AI models to analyze MRI scans of tumors
Prostate cancer screening faces a critical challenge due to the shortage of expert reviewers. Jonas Wallstr枚m has secured research funding to develop AI models that could offer a practical solution.
The funding, administered by the Sahlgrenska Comprehensive Cancer Center鈥檚 research council, aims to strengthen the competitiveness of cancer research at Sahlgrenska University Hospital and Sahlgrenska Academy within the strategic research plan framework.
Jonas Wallstr枚m, a senior physician specializing in abdominal radiology, conducts radiology and urology research at the Institute of Clinical 91探花s. He has been awarded research funding to focus on AI model development linked to tumor progression.
How can AI contribute to cancer care?
鈥滻n this research project, we aim to utilize AI models to track tumor changes over time in men undergoing repeated MRI scans. Initially, we plan to use the model to filter out investigations without significant findings. Subsequently, we intend to employ the model to predict image changes that could indicate the development of serious cancer later on.鈥
Reducing overdiagnosis
How might this method be implemented in clinical practice?
鈥滻鈥檝e been involved in the Gothenburg 2 Screening Study for early prostate cancer detection since its start in 2015, both as a researcher and a radiologist reading MRI scans. Our findings showed that a screening algorithm combining PSA tests and magnetic resonance imaging reduces overdiagnosis of low-risk cancer by half while maintaining effective detection of clinically significant cancer,鈥 explains Jonas Wallstr枚m.
鈥滿anaging the vast volume of images in a future screening program with a limited number of expert reviewers poses a significant challenge. We believe AI models can play a role in addressing this challenge.鈥
鈥漊nique database鈥
How will the research be conducted practically?
鈥漌e have an extensive database of MRI scans for training algorithms, which will undergo evaluation within the ongoing screening study.鈥
Are there any similar examples?
鈥漀o, not precisely. Our screening examination database is unique, and as of now, I don鈥檛 know of any AI group focusing on follow-up studies,鈥 says Jonas Wallstr枚m.
What advantages do you expect for patients in this approach?
鈥滲y supporting and aiding radiologists, high diagnostic quality can be offered to more patients, promoting equitable care. While engaged in other AI-related research projects, I鈥檓 not currently utilizing algorithms in my clinical practice.鈥
View a complete list of the 24 projects that have received research funding distributed through Sahlgrenska Comprehensive Cancer Center's Research Council in 2023.
This article was originally published on