radiologist radiology random reconstruction recorded recruited reduced relevant remaining representative represented representing represents resolution respectively respiratory review saturation saved scheduled score scores segmentation segmented semi separation sharing sixth slice software space spacing specimen subject subjects supervised system temporal therapy tissue train trained training triggered tumor tumors twenty urology utilizing variable view water whose worldwide zonal