A Performance Comparison of Conventional and Adaptive Multi-modal Fusion Frameworks for Alzheimer’s Disease Classification

DOI

10.2991/978-94-6463-948-3_11How to use a DOI?

Keywords

Alzheimer’s Disease; ViT; CNN; Image Fusion

Abstract

Classifying Alzheimer’s Disease (AD) using multimodal neuroimaging data like Magnetic Resonance Imaging (MRI) and Positron Emission Tomography (PET) is a significant challenge in medical imaging. The effectiveness of these models relies heavily on the strategy used to combine data from both modalities. This study conducts an in-depth analysis of two fusion methods. First, we applied the traditional Discrete Wavelet Transform (DWT) technique. The second method is the Modality Attention Gate (MAG), a new fusion approach. The DWT fusion enables static spatial-domain merging. It combines the low and high-frequency elements of MRI and PET images. The MAG presents a dynamic, attention-driven feature fusion method. In this case, the model learns to give different importance to each modality during training.

We used two advanced architectures to test the reliability and adaptability of these fusion methods: a hybrid CNN+BiGRU model and a Vision Transformer (ViT). The hybrid CNN+BiGRU model captures sequential relationships in spatial features, while ViT provides global contextual information of the combined image. The performance of these models is tested on ADNI dataset. This demonstrated that MAG-based fusion consistently outperforms DWT in terms of classification accuracy, F1-score, and ROC-AUC. Moreover, interpretability analyses through attention maps further validate the clinical significance of adaptive fusion. This study positions MAG as a superior and innovative alternative to traditional fusion techniques for diagnosing Alzheimer’s disease.

Copyright

© 2025 The Author(s)

Open Access

Open Access This chapter is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

Download article (PDF)

TY  - CONF
AU  - Amar Dum
AU  - K. V. Kulhalli
PY  - 2026
DA  - 2026/01/06
TI  - A Performance Comparison of Conventional and Adaptive Multi-modal Fusion Frameworks for Alzheimer’s Disease Classification
BT  - Proceedings of the International Conference on Sustainable Innovation with Artificial Intelligence and Machine Learning 2025 (ICSIAIML 2025)
PB  - Atlantis Press
SP  - 165
EP  - 181
SN  - 1951-6851
UR  - https://doi.org/10.2991/978-94-6463-948-3_11
DO  - 10.2991/978-94-6463-948-3_11
ID  - Dum2026
ER  -