Punjabi University, PGI Chandigarh Develop AI Tool that Outperforms Dermatologists in Diagnosing Rare Skin Disorders
Advertisement
The research team plans to further expand the dataset by including additional rare AIBD subtypes and integrating clinical information such as laboratory findings and lesion locations.
Punjabi University and PGI Chandigarh have developed an artificial intelligence-based diagnostic tool capable of identifying rare autoimmune skin disorders with high accuracy, marking a significant step toward expanding access to specialist-supported dermatology care.
Autoimmune Blistering Diseases (AIBDs) are a group of rare and often debilitating skin conditions that cause severe blistering of the skin and mucous membranes.
Diagnosing these disorders is frequently complex because several disease subtypes share similar clinical features, making it difficult to distinguish between them through routine examination alone.
Accurate diagnosis often requires specialised laboratory investigations available primarily at tertiary healthcare centres.
Against this backdrop, researchers from the Department of Computer Science at Punjabi University and the Department of Dermatology at the Postgraduate Institute of Medical Education and Research (PGI), Chandigarh, have developed and validated AI-driven techniques to assist in the diagnosis of these conditions.
The study was carried out by researcher Manbir Singh under the supervision of Maninder Singh and the co-supervision of Dipankar De.
The researchers noted that confirming AIBDs typically involves advanced diagnostic procedures such as Direct Immunofluorescence (DIF), considered the gold-standard test, along with Indirect Immunofluorescence (IIF) and Enzyme-Linked Immunosorbent Assay (ELISA) testing.
To develop the AI system, the research team created a clinically validated dataset using real patient images collected through PGIMER’s Department of Dermatology.
The team assessed nearly 240 artificial intelligence model configurations spanning classical machine-learning approaches, hybrid architectures and advanced deep-learning techniques.
According to the researchers, the AI models consistently achieved better performance than participating dermatologists in classifying different subtypes of autoimmune blistering diseases.
The research team plans to further expand the dataset by including additional rare AIBD subtypes and integrating clinical information such as patient age, sex, demographics, laboratory findings and lesion locations.
Stay tuned for more such updates on Digital Health News
