AI detects prostate cancer more frequently than radiologists. Additionally, AI issues half as many false alarms. This is evidenced by an international study coordinated by Radboudumc and published in The Lancet Oncology. This is the first major study where an international team transparently evaluates AI and compares it with radiologists’ assessments and clinical outcomes.
Radiologists face an increasing workload because men at higher risk of prostate cancer now routinely receive a prostate MRI. Diagnosing prostate cancer with MRI requires significant expertise, and there is a shortage of experienced radiologists. AI can help with these challenges.
AI expert Henkjan Huisman and radiologist Maarten de Rooij, project leaders of the PI-CAI study, organized a large competition between AI teams and radiologists with an international team. Along with other centers in the Netherlands and Norway, they provided more than 10,000 MRI scans. They transparently determined whether each patient had prostate cancer and had various groups worldwide develop AI for analyzing these images. The five best submissions were combined into a sort of super-algorithm for analyzing MRI scans for prostate cancer. Finally, AI’s assessment was compared with that of a group of radiologists on four hundred MRI scans.
Accurate Diagnosis
The PI-CAI community brought together more than two hundred AI teams and 62 radiologists from twenty countries. They compared the findings of the AI and the radiologists not only with each other but also with a gold standard by following up on the men whose scans were used. The men were followed for an average of five years.
From this first international study on AI in prostate diagnostics, it appears that AI detects almost seven percent more significant prostate cancers than the group of radiologists. Additionally, AI identifies fifty percent fewer suspicious spots that later turn out not to be cancer. This means that the number of biopsies can be halved with the use of AI. If these results are replicated in follow-up studies, it could greatly assist radiologists and patients in the future. It could reduce radiologists’ workload, lead to more accurate diagnoses, and result in fewer unnecessary prostate biopsies. The developed AI still needs to be validated and is not yet available to patients in the clinic.
Quality System
Huisman notes that society still has little confidence in AI. “This is because manufacturers sometimes build AI that is not good enough,” he explains. He is working on two things. The first is a public and transparent test to fairly evaluate AI. The second is a quality management system, similar to what exists for the aviation industry. “If planes nearly collide, a commission investigates how to improve the system so it doesn’t happen again. I want the same for AI. I want to develop a system that learns from every mistake, so AI is monitored and can continue to improve. This is how we build trust in AI for healthcare. Good, reliable AI can help make healthcare more efficient by relieving doctors of routine work.”
