Research Article
Multimodal Alzheimer’s Detection: Integrating Deep Learning on MRI Scans with Machine Learning on Genetic Data
@INPROCEEDINGS{10.4108/eai.28-4-2025.2357920, author={Savaram Jayanthi and Manikandan R and K Charitha Sai and P Himaroopa and V Jyothika}, title={Multimodal Alzheimer’s Detection: Integrating Deep Learning on MRI Scans with Machine Learning on Genetic Data}, proceedings={Proceedings of the 4th International Conference on Information Technology, Civil Innovation, Science, and Management, ICITSM 2025, 28-29 April 2025, Tiruchengode, Tamil Nadu, India, Part I}, publisher={EAI}, proceedings_a={ICITSM PART I}, year={2025}, month={10}, keywords={alzheimer's disease genetic information multimodal diagnosis explainable ai (xai) deep learning mri classification and early detection}, doi={10.4108/eai.28-4-2025.2357920} }- Savaram Jayanthi
Manikandan R
K Charitha Sai
P Himaroopa
V Jyothika
Year: 2025
Multimodal Alzheimer’s Detection: Integrating Deep Learning on MRI Scans with Machine Learning on Genetic Data
ICITSM PART I
EAI
DOI: 10.4108/eai.28-4-2025.2357920
Abstract
Alzheimer’s disease (AD) is a progressive neurodegenerative disorder that requires early detection for effective management. Existing detection methods either depends on MRI-based deep learning models or structured-data-driven machine learning approaches, restrictive accessibility due to cost constraints. This paper proposes a new multimodal system that integrates deep learning on MRI scans with machine learning on gene data, offering a flexible and cost-effective diagnostic approach. CNN architectures with a high accuracy of 92.7%, such as InceptionV3, VGG16, and ResNet50, were trained on MRI datasets. In parallel, organized clinical and genomic data were used to train machine learning methods such Random Forest, SVM, and Logistic Regression; Random Forest achieved an accuracy of 91.3%. A unified test module allows users to input MRI scans, gene data, or both for comprehensive predictions. If a patient cannot able to afford an MRI scan, the machine learning model provides an alternative, while those with MRI scans can use either gene approach or a combination for improved reliability. The proposed framework enhances accessibility and affordability in AD detection, making diagnosis adaptable to patient needs. Future work includes expanding dataset diversity, integrating Explainable AI (XAI) for model interpretability, and exploring federated learning for real-world deployment.
