Proscia has released study results on new artificial intelligence (AI) that predicts diagnostic agreement for melanoma, the deadliest form of skin cancer.
The research results, which were presented at the European Conference on Computer Vision (ECCV) 2022, emphasize the technology’s potential to increase diagnostic precision for illnesses with poor pathologist concordance, such as melanoma.
Proscia’s work, “Using Whole Slide Image Representations from Self-Supervised Contrastive Learning for Melanoma Concordance Regression,” was carried out at the University of Florida and Thomas Jefferson University and showed how well the AI performed using 1,412 whole slide photos of skin biopsies.
Three to five dermatopathologists evaluated each picture in order to get the concordance rate.
Potential benefits
Proscia intends to carry out more studies demonstrating the potential advantages of AI in assisting pathologists in making diagnoses of melanoma in addition to the current study.
Reduced misdiagnosis rates for challenging patients are one of these advantages. Pathologists differ on the diagnosis of melanoma 40% of the time since it frequently presents as benign mimickers. As cases are frequently only reviewed by one pathologist, artificial intelligence that anticipates concordance with numerous specialists might aid in improving diagnosis accuracy by acting as an additional pair of eyes.
Additionally, turnaround times for important results might be accelerated. In the United States, more than 15 million skin biopsies are performed each year, and each one might reveal hundreds of illnesses. By identifying situations that were likely to be problematic, AI that forecasts diagnostic agreement might increase efficiency by recommending further testing to give a more thorough look before pathologists analyze the cases.
Cost savings are an additional benefit. In addition to increasing the expense of healthcare, frequent overdiagnosis of melanoma causes people to undergo needless treatment and deal with the stress of being told they have a potentially fatal illness. These burdens could be reduced with improved diagnostic accuracy.
Other diagnoses
Kiran Motaparthi, director of Dermatopathology and clinical associate professor of dermatology at the University of Florida, acknowledged that melanoma can be extremely difficult to identify.
As pathologists increasingly rely on AI to uphold their commitment to providing exceptional patient care, Proscia’s technology heralds an exciting development in the field.
According to Proscia’s findings, the same AI might be used for other diagnoses with poor pathologist agreement. This includes prostate cancer Gleason grading, which is used to assess the disease’s aggressiveness, and breast cancer staging. Both frequently play a significant part in guiding therapy choices.
Sean Grullon, the chief AI scientist at Proscia and the paper’s main author, said, “With this study, we have created the foundation for a new use case of AI in pathology that might have a great influence on patient outcomes.”
Our solution uses self-supervised learning to recognize incredibly subtle patterns, demonstrating the effectiveness of one of the most cutting-edge AI methodologies.