The RODI mHealth app Insight: Machine Learning-Driven Identification of Digital Indicators for Neurodegenerative Disorder Detection

Abstract

Abstract Neurocognitive Disorders (NCDs) pose a significant global health concern, and early detection is crucial for optimizing therapeutic outcomes. In parallel, mobile health apps (mHealth apps) have emerged as a promising avenue for assisting individuals with cognitive deficits. Under this perspective, we pioneered the development of the RODI mHealth app, a unique method for detecting aligned with the criteria for NCDs using a series of brief tasks. We utilized the RODI app in a comprehensive study involving 182 individuals with NCD and healthy participants. The results were then processed through machine learning processes to identify underlying NCD patterns. We prioritize the tasks within RODI based on their alignment with the criteria for NCD's, thus acting as key digital indicators for the disorder. We achieve this by employing an ensemble strategy that leverages the feature importance mechanism from three contemporary classification algorithms. Our analysis revealed that tasks related to visual working memory were the most significant in distinguishing between healthy individuals and those with NCD. On the other hand, processes involving mental calculations, executive working memory, and recall were less influential in the detection process. Our study serves as a blueprint for future mHealth apps offering a guide for enhancing the detection of digital indicators for disorders and related conditions.