Artificial Intelligence Method Can Help Predict Skin Cancer Recurrence

An artificial intelligence (AI)-based method developed by researchers from Massachusetts General Hospital (MGH) can help identify early-stage melanoma patients at risk for cancer recurrence and assist physicians in deciding the level of care a patient may need.

According to the Centers for Disease Control and Prevention (CDC), there were 88,059 new cases of melanoma, a deadly type of skin cancer, in the US in 2019.

A press release also noted that deaths related to melanoma most commonly occur among patients who received a diagnosis of early-stage melanoma and later experienced a recurrence that is generally not discovered until it has spread. Thus, identifying patients at high risk of recurrence is key to boosting survival rates.

This issue led a group of researchers from MGH to create an AI-based method to draw attention to the patients most likely to experience a recurrence, informing clinicians of the patients that need more aggressive treatment.

Generally, patients with early-stage melanoma undergo surgery to eliminate cancerous cells, according to the press release. However, those battling advanced-stage cancer may need immune checkpoint inhibitors, which can strengthen the immune response to tumor cells.

“There is an urgent need to develop predictive tools to assist in the selection of high-risk patients for whom the benefits of immune checkpoint inhibitors would justify the high rate of morbid and potentially fatal immunologic adverse events observed with this therapeutic class,” said senior study author Yevgeniy R. Semenov, MD, an investigator in the Department of Dermatology at MGH, in the press release.

To create the AI-based method, the group of researchers collected 1,720 early-stage melanomas, 1,172 of which came from the Mass General Brigham healthcare system and 548 from the Dana-Farber Cancer Institute. They identified 36 clinical and pathologic features of these melanomas from patient EHRs, of which tumor thickness and rate of cancer cell division were found to be the most predictive.

Using these features, the researchers trained machine-learning algorithms to predict patients’ recurrence risk, validating them on various data sets from Mass General Brigham and Dana-Farber Cancer Institute.

“Our comprehensive risk prediction platform using novel machine learning approaches to determine the risk of early-stage melanoma recurrence reached high levels of classification and time to event prediction accuracy,” said Semenov. “Our results suggest that machine learning algorithms can extract predictive signals from clinicopathologic features for early-stage melanoma recurrence prediction, which will enable the identification of patients who may benefit from adjuvant immunotherapy.”

The use of AI to support cancer recurrence prediction is becoming an increasingly common practice.

In August, Mayo Clinic led a study that described how an AI model incorporating a deep-learning framework could help improve predictions of recurrence and survival in colorectal cancer (CRC) patients.

After gathering thousands of digital slide images of CRC tumors, researchers developed an algorithm to identify different regions of interest within the tumors.

AI has also been incorporated into the predictive efforts surrounding chronic care.

In April, researchers from the National Institute of Health Clinical Center created an AI model that assessed pancreas health and fat levels to determine patient risk for type 2 diabetes. The model was developed using non-contrast abdominal computed tomography images from various datasets.