Artificial intelligence as a digital colleague improves prostate diagnosis

For the diagnosis of prostate cancer, the most common cancer among men, the microscopic analysis of suspected prostate tissue by the pathologist is crucial. But pathologists are subjective and often differ in their assessment. With potentially major consequences for a patient. Wouter Bulten, PhD student at Radboudumc, shows in his dissertation that artificial intelligence (AI) assesses prostate tissue just as well as experienced pathologists. He investigated this together with project leader Geert Litjens, on a dataset of more than a thousand patients from Radboudumc, partly thanks to support from NWO and KWF Alpe d’HuZes.


Based on this conclusion, Bulten and Litjens and their team, Karolinska Institute and Google Health, organized a major follow-up study, the PANDA challenge. The aim is to test whether computers can improve diagnostics worldwide. Radboudumc and Karolinska combined their databases to build the largest freely accessible dataset in pathology, with images of more than ten thousand prostate cancer biopsies. They then challenged AI experts worldwide to develop an algorithm for digital analysis in three months.

“We had more than a thousand participants,” says Bulten. “Everything was public for all participants, which allowed them to work together towards very powerful solutions. After ten days, one of the algorithms was already at the level of the average pathologist and after that it got even better. The creators of the best three algorithms received a cash prize. We then continued with the top fifteen for a more comprehensive and independent analysis. For example, we have developed a blueprint for large-scale evaluation of AI in pathology. AI is now technology-ready for clinical implementation.”


The PANDA study confirms the power of AI for diagnostics on a global scale. Will the computer soon replace the pathologist? Bulten: “I regularly hear this question in the media. But I do not think so. Our follow-up research shows that pathologists who use AI diagnose prostate cancer better and more consistently. AI alone does better than the average pathologist, but the pathologist together with AI is superior. The same applies here: two know more than one.”

Bulten sees an important role for AI in improving diagnostics by pathologists: “The quality improves when a pathologist uses AI as a digital second opinion, especially with less experienced doctors. Thanks to AI, we get better pathologists who agree more often. AI can also speed up diagnostics, making healthcare more affordable. In addition, you can use AI to very quickly assess many tissues in population screening. Then the pathologist only has to look at the suspicious biopsies.”


The question is how long it will take before AI is standard applied in clinical practice. “We now need to demonstrate that more accurate diagnostics with AI actually improves outcomes in prostate cancer. And that the costs of care are going down. To this end, we want to integrate AI into the existing workflow. Currently only ten percent of hospitals worldwide use digital images, the other ninety percent only view biopsies through a microscope. So we can make significant gains in that.”

According to the researchers, their algorithm can show not only whether a structure is tumor tissue, but also how aggressive a tumor is, the so-called tumor grading. Promoter Jeroen van der Laak: “Our software therefore provides a great deal of information. Aiosyn, a spin-off from Radboudumc, is now working further on software for various applications of AI in pathology in the clinic. This is how we create value for the patient.”

Artificial intelligence as a digital colleague improves prostate diagnosis
Source link Artificial intelligence as a digital colleague improves prostate diagnosis

Back to top button