This study examined whether artificial intelligence could help doctors read MRI images for prostate conditions. The team used computer models to analyze scans from 153 patients who had been diagnosed with either prostate cancer or prostatitis. They compared the computer results against the readings made by experienced human radiologists.
The AI models achieved high accuracy rates, reaching up to 99% for identifying prostatitis and 96% for specific cancer lesions in certain scan types. In other areas, the computer programs performed similarly to expert doctors, showing substantial agreement between the two methods. However, the models struggled significantly to detect smaller lesions, missing them completely in some cases.
Because this was a retrospective study looking at past data, there was no randomization or control over how patients were selected. The researchers note that while these tools show potential as decision support, they are not yet proven in forward-looking trials. Readers should view these findings as promising but incomplete until larger, prospective studies confirm the results.