Feature extraction in multiparametric ultrasound for the classification of prostate cancer

Prostate cancer is the most diagnosed cancer in men. The current recommended diagnostic procedures involve systematic prostate biopsy and multiparametric MRI guided biopsy. However, systematic prostate biopsy can lead to overdiagnosis of clinically insignificant cancers, and MRI suffers from high cost. Alternative imaging methods have been explored to accurately detect prostate cancer. Multiparametric ultrasound has shown the ability to detect suspicious regions within the prostate, with comparable accuracy to multiparametric MRI. In collaboration with the Western General Hospital in Edinburgh, we have collected both conventional and super resolution ultrasound imaging data from patients with confirmed prostate cancer. In this seminar, we will describe a new multiparametric ultrasound approach based on super resolution imaging, augmented by B-mode and contrast enhanced imaging. Using this approach, structural and dynamical features are extracted from the data and are used to detect prostate cancer using machine learning classifiers.