PanDerm was evaluated on a wide range of clinical tasks and trained on multiple image types
By Lori Solomon HealthDay Reporter
TUESDAY, June 17, 2025 (HealthDay News) — The PanDerm artificial intelligence (AI) model improves skin cancer diagnosis when used by doctors, according to a research article published online June 6 in Nature Medicine.
Siyuan Yan, from Monash University in Melbourne, Australia, and colleagues developed PanDerm, which was pretrained through self-supervised learning on more than 2 million real-world skin disease images from 11 clinical institutions across four imaging modalities. Then, PanDerm was evaluated using 28 diverse benchmarks, including skin cancer screening, risk stratification, differential diagnosis of common and rare skin conditions, lesion segmentation, longitudinal monitoring, and metastasis prediction and prognosis.
The researchers found that PanDerm’s performance across all evaluated tasks was state-of-the-art, even outperforming existing models when using only 10 percent of labeled data. PanDerm outperformed clinicians by 10.2 percent in early-stage melanoma detection through a longitudinal analysis, improved clinicians’ skin cancer diagnostic accuracy by 11 percent on dermoscopy images, and enhanced nondermatologist health care providers’ differential diagnosis by 16.5 percent across 128 skin conditions on clinical photographs.
“Previous AI models have struggled to integrate and process various data types and imaging methods, reducing their usefulness to doctors in different real-world settings,” co-lead author Zongyuan Ge, Ph.D., also from Monash University, said in a statement. “PanDerm is a tool designed to work alongside clinicians, helping them interpret complex imaging data and make informed decisions with more confidence.”
Several authors disclosed ties to the pharmaceutical and biotechnology industries.
Copyright © 2025 HealthDay. All rights reserved.
