Improving Community Health Literacy through Education on the Use of Artificial Intelligence for Alzheimer’s MRI Image Recognition

Abstract

Alzheimer’s disease and the development of artificial intelligence (AI) in medical imaging require adequate understanding to ensure that the technology is used appropriately and responsibly. This community service activity aimed to improve health literacy and AI literacy through education on the use of AI for Alzheimer’s magnetic resonance imaging (MRI) image recognition. The activity was conducted at RSUD dr. Soehadi Prijonegoro Sragen, Indonesia, from 5–26 August 2026 through education, demonstrations, discussions, and participant understanding evaluation. The materials covered Alzheimer’s disease, MRI functions, the principles of convolutional neural networks (CNN), MRI image classification, and explainable artificial intelligence (XAI) using Grad-CAM. The demonstration used 23,900 MRI images in four categories with an imbalanced class distribution. The CNN model achieved a test accuracy of 80.83%, but showed different performance across classes, with a recall of 0.06 for the Moderate Demented class and 0.91 for the Non-Demented class. These results were used to demonstrate that overall accuracy alone is insufficient to understand AI performance. Grad-CAM was used as a visual medium to introduce model interpretability, with emphasis that AI results are supporting information and not a clinical diagnosis. This activity is expected to strengthen participants’ understanding of the benefits, limitations, and responsible use of AI in healthcare.