Hybrid Multimodal Data Fusion Model for Early Parkinson’s Disease Diagnosis

DOI

https://www.doi.org/10.59256/indjcst.20260501C016

Conference

International Conference on Cyber and AI Security

Organised By

Madanapalle Institute of Technology & Science (MITS)

Conference Date

9 & 10 March 2026

Pages

98-105

Article Type

Research Article

Check for updates

Parkinson's disease (PD) produces motor impairments which affect both speech development and cognitive abilities. The condition remains uncured, yet research demonstrates that early diagnosis with medical treatment leads to symptom reduction. The inability to access medical centres and qualified personnel prevents doctors from diagnosing patients with PD. PD is a widespread condition which causes chronic neurological movement disorders that especially affect older adults. Multimodal data fusion shows potential to deliver additional information about PD pathophysiology. The research uses its new multimodal deep learning techniques to evaluate Parkinson's disease through its ability to differentiate between people who have PD and people who do not have PD. The study used Parkinson's progression markers which were obtained through functional magnetic resonance imaging and Public Access to the PPMI and ADNI databases. The researchers aim to develop predictive models which will identify specific brain regions affected by diseases while simultaneously determining which genetic variants contribute to Parkinson's disease (PD) risk. The study investigates how genetic elements affect both the beginning and development of PD. The results show that our multimodal approach outperforms both unimodal methods and existing multimodal techniques. The Stacked Deep Learning Classifier (SDLC) system produces outstanding results through its achievement of an F1 score measurement of 0.99 and an accuracy rate of 99.4%. Proposed method shows superior performance techniques multimodal approaches. The outcomes demonstrate how our method increases both accuracy and reliability for patient data. The study demonstrates that our proposed method shows better performance results than all other methods which we tested because it achieved higher F1 score improvements in comparison to the alternative methods. The study demonstrates how exercise provides advantages to people who participate in it. The presence of multiple modalities enables estimation to occur even when one modality is not available.