Prostate cancer is the most common cancer in men. To diagnose it, pathologists perform a microscopic analysis of suspicious prostate tissue. But pathologists are people, and people are subjective and not always consistent in their assessment. This can have serious consequences for the patient.
The researchers at Radboudumc wanted to find out if AI will make fewer mistakes. together with Project leader Geert Letgens, PhD candidate Wouter Bolten, has applied artificial intelligence to prostate tissue of more than a thousand patients. He found that they classified them at least as well as experienced pathologists.
Based on this conclusion, they organized a large follow-up study with the Karolinska Institutet and Google Health. With the PANDA Challenge Do they want to check Whether computers can improve prostate cancer diagnosis worldwide. Radboudumc and the Karolinska Institute have combined their databases to build the largest and freely accessible data set in pathology, with images of more than ten thousand prostate cancer biopsies.
They then challenged AI experts from around the world to develop an algorithm for their digital analysis in three months. More than a thousand participants registered in International Programming Competition. “Everything was public for all participants, which allowed them to develop very powerful solutions together,” says Bolten.
second opinion
Ten days later, one of the algorithms was already at the level of an ordinary pathologist and after that it got better. With the top fifteen participants, the team continued the analysis. They have prepared AI technology for use in a clinical setting. The algorithm can show not only whether the structure is neoplastic tissue, but also how aggressive the tumor is.
Does this mean that the computer will soon replace the coroner? “I don’t think so,” Bolton says. The PANDA Challenge showed that AI itself performs better than the average pathologist, but also that pathologists using AI do a better job. With two you also know here more than alone.
Bulten sees more advantages. Quality improves when a pathologist uses AI as a digital second opinion, especially when it comes to less experienced clinicians. AI can speed up diagnosis, making healthcare more affordable. And you can use AI in population screening to assess many tissues very quickly. Then the pathologist only has to look at the suspicious biopsies.
