A study of the evidence for the use of AI solutions in diagnosing skin cancer has boosted the development of an algorithm from the University of Gothenburg in order to determine the severity of melanoma as precisely as dermatologists.
Scientists at the University of Gothenburg have shown that the technology can perform at the same level as dermatologists in assessing the severity of skin cancer. The system was developed to help doctors determine the stage of skin cancer, and the study was published in the Journal of the American Academy of Dermatology.
And while patients often independently find melanomas by discovering a new mole or changing an existing one, even dermatologists can struggle to decide whether or not it is harmful.
The researchers doubted that AI could help them with the task, and they classified melanomas using a convolutional neural network, an effective image analysis method that has proven adept at identifying various skin lesions.
The study was conducted at Sahlgrenska University Hospital in Gothenburg, where researchers trained and validated the system with 937 images of melanomas collected through a melanoma dermatoscope, a portable instrument used to examine the skin.
They tested the algorithm’s evaluations on 200 diagnosed cases from a dermatologist, and when they compared the system’s performance with an analysis of seven independent dermatologists, the result was a draw.
“The results of the study are interesting,” said Sam Polesie, a researcher at the University of Gothenburg and a medical specialist at University Hospital Sahlgrenska, and author of the study.
“The hope is that the algorithm can be used to support clinical decision in the future, as no dermatologist has significantly outperformed the algorithm,” he added.
The researchers acknowledge that the algorithm still needs to be further refined and evaluated in the long term in a clinical setting and future studies that monitor patients over time.
However, the study shows that AI can help assess the severity of melanoma before surgery, which affects how comprehensive the process should be.