MACDILL AIR FORCE BASE, Fla. -- The dermatology clinic at MacDill Air Force Base, Florida boasts a machine that can help patients log and track various skin conditions over time.
It’s ability to use high resolution photos of the patients’ body and intelligently detect when new marks appear and grow larger allows Maj. Thomas Beachkofsky, the 6th Healthcare Operations Squadron dermatologist, easily monitor areas of concern with his patients.
However, a new software upgrade that takes advantage of machine learning has opened up new opportunities to use this machine, which is one of two in the Air Force.
“Our new software that works with our body scanner uses artificial intelligence and machine learning to analyze a lesion or mark on the skin and uses an algorithm to rate the likelihood that the spot is harmful,” said Beachkofsky. “With training, our dermatology technicians can use this program to efficiently scan and process questionable spots.”
Beachkofsky explained that although the machine makes an educated guess on the severity of the lesion, it is up to a fully-trained dermatologist to make a diagnosis, and recommend treatment.
So far, Lt. Col. Kurtis Kobes, the 6th Operational Medical Readiness Squadron dental flight commander was among the first to benefit from the new software, after seeking a second glance at MacDill’s clinic for a spot on his forearm.
“Based on how it looked, and the results from the scan, I ordered a biopsy which came back as melanoma in situ,” remarked Beachkofsky.
Melanoma in situ, also called stage zero melanoma is a very early stage of cancer where the cancerous cells only affect the epidermis and haven’t spread to deeper layers of the skin.
“It’s very fortunate that something like this was caught in as early of a stage as it did,” remarked Beachkofsky. “Melanoma can be deadly if left to spread, so treating it while it’s in situ allows a simple procedure with a fast recovery.”
With the new software upgrade, the dermatology office hopes to give its patients the peace of mind that their questionable spots can be checked accurately and efficiently.
“I’m very grateful for the dermatology clinic quickly verifying and handling the suspicious area on my forearm,” said Kobes. “I’ve had this spot for over a year, and after having it looked at by other clinics, I was only told it could be monitored, but it didn’t look alarming.”
In a study named “Man against machine,” the deep-learning algorithm used by the analyzing software was able to correctly identify 95 percent of malignant skin tumors. This data was compared to 58 dermatologists across 17 nations, who were able to successfully identify 86.6 percent of the same tumors.
“It’s definitely not a replacement for doctors, nor is AI taking over healthcare,” laughed Beachkofsky. “It’s mostly a tool for a dermatologist to get a second opinion from a system that has analyzed tens of thousands of lesions and is constantly learning.”