New study explores how a crowdsourced tech competition led to innovations in prostate cancer assessment
A technology competition has resulted in the development of an artificial intelligence (AI) system for diagnosing and grading prostate cancer that has been shown to be equally as reliable as human pathologists, according to a new study in the journal Nature Medicine.
This research offers hospital leaders the opportunity to reduce the risks of human error when diagnosing prostate cancer by adding a supportive tool for physicians.
In 2020, Radboud University Medical Center in the Netherlands and Karolinska Institutet in Sweden organized an international competition called PANDA (Prostate cANcer graDe Assessment) to develop an AI system that could evaluate prostate cancer. The goal of organizing PANDA was not just to create an AI algorithm for prostate cancer diagnostics, but to develop a robust system that could be widely applied internationally.
The PANDA challenge attracted almost 1,300 developers and examined a database of over 10,000 prostate biopsies. The crowdsource-type approach to development was intended to reduce the problems and biases that can occur when a single group both develops and validates an AI system.
Helpful Results from PANDA Came Rapidly
PANDA quickly yielded useful results. “Only 10 days into the competition, algorithms matching average pathologists were developed,” said Kimmo Kartasalo, PhD, in a news release. “Organizing PANDA shows how competitions can accelerate rapid innovation for solving specific problems in healthcare with the help of AI.” Kartasalo is a Postdoctoral Researcher at the Department of Medical Epidemiology and Biostatistics at Karolinska Institutet and corresponding author of the Nature Medicine study.
“The results from PANDA show, for the first time, that AI systems can produce an equally accurate diagnosis and grading of prostate cancer in an international setting as human pathologists,” said Martin Eklund, PhD, Associate Professor at the Department of Medical Epidemiology and Biostatistics at Karolinska Institutet. “The next step is controlled studies for evaluating how to best introduce AI systems in patient care.”
While this AI system is not quite two years old, its reliable validation method indicates it could be integrated into clinical practice relatively quickly.
AI Shouldn’t Displace Pathologists, Researchers Note
One potential consideration that researchers recognize is the concern that the AI system could displace the role of pathologists.
“The idea is not for AI to replace human experts, but rather to function as a safety net to avoid pathologists missing cancer cases and to help in standardizing the assessments,” explained study co-author Lars Egevad, MD, PhD, Professor at the Department of Oncology-Pathology at Karolinska Institutet. “AI can also be an option in those parts of the world that today completely lack pathology expertise.”
The use of AI-based diagnostics in developing countries is particularly important, as many of those nations have less than one pathologist per 1 million people. AI software could support overstretched pathologists in these countries and also act as a second set of eyes in areas where pathologists are more readily available.
Future development on this system includes both determining how to introduce it into clinical environments and further enhancing the technology.
“Future studies should also include a larger variety of rare but unusually challenging biopsies as well as samples from countries with more varied ethnicity and demographics,” Eklund explained, describing the next areas of improvement that this technology could focus on.
While there are many potential applications and further developments that this new AI system could include, healthcare and oncology leaders recognize the immediate potential for improved accuracy when diagnosing prostate biopsies. By using AI systems to support human efforts, the likelihood of misdiagnoses can be reduced, improving the care that prostate cancer patients receive.