Today, patients with high-risk stage II and stage III colorectal cancer (CRC) disease receive adjuvant treatment, whereas surgery is considered sufficient for stage I and low-risk stage II cancers. However, about 25% of patients with low-risk stage II cancers experience recurrence after surgery. Better methods are needed to identify those who could benefit from additional treatment. A histotyping method based on standard pathological samples and AI/deep learning automatically predicts the outcome of a patient and provides information about patients in need of additional treatment, potentially improving survival rates and quality of life.
With a worldwide implementation of digital pathology combined with the limitation in pathologist capacity and the ever increase in pathology samples, the automated outcome prediction provided by the method has potential to become a new state-of-the-art prognostic tool.
Inven2 seeks development partners and licensees for the technology.
Leading scientists at the Institute for Cancer Genetics and Informatics at Oslo University Hospital use deep learning to develop methods for automated analyses of pathology samples for improved cancer diagnostics. The methods analyze pathological images, such as tissue sections stained with haematoxylin and eosin (HE) scanned with high-resolution scanners. The prediction method of patient outcome in colorectal cancer is almost ready, and similar methods for other cancer types are under development. In a validation study, the results of the CRC method graded 88% of the patients correctly while 12% of the patients where given an uncertain or unresolved grading. For comparison, histologic grading from pathologists grouped more than 80% of the patients in the unresolved grading. Although scientifically validated, further testing of the CRC method in clinical settings is vital.
Atomization of manual routine of the pathologist´s review of a tissue sample with a far more precise prognosis.
Skrede et al., Lancet. 2020 Feb 1;395(10221):350-360