Grant: Operating Grant - PA: Cancer Research (bridge funding)

Prostate cancer is a complex phenomenon on the molecular level. Our central research question is whether these complexities can be captured using imaging technologies. My proposal aims at developing methodologies for image-based prognostic evaluation of the prostate gland that are multiparametric, multimodal, and are linked to molecular and pathologic biomarkers of the disease, with the intention of the imaging eventually replacing or complementing these biomarkers. We propose the extraction and analysis of quantitative features from multiparametric MRI (T2, perfusion and diffusion) and registered multiparametric ultrasound (B-mode, Doppler, raw ultrasound RF signals, and elastography) along with texture parameters that quantify the statistical distribution and higher order relationships of the pixel intensities in all MRI and ultrasound modalities. This will create a large set of parameters that characterize a region of interest or lesion in the prostate tissue. These imaging features will form the radiologic profile of the tissue. After surgery, we will acquire the molecular profile with genomic and proteomic technics. Using this data, we will work towards the following goals: Imaging and registration - To build the technology and acquire data for registration of multiparametric ultrasound and multiparametric MRI and to create a comprehensive radiologic profile of prostate tissue. Fusion of imaging and molecular signature - To study the associations between the radiologic profile and molecular profile of prostate cancer in order to find feature subsets in the radiologic profile that can be used as surrogates for the molecular profile. Classification for diagnosis and prognosis - To develop a machine learning framework based on supervised learning in which the molecular profile and the radiologic profile inform one another to improve the prognostic value. This work could elevate "imaging markers" to reliable and valid prognostic biomarkers.