Pitt Researchers Create Program To Diagnose Skin Lesions
Skin lesions are a heath concern that many will face in their lifetimes, with 76,000 Americans being diagnosed every year with skin cancer. New research, blending technology and medicine, hopes to make the detection process easier and more accurate.
In collaboration with Carnegie Mellon University, scientists at Pitt and UPMC have created a computer program that can scan photos of skin lesions and assess whether or not they will require further treatment.
To test it, they compiled images of moles that were already accurately assessed by physicians to see how the program would rate the moles.
“We trained a computer, or a classifier, to be able to look at certain features within each image and come up with an algorithm that would predict which ones were benign and which ones were malignant,” said Laura Ferris, associate professor of dermatology at the University of Pittsburgh.
Researchers then ran a new set of photos through the program, asking it to rate lesions either benign or malignant using the developed algorithm.
“The computer was able to accurately identify 38 of 39 melanomas as being malignant,” Ferris said.
When tested against the human eye, the program has continued to perform accurately. A board-certified panel of dermatologists examined the same photos as the programs and were asked to assess the status of the lesion. The program was 96% accurate in picking melanomas whereas the doctors only scored 70.8%.
Researchers will now continue to test the software to see if the same results are seen using lesions submitted by primary care physicians, not dermatologists. The goal is to make the technology available to primary care doctors whose specialty is not skin cancer detection.
“When the patient’s in the office and they say, ‘by the way, I have this mole, what do you think?’ this could be a tool that would be helpful. It would give them rapid feedback and if it comes up with a concerning score then the primary care doctor could alert the dermatologist,” Ferris said.
This hybrid of health care and computer technology is thanks to the collaboration between UPMC and the two universities, says Ferris.
“Because we’re here in Pittsburgh we have specialists in medicine and also specialists in computer science. This is an example of how having these two groups being able to work together, we can approach a problem that either one of us couldn’t have done on our own.”