Trial lecture
“Foundation models in cancer research”
Time and place: April 26, 2024 11:15 AM, Kristen Nygaards sal (5370), Ole-Johan Dahls hus
Main research findings
We developed a method to automatically assess the prognosis of a patient surgically treated for colorectal cancer. The method analyses microscope images of the resected tumour, and consists of two parts: one that finds cancerous regions in an image and one that scores them for severity. Tissue sections from about 2500 patients were used to develop the method using deep learning. The result was integrated with other risk factors to increase the ability to discriminate between good and poor prognosis. Later, the method that finds cancerous regions was made more robust, and can now handle images from different cancer types.
The method was thoroughly tested in an independent cohort with over 1000 patients from different countries. It grouped patients into good, intermediate, or poor prognosis groups with clear differences in survival rates. Using current guidelines results in prognosis groups with similar survival rates, but our good prognosis group was three times larger than the guideline-based good prognosis group. This means that our method identified many low-risk patients who are now perhaps recommended unnecessary adjuvant therapy.
We founded the company DoMore Diagnostics to get our method into the clinic so that patients and physicians can use it to inform further treatment decisions.
Adjudication committee
- Professor Paul J van Diest, Department of Pathology, University Medical Center Utrecht, the Netherlands
- Professor emeritus Arvid Lundervold, Department of Biomedicine, University of Bergen, Norway
- Professor Anne Solberg, Department of Informatics, University of Oslo, Norway
Supervisors
-
Professor Emeritus Fritz Albregtsen, Department of Informatics, UiO, Norway
-
Professor Håvard E. Danielsen, Institute for Cancer Genetics and Informatics, Oslo University Hospital, Norway
Chair of defence:
Contact information at Department: Mozhdeh Sheibani Harat