A Next-Generation Feature-Selective Deep Hybrid Architecture for Attention-Driven Parkinson's Disease Classification
Engineering, Technology & Applied Science Research · 06.06.2026 · 16 (3) · 35881-35887
DOI: 10.48084/etasr.18067 ↗Kurz gefasst
Fachartikel in der Zeitschrift Engineering, Technology & Applied Science Research, online erschienen am 06.06.2026. Verfasst von Raghuvira Pratap Adimulam und 1 weiteren Autor:innen. Die Arbeit nennt 32 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Parkinson's Disease (PD) is among the most widespread neurodegenerative disorders, characterized by cardinal symptoms including bradykinesia, shortened strides, abnormal gait and posture, and other mobility impairments. PD detection primarily relies on clinician observation and evaluation of cardinal motor symptoms. However, conventional diagnostic methods are often constrained by subjectivity, as they depend on interpreting subtle movement alterations, which may lead to misclassification. Meanwhile, non-motor signs of PD may initially be subtle and could be attributed to several other conditions. Thus, these signs are frequently neglected, making the detection of PD at an early stage difficult. To overcome these obstacles and enhance the identification and assessment of PD, artificial intelligence, particularly Machine Learning (ML) and Deep Learning (DL), has been applied. Therefore, this study presents a Robust Parkinson's Disease Classification Framework using Advanced Deep Learning (RPDCF-ADL) model. The aim is to develop a reliable model for the accurate classification of PD patients and healthy control subjects. Initially, raw data undergo preprocessing, including data cleaning and data transformation, to improve data quality and consistency. Following that, a minimum Spearman Maximum Mutual Information (mSMMI) method is utilized to select the most informative features. The selected features are then fed into a hybrid classification model that integrates a Transformer with a Deep Belief Network (T-DBN) to effectively capture complex feature dependencies and non-linear patterns for precise PD classification. Furthermore, the AdamP optimization algorithm is adopted for weight optimization. The experimental validation of the RPDCF-ADL method achieved superior accuracy values of 98.61%, 98.86%, and 97.62% compared with existing models on the PD audio, PD, and HandPD datasets, respectively.
Zitieren
APA
Adimulam, R. P., & Annamalai, S. (2026). A Next-Generation Feature-Selective Deep Hybrid Architecture for Attention-Driven Parkinson's Disease Classification. Engineering, Technology & Applied Science Research, 16(3), 35881-35887. https://doi.org/10.48084/etasr.18067
Vancouver
Adimulam RP, Annamalai S. A Next-Generation Feature-Selective Deep Hybrid Architecture for Attention-Driven Parkinson's Disease Classification. Engineering, Technology & Applied Science Research. 2026;16(3):35881-35887. doi:10.48084/etasr.18067
BibTeX
@article{Adimulam2026,
author = {Adimulam, Raghuvira Pratap and Annamalai, Suresh},
title = {A Next-Generation Feature-Selective Deep Hybrid Architecture for Attention-Driven Parkinson's Disease Classification},
journal = {Engineering, Technology & Applied Science Research},
year = {2026},
volume = {16},
number = {3},
pages = {35881--35887},
doi = {10.48084/etasr.18067}
}
EUVIMED