Akademik Veri Yönetim Sistemi
2025 33rd Signal Processing and Communications Applications Conference (SIU), İstanbul, Türkiye, 25 - 28 Haziran 2025, ss.1-4, (Tam Metin Bildiri)
Early diagnosis of developmental, neurological, and functional disorders is of critical importance for improving individuals’ quality of life and developing appropriate intervention strategies. This study aims to distinguish between dyslexia, obstetric brachial plexus injury (OBPI), intellectual disability (ID), and typical development (TD) classes through the analysis of physiological signals. Data were collected while participants played two different serious games while wearing Empatica E4 wristbands; blood volume pulse (BVP), electrodermal activity (EDA), and skin temperature (ST) signals were obtained. Stress labels given by three different experts were added to each segment. Using the obtained data, logistic regression and multilayer perceptron artificial neural network models were developed and comparative classification analysis was performed. The logistic regression model performed classification using a one-vs-all approach, while the neural network model directly produced multi-class outputs. With the logistic regression model, 60%, 85%, 78%, and 95% accuracy was achieved in dyslexia, OBPI, intellectual disability, and typical development classes, respectively. The neural network model provided accuracy of 93%, 92%, 92%, and 99% in the same order. The overall accuracy rate reached 92% in the neural network model. The findings indicate that physiological signals are an effective tool in the classification of neurodevelopmental disorders, and artificial intelligence-based models can support diagnostic processes.