{shortcode-339692dea37d870eb09ec0a1a1148467f2b68f91}
A team of scientists at Massachusetts General Hospital has developed a new tool that uses artificial intelligence to predict which patients are most likely to experience a recurrence of melanoma, the most deadly form of skin cancer.
The study — published on Oct. 31 in the npj Precision Oncology journal — was led by Harvard Medical School professor Yevgeniy R. “Eugene” Semenov, and Guihong Wan, an HMS research fellow.
To develop the tool, the researchers analyzed 1,720 early-stage melanomas from the Mass General Brigham healthcare system and the Dana-Farber Cancer Institute.
The research may lead to more effective treatment for patients in the early stages of melanoma, according to the study’s authors.
The “overwhelming majority of fatalities” stem from early-stage melanoma patients who do not have access to the “life-saving therapies” reserved for later-stage patients, Semenov said.
“Our, really, motivation was to try to understand, within this group of people who — quote-unquote — have early-stage disease, is there a way for us to tell which ones are going to get the more high risk for recurrence and progression from the ones who are not,” he said.
Wan explained that the tool could be used to identify who would benefit most from receiving immunotherapy at an earlier stage of their disease, allowing physicians to weigh the potential benefits of the treatment against its possible side effects.
“The tool possibly can be used to identify patients who may experience a recurrence,” she said. “Maybe they can receive the immunotherapy.”
Research on the recurrence of early-stage melanoma currently faces many limitations, including insufficient sample size, according to the paper’s authors.
“There have been other studies trying to do something similar to what we're doing, but they largely haven't focused on the earliest stage population. They looked at all the patients, including stage three patients, and they didn't really use large cohorts,” Semenov said. “So, for many reasons, there really hasn't been a comprehensive approach like the one we're leading.”
Utilizing data from “the entire healthcare system in Boston,” as well as “more reliable labeling,” the researchers were able to come up with more accurate models, Semenov said.
Beyond predicting early-stage melanoma recurrence, Semenov said there are many more potential applications of machine learning in healthcare.
“I think this is just the tip of the iceberg, and anyone who is in training now, I would heavily encourage really to start thinking about combining computational training with an interest in healthcare, because this sector is really going to explode and grow rapidly in the next 20, 30 years,” Semenov said.
