Explainable Artificial Intelligence Analysis of Transfer Learning Models for Alzheimer’s Disease MRI Classification

2026-01-15T06:42:03+00:00

Alzheimer’s disease is a progressive neurodegenerative disorder that leads to cognitive decline and requires early and accurate diagnosis to slow disease progression. Magnetic resonance imaging (MRI) is widely used to detect structural brain changes associated with Alzheimer’s disease; however, manual interpretation of MRI scans is time-consuming and subject to observer variability. Deep learning approaches have shown strong potential in automated MRI analysis, but their black-box nature limits clinical trust and interpretability. This study proposes a transfer learning–based deep learning framework for Alzheimer’s disease classification, complemented by explainable artificial intelligence (XAI) techniques to analyze model predictions. A pretrained VGG16 model is employed to classify MRI images into four cognitive impairment categories: no impairment, very mild impairment, mild impairment, and moderate impairment. To enhance transparency, Grad-CAM, Saliency Maps, and Guided Grad-CAM are applied to visualize brain regions that contribute most to model predictions. Experimental results demonstrate that the proposed approach achieves 95.41% accuracy, indicating that a well-balanced network architecture combined with integrated explainability techniques leads to effective, interpretable classification. The visual explanations highlight clinically meaningful brain regions that align with known Alzheimer’s disease–related structural changes. These findings suggest that combining deep transfer learning with explainable artificial intelligence can provide accurate and interpretable decision support for Alzheimer’s disease diagnosis. This study is limited by the use of a single publicly available dataset and two-dimensional MRI slices, which may affect generalizability across clinical environments.

RPG-Based Educational Game for Personal Data Security Awareness in Elementary School Students: A Design and Usability Study

2026-01-13T12:57:20+00:00

As more and more elementary school-aged children use the internet, they are more likely to be exposed to cybersecurity threats, especially when it comes to keeping their personal information safe. Various educational media have been developed to introduce cybersecurity concepts to children, but most remain passive and do not engage children in simulated real-life digital risk situations. This research addresses this gap by proposing an RPG-based educational game that integrates personal data security concepts into gameplay missions tailored to the cognitive characteristics of children aged 10–12. The goal of this study was to create and assess an educational game that could serve as a substitute learning tool for personal data security. The game was developed using the Game Development Life Cycle framework and implemented using RPG Maker MV. Usability testing involved 20 elementary school students and was carried out through direct observation of 13 game scenes. The success rate indicates the number of students who were able to complete each scene independently. The test results showed that the beginning and end of the game had low success rates, indicating issues with text readability, navigation clarity, and reflective elements. The results showed that iterative improvements in the beta phase improved interface clarity and the gameplay experience. The findings in this study indicate that usability-based improvements have an important role in the design of educational games for children, and RPG-based educational games have the potential to be interactive and contextual personal data security education media.

Journal of Information Technology and Cyber Security

Explainable Artificial Intelligence Analysis of Transfer Learning Models for Alzheimer’s Disease MRI Classification

RPG-Based Educational Game for Personal Data Security Awareness in Elementary School Students: A Design and Usability Study