Analysis of breast and prostate cancer with artificial intelligence for improved care

DeepGrading: Improved breast and prostate cancer grading for accurate diagnosis

Providing the right care for individual prostate cancer patients is hampered by variability in assigning the histopathological Gleason grade. It is shown that it is possible to leverage artificial intelligence to help pathologists reduce this variability and increase the quality of their diagnostics. Within this project, together with Aiosyn, a start-up focusing on bringing AI algorithms to pathologists, the next step will be taken to integrate these algorithms in an efficient way into the workflow of the pathologist.

Breast and prostate cancer are the most common cancers among women and men in the Netherlands, with over 15,000 and 12,000 new cases per year, respectively. Many of these patients have low-grade cancers, which do not require aggressive treatment. However, due to variability in grading patients sometimes are either treated with aggressive therapy unnecessarily, reducing their quality of life, or not treated aggressively enough, causing their cancer to progress. By supporting pathologist with AI tools, better, more personalised grades can be obtained for individual patients and ensure they receive appropriate treatment.

Within the research AI algorithms have been developed that can grade cancers at the level of expert pathologists. However, it is unclear how pathologists can use these algorithms in practice effectively. Within this project it will be investigated which algorithm-supported workflows results in the most efficient and accurate diagnostic pathway. The AI-supported pathologists will be able to provide better, more consistent grades for breast and prostate cancer, thus contributing to the increase of effective and appropriate treatment for patients.

At the end of the project the goal is to have fully, clinically validated AI algorithms for supporting pathologists in accurately grading breast and prostate cancer. In addition, the algorithms will be able to handle unseen, rare subtypes of cancer and flag these for the pathologist, preventing oversight.

The cancer grade, provided by pathologists, is the most important predictor of patient outcome, but suffers from inter- and intra-pathologist variability, reducing its usefulness for individual patients. An expert-level AI system will be provided to support pathologist and help reduce this variability and make their diagnostic practice more accurate and efficient.
Technology Readiness Level (TRL)
4 - 7
Time period
48 months
Logo Radboud UMC