Machine learning trumps conventional analysis in detecting lymphedema

Machine learning algorithms can now identify lymphedema—a chronic side effect of breast cancer treatment—with 94 percent accuracy, New York University researchers reported this month in mHealth.

Lymphedema isn’t a curable disease, study author Mei R. Fu, PhD, RN, and colleagues wrote. The swelling, heaviness, aches, burning and limited mobility make it an unattractive prospect. Early detection and intervention are the only ways physicians can reduce those symptoms and keep the condition from progressing.

“Clinicians often detect or diagnose lymphedema based on their observation of swelling,” Fu, an associate professor of nursing at the NYU Rory Meyers College of Nursing, said in a release. “However, by the time swelling can be observed or measured, lymphedema has typically occurred for some time, which may lead to poor clinical outcomes.”

It can occur after cancer surgery or as late as 20 years after, Fu and co-authors said. But within a decade of treatment, 41 percent of breast cancer patients experience it.

Fu’s team focused on machine learning since the technology excels in processing handfuls of data points that are independent from one another, just like lymphedema symptoms. Three hundred and fifty-five women who had undergone breast cancer treatment were included in the study, in which the researchers collected demographic and clinical information before asking patients whether they were experiencing any of 26 lymphedema systems.

The researchers then inputted symptom information into five different machine learning algorithms, including two Decision Tree models, a gradient boosting model, an artificial neural network and a support vector machine. They said all five modalities identified lymphedema more accurately than the standard statistical approach, but the artificial neural network was the most successful, with 93.8 percent accuracy.

“Using a well-trained classification algorithm to detect lymphedema based on real-time symptom reports is a highly promising tool that may improve lymphedema outcomes,” Fu said. “Such detection accuracy is significantly higher than that achievable by current and often-used clinical methods.”

Fu said the method also encourages self-monitoring of symptoms, since it has the ability to send alerts to patients at the highest risk for developing lymphedema.

“This has the potential to reduce healthcare costs and optimize the use of healthcare resources through early lymphedema detection and intervention, which could reduce the risk of lymphedema progressing to more severe stages,” Fu said.