Chelyabinsk Student Develops Neural Network to Predict Arrhythmia Recurrence
The system not only identifies patients at risk but also explains how it reaches its conclusions.
Alexey Sedelkov, a young researcher at Chelyabinsk State University, has developed an AI system designed to assess the risk of recurrent atrial fibrillation, one of the most dangerous forms of arrhythmia.
Crucially, the algorithm does more than identify patients at risk: it also explains which indicators led to each prediction.
The neural network was trained on data from 97 patients treated between 2010 and 2013. Each case was initially described using 27 clinical features, including laboratory test results and measurements from diagnostic procedures. After preprocessing, the number of factors was reduced to 18, which the algorithm uses to make its predictions. It classifies patients as high risk only when their probability of recurrent atrial fibrillation exceeds 91%. The system is now a ready-to-use decision-support tool for clinical cardiology.
