The study presents an Artificial Intelligence (AI) system designed to enhance the detection of clinically significant prostate cancer (csPCa) on b-mode transrectal ultrasound (TRUS) images, a procedure often challenged by imaging artifacts and benign confounders. The AI system learns ultrasound biomarkers by correlating TRUS with MRI and histopathology images during training, eliminating the need for MRI in future applications. Trained on a cohort of 289 men and validated with an additional 1573, the AI demonstrated patient-level sensitivity and specificity of 65% and 81%, respectively, and lesion-level sensitivity of 60%, outperforming an average human reader. The findings suggest that AI can effectively identify and localize csPCa in TRUS images, facilitating targeted biopsies without reliance on MRI.