UCLA

Prostate cancer (PCa) is one of the most common cancer-related diseases among men in the United States. Multi-parametric magnetic resonance imaging (mp-MRI) is considered the best non-invasive imaging modality for diagnosing PCa. The core components of mp-MRI include T2-weighted imaging (T2w), diffusion-weighted imaging (DWI), and dynamic contrast-enhanced imaging (DCE), each of which provides distinct anatomical or functional information. However, mp-MRI for PCa diagnosis is currently limited by the qualitative or semi-quantitative interpretation criteria, leading to inter-reader variability and a suboptimal ability to assess lesion aggressiveness. Deep learning is a class of methods designed to automatically learn multi-layer artificial neural networks from the training data for various tasks, including image classification, object detection, and segmentation. Here, deep learning methods specific to multi-parametric imaging were proposed to detect, segment PCa lesion and assess the lesion aggressiveness. In addition, an alternative learning method using unannotated dataset was designed, due to the inaccessibility of accurate annotated dataset in many institutions.