Hybrid 3D CNN and Transformer-Based Multimodal Alzheimer’s in Privacy Constraint Smart Health Care Systems

View at publisher Summarise with VIRAT
Authors: B. Aravindha Roshan, S.P. Ramya
Cited by: 0
References: 11

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Alzheimer’s disease (AD) is an irreversible disorder that causes the loss of cognitive abilities over time including memory and the ability to conduct day to day activities. To treat AD and manage symptoms, it is extremely important that the disease be diagnosed very early in the mild cognitive impairment (MCI) stage. Unfortunately, current methods for diagnosing AD (i.e., neuro-imaging and clinical assessments) take a lot of money and time and are limited by the need for patient confidentiality in decentralized smart healthcare. This paper describes FCMT-Net a multi-modal hybrid (3D CNN and Transformer) framework for predicting MCI stage AD for use in privacy-constrained Internet of Medical Things (IoMT) enabled smart healthcare systems. The hybrid framework integrates 3D structural Magnetic Resonance Imaging (MRI) data with biomarkers that were extracted from speech processed at different levels of linguistic analysis and cognitive test score data from a single deep learning architecture. Federated learning (FL) is an employee to preserve patient privacy through decentralized model training without sharing raw data. The experimental results of FCMT-Net indicate a classification accuracy of 96.4% which is better than the accuracy of existing Convolutional Neural Network (CNN) and transfer learning based methodologies. Therefore, the results of this study demonstrate the efficacy, scalability, and confidentiality of the proposed method for next generation smart healthcare systems.

Fields of Study & Subjects

Dementia and Cognitive Impairment Research Artificial Intelligence in Healthcare Multimodal Machine Learning Applications

Publication details

Journal / source
International Conference on Computing, Communication, Control and Cyber-Physical Systems (I5CPS)
Pages
1–6
Publication date
26 February 2026
Language
EN