Convolutional neural networks and misdiagnosis of melanoma

Posted in Firm News, Medical Malpractice on November 18, 2020

Patients in Washington and around the country want certainty when it comes to their cancer diagnoses. The idea that someone can misidentify a lesion is terribly scary for any patient. There can be serious health consequences when a doctor error happens.

Understanding melanoma

Melanoma is the most serious form of skin cancer. It can even be fatal if it’s not treated. When someone has a suspicious lesion that could be cancerous, it’s very important that they get it looked at. It’s also crucial to get an accurate diagnosis. Dermatologists have all kinds of protocols to ensure this happens. They look for changes to shape, size and color. They’ll check suspicious spots every six months.

Recently, artificial intelligence has been becoming more important in medicine. One example of this is the advancement of convolutional neural networks. This technology has been touted as similar in accuracy to observation and diagnosis by a dermatologist. However, a recent letter from researchers, printed in the Journal of Investigative Dermatology, raises doubts about this.

These researchers found serious errors with the way the technology works. Convolutional neural networks seem to have real difficulty distinguishing between benign melanocytic nevi and melanomas. Some of these misses are very serious. The researchers compared an example to mistaking a panda bear for a primate like a gibbon. For now, it’s important to do more work on these systems before they become a go-to diagnostic tool.

Anyone who’s suffered a cancer misdiagnosis may want to consider contacting a lawyer. An experienced attorney may be able to help determine if there’s a valid medical malpractice claim. An attorney may advise their client on how to proceed and what damages the patient might be eligible for.