Procia has released the results of a new artificial intelligence (AI) study that predicts diagnostic agreement for melanoma, a deadly form of skin cancer.
The findings, presented at the European Conference on Computer Vision (ECCV) 2022, highlight the technology’s potential to improve diagnostic accuracy for melanoma and other diseases with little pathologist coordination.
A Proscia study conducted at the University of Florida and Thomas Jefferson University “Using Whole Slide Image Representations from Self-Supervised Contrastive Learning for Melanoma Concordance Regression” demonstrated AI performance on whole slide images of 1,412 skin biopsies.
Each image was evaluated by three to five dermatologists to establish a concordance rate.
In addition to this study, Procia plans to conduct additional research detailing the potential benefits of AI in helping pathologists diagnose melanoma.
One such benefit is reducing the misdiagnosis rate for difficult cases. Melanoma often looks like benign mimics, causing pathologists to disagree on the diagnosis 40% of the time. Since cases are often evaluated by only one pathologist, AI that evaluates coordination with multiple experts can help improve diagnostic accuracy by acting as a second set of eyes.
Also, turnaround time for critical results can be accelerated. More than 15 million skin biopsies are taken annually in the US, each of which can reveal hundreds of diagnoses. AI that predicts diagnostic agreement can flag challenging cases by suggesting additional tests to provide a more complete look before pathologist review, driving efficiency gains.
Another benefit is cost reduction. Frequent over-diagnosis of melanoma not only leads to additional costs for health systems, but also leads to patients undertaking unnecessary treatment and dealing with the stress of believing they have a malignant disease. Increased diagnostic accuracy can help eliminate these burdens.
“Diagnosing melanoma can be very challenging,” said Kiran Motaparthy, MD, director of dermatopathology and clinical associate professor of dermatology at the University of Florida.
“Procia’s technology represents an exciting breakthrough as pathologists increasingly turn to AI to deliver on their commitment to excellent patient care.”
Procia’s research suggests that the same AI could be extended to other diagnoses that exhibit low pathologist agreement. This includes staging of breast cancer and Gleason grading of prostate cancer, which is used to assess the aggressiveness of the disease. Both often play an important role in informing treatment decisions.
“With this study, we’ve laid the groundwork for a new use case for AI in pathology that can have a tremendous impact on patient outcomes,” said Sean Grullan, Procia’s chief AI scientist and lead author of the study.
“Our technology relies on self-supervised learning to detect incredibly subtle patterns, demonstrating the power of one of the most advanced approaches in AI.”