More global coordination needed for AI-enhanced diagnostics, treatment to save lives of prostate cancer patients

The largest prostate cancer biopsy dataset — containing more than 95,000 images — was created by researchers in Sweden to ensure AI can be trained to diagnose and assess prostate cancer for real-world clinical applications.

The researchers will today, at the annual conference of the European Association of Urology (EAU22), call for large-scale clinical trials of artificial intelligence (AI) algorithms and greater global coordination to ensure that AI-enhanced diagnostics, prognosis and treatment selection can help. Save lives.

There is a shortage of pathologists around the world, both generalists and specialists in urology. AI can help detect early-stage prostate cancer, but due to the vast differences in how clinics prepare samples, scan images, and in the diverse patient populations they serve, many algorithms are not universally applicable.

The team, from the Karolinska Institutet, teamed up with colleagues from the Radboud University Medical Center in the Netherlands, the University of Turku in Finland and Google Health in the US to conduct an AI competition involving nearly 1,300 developers from across the globe. whole world. The developers have created algorithms that can assess prostate cancer tumors and trained them using 10,000 international biopsy images. The best performing algorithms outperformed generalist pathologists and matched the average performance of specialist uropathologists.

dr. Kimmo Kartasalo, who will present the results of the competition at EAU22, said: “Assessment of prostate cancer is an important step in deciding on the right treatment, but it is quite a subjective process and the differences between assessments by pathologists can sometimes be large. AI can provide additional expert advice, compensating for the pathologist shortage and standardizing assessment. While many algorithms are not generally applicable, the algorithms developed in our competition maintained their performance across different patient cohorts.”

PhD candidate Nita Mulliqi collaborated with colleagues from the Karolinska Institutet to prepare the comprehensive dataset of 95,000 prostate biopsy images, the equivalent of more than three years of work from a single uropathologist. They used biopsies from a clinical trial in Stockholm that lasted about four years as of 2012, and obtained images from nine other European labs, and many rare disease subtypes from colleagues in Australia.

Mulliqi now uses the dataset to train and test a clinically applicable robust AI based on integrating the best elements of the competition’s best performing participants into a single enhanced algorithm. The comprehensive dataset will ensure that the algorithm can deal with the kind of additional complexity that can be found in a real clinical situation, such as rare cancer types and situations that mimic cancer, but are benign.

Through the research, Mulliqi identified four key areas that require specific attention to ensure that better assessment and prognosis of prostate and other cancers can be achieved using AI, and that the algorithms can be responsibly introduced into clinical use. .

The four areas are:

  1. Scanner calibration: making sure the setup is the same everywhere scanning takes place
  2. Improved Algorithms: Using state-of-the-art AI methodology to ensure robust performance and wide applicability of the algorithms
  3. Dataset upscaling: delivering larger international datasets to ‘learn’ the AI
  4. Modeling morphological heterogeneity: looking at different subtypes of the same disease

Mulliqi presents these findings today at EAU22. She said: “AI holds promise and can benefit patients everywhere, but to deliver on this promise we need an international effort to collect data sets that are representative of the variation in technical approaches and between patients. The combination of our huge database and our colleagues’ algorithms are starting to show how we can really work together to make a big difference for clinicians and patients.”

Professor Jochen Walz heads the Department of Urology at the Institut Paoli-Calmettes Cancer Center in Marseille, France, and is a member of the Scientific Congress Office of the EAU. He said: “AI is becoming a routine tool, which will not replace pathologists and urologists, but will help them make more consistent decisions. There is currently a lot of variation in the assessment of prostate cancer, especially outside specialized centers.

“This research has used a clever way – crowdsourcing expertise – to develop AI to improve tumor classification and has taken the next step by validating it against a very diverse range of images. This shows that it can be used in the general practice.

“Until now, AI has only replicated the rating system used by urologists. But it has the potential to go further – to identify elements in the images that can directly predict clinical outcomes. That’s the next challenge for AI.”

Leave a Comment

Your email address will not be published.