<![CDATA[Background: Peripheral neuropathy (PN) is a common complication of metabolic and systemic diseases, particularly diabetes mellitus, resulting in sensory loss, pain, and motor impairment. Conventional diagnostic tools often detect PN only after irreversible nerve injury. Artificial intelligence (AI), especially machine learning (ML), has emerged as a promising tool for early diagnosis and risk prediction by integrating clinical, imaging, and genetic data. Methods: Following PRISMA 2020 guidelines, PubMed, EMBASE, IEEE Xplore, and Scopus were systematically searched up to September 2025. Studies applying ML or deep learning algorithms to PN were included, while reviews, grey literature, and studies lacking methodological details or performance metrics were excluded. Result: Our study included participants with diabetic, chemotherapy-induced, or pain-related neuropathies. Deep learning models, such as multilayer perceptrons and neural networks, achieved diagnostic accuracies of 87–93%, while classical algorithms including random forest, XGBoost, and SVM reported AUCs of 0.80–0.93. Radiomics-based SVMs using ultrasound showed external validation AUCs of 0.70–0.90. Key predictors included HbA1c, diabetes duration, lipid profile, and BMI. Conclusions: Machine learning demonstrates strong potential for improving the prediction, diagnosis, and phenotypic classification of PN. However, heterogeneity in datasets and limited external validation restrict clinical translation. Future work should focus on standardized data, multicenter validation, and interpretable AI models to facilitate integration into clinical practice.]]>
Copyrights © 2023
