Sechenov University and the Higher School of Economics have developed an AI-based diagnostic system for erythroderma, a life-threatening skin condition. The tool aims to accelerate the initial diagnosis process, allowing general practitioners to initiate treatment before a specialist consultation.
The system operates via a web platform where doctors upload patient skin images and clinical symptom data. An AI model analyzes these inputs to provide a preliminary diagnosis and treatment recommendations. The algorithm was trained over six months using anonymized data from 240 patients.
The diagnostic process involves a two-stage analysis: a computer vision model identifies skin anomalies, followed by a language model that correlates these findings with clinical symptoms. Testing on a small sample set showed a 77% accuracy rate for the system.
Developers are currently preparing to patent the technology and obtain a registration certificate for nationwide clinical implementation. Future plans include integrating histological scans to improve diagnostic precision and conducting comprehensive expert evaluations by dermatologists.