Artificial Intelligence and Biparametric MRI in Prostate Cancer Detection: From Benchmarking to Workflow Strategies
Prostate cancer is one of the most common cancers in men. MRI helps detect aggressive cancer accurately and on time, but interpretation is complex and requires a high level of expertise. At the same time, pressure on healthcare systems is increasing due to a growing number of patients.
This dissertation investigates two solutions: artificial intelligence (AI) and shorter MRI scans without the use of contrast agents. In an international study involving more than 10,000 MRI scans, 800 AI developers, and 62 radiologists, AI detected aggressive prostate cancer as well as radiologists, with fewer false alarms. It also showed that non-contrast MRI scans produced results comparable to standard scans.
Follow-up research focused on AI applications. AI support improved radiologists’ accuracy in detecting prostate cancer. Alternatively, when AI is used as an independent filter, nearly 50% of scans are diagnosed with very high reliability. This offers opportunities to significantly reduce radiologists’ workload.