What question did this study set out to answer?

This study aims to identify cognitive biomarkers of Parkinson's disease by comparing machine learning models based on EEG and pupillometry data.

July 26, 2026Brain SciencesOpen Access

A Comparison of Machine Learning Models for Classification of Parkinson’s Disease During a Working Memory and Sustained Attention Task

View Full Paper

Ask AI

Bookmark

Authors

Randomized trial compares machine learning models for classifying Parkinson's disease cognitive deficits, implying potential for biomarker discovery.

This study aims to identify cognitive biomarkers of Parkinson's disease by comparing machine learning models based on EEG and pupillometry data.
EEG and pupillometry data recorded from Parkinson's disease patients and healthy controls during cognitive tasks.
108 features extracted and processed using PCA and recursive elimination, classified through SVM-RBF with 5-fold cross-validation.
Class balancing performed using SMOTE and feature selection limited to training folds to prevent leakage.
WM achieved 63% accuracy (F1 = 0.626) and ATTN achieved 70% accuracy (F1 = 0.623) on an independent hold-out set.
ATTN showed higher accuracy (73%) and precision while WM demonstrated higher recall (71%).
Univariate analyses indicated significant features linked to ATTN and corroborated important neurophysiological signatures.

Terry et al. (2026) studied this question.

View Full Paper

Ask AI

Bookmark

Randomized trial compares machine learning models for classifying Parkinson's disease cognitive deficits, implying potential for biomarker discovery.