When a patient is diagnosed with breast cancer, a pathologist uses a system called grading to help determine the treatment course based on how abnormal the tissue appears. Typically, this looks at just the cancerous cells in the patient’s body. A new method using artificial intelligence also examines the patient’s non-cancerous and immune cells, and it may be effective enough to determine if a patient can cut back on chemotherapy that isn’t needed.
According to research recently published in the journal Nature Medicine, an AI system that analyzes 26 different breast tissue properties to come up with an overall prognostic score performed better than expert pathologists at determining how the disease will play out. This multi-faceted analysis bases the prognosis not just on cancer cells, but also on immune system cells and non-cancerous cells that provide structure for tissue.
Lee Cooper, study co-author and associate professor of pathology at Northwestern University Feinberg School of Medicine, says, “Our study demonstrates the importance of non-cancer components in determining a patient’s outcome. The importance of these elements was known from biological studies, but this knowledge has not been effectively translated to clinical use.”
The AI model the team built and tested measures the appearances of these non-cancerous cells, as well as cancerous cells and how the two interact. Cooper says it can be difficult for a pathologist to evaluate these relationships, and the model presents the information in such a way that the pathologist can understand how it came to its conclusions.
The researchers hope that this model can provide patients with a better idea of how high-risk their disease is, which may help them better decide how to proceed with treatment. They also say that changes in the tissue over time could determine if treatment should be cut back or escalated, particularly when it comes to chemo.
It could address access to care issues, as well.
Cooper explains, “We also hope that this model could reduce disparities for patients who are diagnosed in community settings. These patients may not have access to a pathologist who specializes in breast cancer, and our AI model could help a generalist pathologist when evaluating breast cancers.”
The next step is evaluating the model to see if it’s appropriate for clinical use.
You can read more about the study, which was conducted in partnership with the American Cancer Society, here.