Prostate Cancer Radio-Pathomics for Differentiating Clinically Significant Disease

Abstract Prostate cancer is the most commonly diagnosed non-cutaneous cancer, affecting one in seven men. Even when treated with a radical prostatectomy, historically about 20% of patients exhibit tumor recurrence. This proposal will focus on the integration of two separate, complimentary datasets to better differentiate high risk patients: multi-parametric magnetic resonance imaging (MP-MRI) and whole-mount post-surgical prostate pathology samples. We will develop radio-pathomic algorithms capable of predicting underlying pathomic features from non-invasive imaging in order to differentiate prostate cancer with high metastatic potential. Our overarching hypothesis is that microscopic, heterogeneous pathomic features of prostate cancer are reliably detectable and quantifiable with macroscopic quantitative MP-MRI. Non-invasively mapping these features will provide a clinically useful tool for differentiating aggressive from indolent prostate cancer, and for potentially targeting with radiation. This proposal includes two specific aims in response to the goals outlined in PAR-19-264. Specific to the funding opportunity announcement: Aim 1 will develop radio-pathomic approaches for defining imaging- based biomarkers capable of distinguishing aggressive from indolent prostate cancer. This will be done at the microscopic level in Aim 1.1 with histology, and then at the macroscopic level in Aim 1.2 with MP-MRI. Aim 1.3 will test the resilience of the radio-pathomic algorithm by intentionally perturbing the system and algorithms. Combining the Rad-Path datasets with clinical variables in Aim 1.4 will look to improve sensitivity and specificity for early detection and differential diagnosis, by correlating our radio-pathomic maps with other omics. Additionally, included in Aim 1, are extensive validation experiments meant to further establish the robustness of the radio-pathomic algorithm. In Aim 2, this project will translate the radio-pathomic algorithms to the clinic. This will include in Aim 2.1 adapting our algorithms to two clinical MR imaging systems (GE and Siemens), and in Aim 2.2 developing a radio-pathomic driven MRI protocol for serial imaging on a combined MR-LINAC system, one of only two operational in the US. Completion of this project will provide a powerful set of quantitative imaging tools to clinicians for improved differentiation of high-risk prostate cancer and for measuring response to prostate cancer therapy.