Permanent Neonatal Diabetes Mellitus Detection Using Machine Learning

This study aids in the early diagnosis and treatment of Permanent Neonatal Diabetes Mellitus (PNDM), enhancing the quality of life for those affected. In our study, we employed a variety of machine learning techniques, including “Logistic Regression,” “Random Forest,” and “Support Vector Machine,” to identify cases of PNDM from non-PNDM instances based on clinical signs and biomarkers. Our findings demonstrate that our proposed approach can reliably and specifically identify PNDM patients with the added benefit of using LIME (Local Interpretable Model-Agnostic Explanations) for interpretability, providing insights into the most influential diagnostic factors. Notably, the Random Forest model achieved an exceptional accuracy of 99.99%. This breakthrough underscores the broader applicability of machine learning in pediatric healthcare, offering a valuable tool for healthcare professionals and researchers for neonatal diabetes mellitus. Our approach can also be extended to other uncommon genetic diseases, enhancing early diagnosis and patient outcomes while expanding our understanding of these rare conditions.