.svg)
AI Outperforms Radiologists in Detecting Prostate Cancer on MRI Scans. Research coordinated by Henkjan Huisman, Lab Director of the ICAI Healthy AI Lab
An international study led by Radboud University Medical Center has revealed that artificial intelligence (AI) can detect prostate cancer on MRI scans more effectively than radiologists, while significantly reducing false positives. Published in The Lancet Oncology, these findings represent a major advancement in medical diagnostics for prostate cancer.
As MRI usage for high-risk prostate cancer patients becomes routine, the demand on radiologists has increased, exacerbated by a shortage of experienced professionals. AI offers a promising solution by enhancing the diagnostic process. The PI-CAI study, coordinated by AI expert Henkjan Huisman , Lab Director of the ICAI Healthy AI Lab, and radiologist Maarten de Rooij, involved a large-scale competition between AI teams and radiologists, incorporating international contributions from centres in the Netherlands and Norway.
The study utilised over 10,000 MRI scans to develop and test AI algorithms. The top five AI models were combined into a super-algorithm and compared against radiologist assessments on 400 prostate MRI scans. This research is the first to transparently evaluate and compare AI performance against radiologists on such a large scale, involving over 200 AI teams and 62 radiologists from 20 countries. Accuracy was measured against a gold standard, with patient outcomes tracked over five years.
Results showed that AI detected nearly seven percent more significant prostate cancers than radiologists and reduced false positives by 50 percent, potentially halving unnecessary biopsies. These outcomes indicate that AI can significantly reduce radiologists' workload, improve diagnostic accuracy, and minimise unnecessary procedures.
Huisman emphasised the need for trust in AI within the medical community, highlighting the importance of transparency and quality management systems akin to those in the aviation industry. He is developing a public, transparent test for AI and a system to learn from errors, ensuring continuous improvement.
Although the AI developed in this study is not yet available for clinical use and requires further validation, researchers believe it holds great potential for enhancing prostate cancer diagnostics by reducing radiologists' workload and increasing diagnostic precision.
Read the full article from Inside Precision Medicine here.
Read the finding published in The Lancet Oncology here.
Newsletter
Stay in the loop
