Summary: Researchers have achieved near-perfect accuracy in detecting Parkinson’s disease by analyzing brain responses to emotional images and videos using EEG recordings combined with artificial intelligence. The study shows that people with Parkinson’s process emotions differently—responding more to emotional intensity than to whether an emotion is pleasant or unpleasant—and have particular difficulty recognizing fear, disgust and surprise.
Electroencephalography (EEG) data from 20 Parkinson’s patients and 20 healthy control participants were processed with machine learning methods, producing an F1 score of 0.97 for diagnostic classification. This objective, non-invasive approach to identifying Parkinson’s from emotional brain activity could complement traditional clinical evaluation and improve early detection and monitoring.
Key Facts
- Diagnostic accuracy: Emotional EEG analysis combined with AI reached an F1 score of 0.97 for distinguishing Parkinson’s disease from healthy controls.
- Distinct emotion patterns: Parkinson’s patients showed stronger sensitivity to emotional arousal (intensity) than valence (pleasantness vs. unpleasantness) and commonly confused emotions across opposite valences.
- AI integration: Multiple EEG descriptors and machine learning frameworks, including convolutional neural networks, were used to extract and classify characteristic brain-response patterns.
Source: Intelligent Computing
A collaborative research team from the University of Canberra and Kuwait College of Science and Technology has demonstrated a highly accurate method to detect Parkinson’s disease by analysing EEG responses elicited during emotional tasks such as watching curated video clips and viewing emotionally evocative images.
Moving beyond subjective assessments and purely observational diagnosis, the study provides evidence that implicit brain responses to emotions contain reliable biomarkers for Parkinson’s disease. The technique is non-invasive, repeatable, and based on measurable electrical activity recorded from the scalp, making it a promising candidate for clinical and research applications.
Published Oct. 17 in Intelligent Computing, the article titled “Exploring Electroencephalography-Based Affective Analysis and Detection of Parkinson’s Disease” reports how combining affective neuroscience with machine learning enables precise detection of Parkinson’s-related changes in emotional processing.
The study focused on implicit emotional reactions—automatic brain responses measured while participants viewed stimuli designed to elicit specific emotions—rather than relying on self-report. This approach allows researchers to probe subtle differences in emotional perception and processing that may not be apparent in clinical interviews or behavioural tests.
Using EEG recordings from 20 people diagnosed with Parkinson’s and 20 healthy controls, the team processed multiple EEG descriptors that capture different aspects of neural activity. These descriptors included spectral power vectors, which summarize power across frequency bands linked to affective states, and common spatial patterns, which enhance differences between classes by optimizing spatial filters.
After extracting relevant features from the EEG signals, researchers transformed them into visual representations and fed these into machine learning models—such as convolutional neural networks—to automatically learn distinguishing patterns. This pipeline enabled near-perfect separation of Parkinson’s patients from controls based solely on emotional brain responses.
Results indicate that Parkinson’s participants were better at encoding emotional arousal—how intense an emotion feels—than valence, the positive or negative quality of the emotion. Among specific emotions, fear, disgust and surprise were the most difficult for Parkinson’s patients to recognize, while sadness was identified more reliably. Mislabeling patterns in the data revealed frequent confusions between emotions of opposite valence, underscoring a valence-related impairment in emotion perception.
The authors emphasize that emotional EEG monitoring could be a practical and ecologically valid complement to existing diagnostic tools. By combining neurotechnology, affective computing and AI, clinicians may gain an objective, scalable assessment to support diagnosis, track disease progression, or evaluate treatment effects in Parkinson’s disease.
About this Parkinson’s disease, emotion, and AI research news
Author: Xuwen Liu
Source: Intelligent Computing
Contact: Xuwen Liu – Intelligent Computing
Image: The image is credited to Neuroscience News
Original Research: Open access. “Exploring Electroencephalography-Based Affective Analysis and Detection of Parkinson’s Disease” by Ramanathan Subramanian et al., Intelligent Computing. DOI: 10.34133/icomputing.0084
Abstract
Exploring Electroencephalography-Based Affective Analysis and Detection of Parkinson’s Disease
Although Parkinson’s disease is primarily known for motor symptoms, deficits in emotion perception are also reported. This study evaluates whether EEG signals recorded during emotional tasks can reveal differences between Parkinson’s patients and healthy controls and whether these signals can support automated detection.
Applying both traditional machine learning and deep learning approaches to a variety of EEG descriptors, the study examined dimensional (arousal and valence) and categorical emotion recognition, and compared classification performance for Parkinson’s versus healthy control groups. Findings show that Parkinson’s patients more reliably encode arousal than valence, struggle with recognizing fear, disgust and surprise, and identify sadness more accurately than other emotions. Confusions between opposite-valence emotions were common in Parkinson’s data. Crucially, emotional EEG responses provided near-perfect discrimination between Parkinson’s and healthy participants, supporting the potential of affective EEG analysis as an effective, practical tool for objective Parkinson’s diagnosis.