A Hybrid Framework for Symptom-Based Nerve Weakness Detection Using Machine Learning and Rule-Based Methods

Pawan Kumar Badhan,
A Hybrid Framework for Symptom-Based Nerve Weakness Detection Using Machine Learning and Rule-Based Methods,
Franklin Open,
2025,
100414,
ISSN 2773-1863,
https://doi.org/10.1016/j.fraope.2025.100414.
(https://www.sciencedirect.com/science/article/pii/S2773186325002026)
Abstract: Nerve weakness, marked by symptoms such as numbness, muscle stiffness, and memory loss, poses significant diagnostic challenges due to its diverse clinical manifestations. This study introduces a multimodal diagnostic framework that combines rule-based reasoning with advanced machine learning and deep learning techniques to improve the accuracy of detecting nerve-related disorders. The first objective is to develop a robust rule-based system that maps specific symptoms to underlying neurological conditions using a comprehensive and heterogeneous dataset. The second goal is to enhance this baseline model by integrating multimodal data inputs and leveraging ML/DL algorithms to boost diagnostic precision. The third aim is to enable early detection, timely intervention, and potential prevention of neurological and systemic disorders such as Multiple Sclerosis (MS), Parkinson’s Disease, Alzheimer’s Disease, and related conditions. The dataset includes over 1,000 patient records collected from Punjab and other regions of India, capturing symptom onset and neurological indicators such as muscle weakness, headaches, and memory decline. Data sources include clinical interviews, audio/video recordings, wearable sensors, and diagnostic tools such as Electromyography (EMG), Magnetic Resonance Imaging (MRI), and Electroencephalography (EEG). Various machine learning and deep learning models—including Convolutional Neural Networks (CNNs)—were employed to process and analyze these multimodal inputs. The proposed hybrid diagnostic framework achieved accuracy levels between 91% and 97%, notably outperforming the standalone rule-based model, which achieved between 82% and 92%. Additionally, Extreme Gradient Boosting (XGBoost) was applied to further enhance predictive performance through gradient-based optimization.
Keywords: Symptom analysis; Neurological Disorders; Nerve Conduction Studies; Clinical Decision Support; Nerve weakness