Synapse
⌘+K
Synapse
PulseExploreClubsResearchersJournals
Instagram
HomeClubsExplore
September 2, 2026Current Neurovascular Research

EEG-based diagnostic framework achieves ~99.8% accuracy for classifying Parkinson's disease.

View Full Paper
Ask AI
Bookmark
Share

Why the study?

There is a clinical need for reliable, non-invasive methods to diagnose Parkinson's disease across ON and OFF medication states using machine learning and deep learning on electroencephalogram signals.

Does a hybrid signal processing and deep learning framework using EEG signals accurately classify Parkinson's disease and medication states?

Population

15 patients with Parkinson's disease and 16 healthy controls

Comparison

Parkinson's disease (ON and OFF medication) vs healthy controls

Key result

An EEG-based diagnostic framework using Variational Mode Decomposition and a 2D-CNN achieved 99.80% accuracy, 99.79% sensitivity, and 99.80% specificity for classifying Parkinson's disease.

Authors

FLFatma LatifoğluFOFırat OrhanbulucuSPSultan Penekli

Discussion

Loading...

Member takes

Overview

Hypothesis-generating for EEG-based PD classification; requires prospective validation before clinical consideration.

Key Points

  • To develop an automated framework using signal processing and deep learning to classify Parkinson's disease and distinguish between ON and OFF medication states from EEG data.
  • Obtained EEG recordings from 15 patients with Parkinson's disease (evaluated in ON and OFF medication states, mean age 62.60 years) and 16 healthy controls (mean age 63.50 years).
  • Decomposed raw EEG signals into subbands using Variational Mode Decomposition (VMD) and generated spectrogram images with Wavelet Coherence (WC).
  • Classified the resulting 2D spectrogram representations using a two-dimensional convolutional neural network (2D-CNN).
  • Achieved a peak classification accuracy of 99.80% and a sensitivity of 99.79% using optimized EEG subband analysis.
  • Attained a specificity of 99.80% and an F1-score of 99.79% across experimental classification tasks.

Study Design

Type

Cross-Sectional (n=31)

Structured PICO

Does a hybrid signal processing and deep learning framework using EEG signals accurately classify Parkinson's disease and medication states?

P
Population
31 participants, comprising 15 patients with Parkinson's disease and 16 healthy controls, evaluated using EEG signals.
E
Exposure
Hybrid signal processing (Variational Mode Decomposition and Wavelet Coherence) and deep learning (2D-CNN) framework applied to EEG signals
C
Comparator
Healthy controls
O
Outcome
Classification accuracy, sensitivity, specificity, and F1-Scoresurrogate

A novel deep learning framework using EEG signals demonstrated exceptionally high accuracy in classifying Parkinson's disease and medication states, offering a potential non-invasive diagnostic tool.

Cite This Study

Latifoğlu et al. (2026) conducted a cross-sectional in Parkinson's disease (n=31). EEG-based signal analysis and 2D-CNN was evaluated on Classification accuracy. An EEG-based diagnostic framework using Variational Mode Decomposition and a 2D-CNN achieved 99.80% accuracy, 99.79% sensitivity, and 99.80% specificity for classifying Parkinson's disease.

synapsesocial.com/papers/6a97e2eac562ede874ec751fhttps://doi.org/10.2174/0115672026485254260821045826
View Full Paper
Ask AI
Bookmark
Share