# Developing an explainable machine learning and fog computing-based visual rating scale for the prediction of dementia progression

## TL;DR

This study introduces a fog computing-based system that uses mental test scores to predict dementia progression more accurately than traditional methods.

## Contribution

A novel fog computing and machine learning framework for real-time dementia screening using MMSE and MoCA scores.

## Key findings

- The proposed ML model achieved 0.93 accuracy, outperforming MoCA (0.90) and MMSE (0.83).
- The system effectively categorized 450 subjects into MCI, PD, and AD classes with high precision.
- The fog computing approach enables real-time, context-aware monitoring for early dementia detection.

## Abstract

Recently, dementia research has primarily concentrated on using Magnetic Resonance Imaging (MRI) to develop learning models in processing and analyzing brain data. However, these models often cannot provide early detection of affected brain regions. Alternatively, mental test scores such as the Mini-Mental State Examination (MMSE) and Montreal Cognitive Assessment (MoCA) offer valuable insights into the likelihood of dementia and cognitive impairments. The main objective of this study is to introduce an innovative and dependable context-aware health monitoring system based on fog computing to measure mental impairment in the elderly population. The framework provides screening tests utilizing MMSE and MoCA to achieve accurate and real-time monitoring of cognitive function, allowing for early detection and treatment of mental disorders. To assess the effectiveness of our screening test, we evaluated a dataset comprising 450 subjects with Mild Cognitive Impairment (MCI) from Kaferelshikh University. The aggregated dataset is categorized into three classes: (1) 150 patients with MCI, (2) 150 subjects with subcortical diseases, Parkinson’s Disease (PD), and (3) 150 subjects with cortical diseases, Alzheimer’s Disease (AD). To accurately determine health risks, we employ an ensemble AdaBoost model, providing superior performance in accuracy, precision, recall, F-score, and Area Under the Curve (AUC). To validate the effectiveness of our Machine Learning (ML) model on unseen data, we evaluate an additional 18 subjects using the proposed scoring test, with six subjects from each class. The results indicate that our proposed ML model achieves an impressive accuracy of 0.93, outperforming the MoCA score (0.90) and MMSE score (0.83). Through our research, we demonstrate the potential of our context-aware fog computing approach in significantly enhancing early diagnosis of dementia, leveraging mental test scores as valuable indicators.