Reviewed by Lexie CornerJun 13 2024
An international study led by Radboud University Medical Center compares and assesses AI with clinical outcomes and radiologist assessments. It finds that AI is more accurate than radiologists in detecting prostate cancer and reduces the frequency of false alarms by half. These findings are published in The Lancet Oncology.
Men at a higher risk of prostate cancer undergo routine prostate MRIs; thus, radiologists are dealing with an increasing amount of work. Significant experience is required to diagnose prostate cancer with MRI, and there is a shortage of qualified radiologists. AI can help with these difficulties.
The PI-CAI study’s project leaders, radiologist Maarten de Rooij and AI specialist Henkjan Huisman, arranged a significant competition between AI teams and radiologists with a global team. In collaboration with other facilities in Norway and the Netherlands, they performed over 10,000 MRI scans. They transparently ascertained whether prostate cancer was present in each patient. They permitted different teams worldwide to create AI to analyze these photos. A super-algorithm was created by combining the top five submissions to analyze MRI scans for prostate cancer. Eventually, four hundred prostate MRI scan evaluations from AI were contrasted with those from a group of radiologists.
Accurate Diagnosis
More than 200 AI teams and 62 radiologists from 20 different countries joined the PI-CAI community. They tracked the results of the men whose scans originated, comparing the results of AI and radiologists with one another and a gold standard. The men were monitored for an average of five years.
According to the results of this first international study on AI in prostate diagnostics, AI can identify approximately 7 % more significant prostate cancers than radiologists. AI also detects suspicious areas later determined not to be cancer 50 % less frequently. This suggests that using AI could cut the number of biopsies in half. In the future, radiologists and patients may benefit greatly if these findings are confirmed in additional research.
It might lessen radiologists' workload, offer more precise diagnoses, and reduce pointless prostate biopsies. However, the AI created has not yet been made available to patients in clinical settings and still requires validation.
Quality System
Huisman observes that society has little trust in AI. Huisman adds, “This is because manufacturers sometimes build AI that isn't good enough.”
He is working on two initiatives. The first involves developing a public and transparent test to evaluate AI fairly. The second initiative focuses on implementing a quality management system similar to those found in the aviation industry.
If planes almost collide, a safety committee will look at how to improve the system so that it doesn't happen in the future. I want the same for AI. I want to research and develop a system that learns from every mistake so that AI is monitored and can continue to improve. That way, we can build trust in AI for healthcare. Optimal, governed AI can help make healthcare better and more efficient.
Henkjan Huisman, AI Specialist, Radboud University Medical Center
The research was funded by the European Union under grant agreement number 952159 and by Health~Holland, Top Sector Life Sciences & Health.
Journal Reference:
Saha, A., et al. (2024) Artificial intelligence and radiologists in prostate cancer detection on MRI (PI-CAI): an international, paired, non-inferiority, confirmatory study. The Lancet Oncology. doi.org/10.1016/S1470-2045(24)00220-1