The Computational Pathology Group is developing software models based on artificial intelligence (AI) that can help pathologists diagnose tumours. The group was one of the first in the world to demonstrate that AI can match and even exceed the performance of human experts in specific diagnostic tasks. The team is currently working on the world’s first virtual pathologist, ANTONI.
AI turns computers into virtual pathologists
Our healthcare system is under enormous pressure. There is a looming shortage of healthcare providers while people, on average, live longer and need more care. For instance, there is a worldwide shortage of pathologists, medical specialists who examine tissue from patients to make a diagnosis. To reduce pressure on current specialists and guarantee the quality of care for the future, healthcare innovations are urgently needed.
The work of the Computational Pathology Group may provide the answer. Over the past decade, the team has developed several software models based on artificial intelligence (AI) that can help pathologists diagnose diseases such as cancer. By training algorithms with large digital datasets, consisting of images of often thousands of patients, the group has designed innovative AI models. These models can accurately analyse digitised images of diseased tissue and make diagnoses and even predictions based on them.
"Working with AI resulted in more consistent and accurate assessments"
The group was one of the first in the world to show that AI can match and even exceed the performance of human experts in diagnostic tasks, including assessing the severity of breast and prostate cancer. A follow-up study researched how this affected the daily work of a pathologist. Pathologists performed their tasks without and then with AI modelling. It was found that working together with AI resulted in more consistent and accurate assessments. Remarkably, even a trainee pathologist working in a country with little economic clout suddenly scored as well as a highly specialised pathologist in the Netherlands when aided by AI.
Meanwhile, the researchers are ready for their next ambitious step: they want to develop one of the world’s first virtual pathologists. This new AI model, which they call ANTONI, will be able to independently make clinical diagnoses for various tissue types including breast, prostate and colon cancer. What makes ANTONI special is that it will offer a form of ‘explainable AI’, meaning that medical specialists can ask about the reasoning behind the results to support diagnoses. This feature could boost clinicians’ confidence and open the door to broad applications of AI within healthcare.
"AI is not an end in itself, but a means to improve patient care"
This interdisciplinary research group brings together expertise in computer science, pathology, biomedical and software engineering, applied mathematics and computer vision. Together, they are a strong example of team science, with researchers with unique skills effectively working together to achieve something greater than the sum of its parts. Also highly admirable is their commitment to open science. Unlike many commercial AI developers, the group make all their models, codes and data freely available, allowing other scientists to build on their work.
The Computational Pathology Group focuses on both fundamental questions about artificial intelligence and its clinical application. All their work takes place within hospital walls and they involve a wide range of stakeholders, ranging from patients to clinicians, in their research. They do this based on the belief that AI is not an end in itself, but rather a means to improve patient care and make diagnostics as effective and efficient as possible. In short, this team’s research not only contributes to long-term scientific progress, but also directly benefits patients and pathologists worldwide.
Francesco Ciompi (1980) obtained an MSc in Electronic Engineering from the University of Pisa in 2006, and an MSc in Computer Vision and Artificial Intelligence in 2008. In 2012, he obtained a PhD cum laude from the University of Barcelona. Between 2013 and 2015, he was a postdoctoral researcher at the Computer Vision Centre (Spain) and at the Radboud University Medical Center’s Department of Radiology. Since 2022, he has been Associate Professor in Computational Pathology and Research Group Leader at the Department of Pathology of Radboudumc. His research is funded by several European research grants, including a grant from the Dutch Cancer Society (KWF) and a personal Vidi grant from NWO.
Jeroen van der Laak (1967) studied Computer Science at Radboud University in Nijmegen. Since 1991, he has worked at the Department of Pathology at Radboud University Medical Center (Raboudumc) in Nijmegen. In 2015, he started his own research group in the field of artificial intelligence in Pathology. He leads a number of research lines, including one on breast cancer diagnostics. His research is funded by grants from, among others, the Dutch Cancer Society (KWF), the Kidney Foundation, the European Union, and the foundation IT Projects Nijmegen. He is also a visiting professor at Linköping University (Sweden) and in 2021 he founded Aiosyn, a spin-off of Radboudumc where he is currently Chief Scientific Officer.
Geert Litjens (1985) studied Biomedical Engineering at Eindhoven University of Technology. He obtained his PhD in 2015 for his research on computer-aided detection of prostate cancer in MRI images at Radboudumc. After a postdoctoral period at the University of Heidelberg in Germany, he returned to Radboudumc in 2016. He is now a full professor, leading his research group as it focuses on developing multimodal AI solutions within healthcare, spearheaded by pathology and radiology. Litjens has acquired several research grants, including both a Veni and Vidi grant from NWO and an ERC Starting Grant.
