During MRI-guided Transurethral Ultrasound Ablation (TULSA) of the prostate, the physician manually outlines the intended treatment zone on intra-procedural axial T2w images before starting ablation. This contouring task is repetitive, time-consuming, and subject to intra- and inter-reader variability. We present the validation of a deep learning model for artificial intelligence (AI)–assisted whole prostate segmentation compared to manual segmentation with two goals: 1) non-inferior accuracy and 2) time savings.
What's Your Reaction?






