AI-Enhanced Explainable Models for Early Detection of Neurodevelopmental Impairments in Children

Abstract

Neurodevelopmental impairments (NDIs) affect children’s brain development, affecting their physical development, speech, vision, sense of hearing, cognition, and behavior. Early detection may reduce these issues, but in rural Bangladesh, a lack of awareness and access to professionals often ends in delays. This paper introduces a web-based system utilizing the Rapid Neuro-Developmental Assessment (RNDA) framework to enable parents to remotely screen their children, thereby overcoming geographical barriers. Machine learning analysis of data from 151 patients achieved a Silhouette score of 0.30, a Davies-Bouldin Index of 1.50, and a Calinski-Harabasz Index of 14.28 for clustering. Explainable AI tools, SHAP and LIME, clarify key predictors like family income and age, ensuring transparent results. This method suggests the potential for early detection of NDI, thereby improving outcomes in environments with limited resources.

Ähnliche Arbeiten