Deep learning models accurately classify Parkinson's disease from eye-tracking fixation data
Gonzalo Uribarri1, Simon Ekman von Huth2, Josefine Waldthaler3
1MIDAS, Department of Computer and Systems Sciences, Stockholm University, Stockholm, Sweden.
International Journal of Medical Informatics
|July 29, 2026
Summary
Deep learning models accurately classify Parkinson's disease using eye-tracking fixation data from saccade experiments. This approach identifies novel biomarkers for neurodegenerative disease research.
Area of Science:
- Neuroscience
- Biomedical Engineering
- Machine Learning
Background:
- Eye-tracking is a non-invasive tool for assessing motor and cognitive function, relevant for neurodegenerative diseases like Parkinson's disease (PD).
- Saccade experiments show promise for studying PD progression, but a reliable eye-movement biomarker for differentiating patients from healthy individuals is lacking.
- Current methods often rely on hand-crafted features, potentially missing crucial information within raw eye-tracking data.
Purpose of the Study:
- To investigate the efficacy of deep learning models in classifying Parkinson's disease using raw eye-tracking time-series data.
- To explore the potential of fixation intervals, rather than saccade-derived features, for PD classification.
- To assess the interpretability and robustness of the best-performing classification model.
Main Methods:
- Analysis of eye-tracking data from 84 participants (54 PD patients, 30 healthy controls) during prosaccade and antisaccade tasks.
- Utilized raw fixation intervals (approx. 1.5 seconds) from the task's preparatory phase as input for deep learning models.
- Compared three time-series classifiers: InceptionTime, ROCKET, and Detach-ROCKET.
Main Results:
- The ROCKET and Detach-ROCKET models achieved high classification accuracies of 92% and 95%, respectively, on unseen individuals.
- InceptionTime achieved 78% accuracy, demonstrating varying performance among the tested deep learning architectures.
- Interpretability analysis linked model confidence scores to patient metadata and identified key features, suggesting disease-specific patterns in fixation data.
Conclusions:
- Raw fixation data from saccade experiments contains valuable information for machine learning-based Parkinson's disease classification.
- Deep learning models, particularly ROCKET variants, can effectively classify PD patients using eye-tracking fixation data, outperforming traditional feature-based approaches.
- This study supports the development of eye-tracking-based biomarkers for neurodegenerative disease discovery, offering a robust and potentially more accessible diagnostic aid.
Related Concept Videos
01:24Parkinson Disease l: Introduction
Parkinson’s disease is a chronic, progressive neurodegenerative disorder that primarily affects movement. It is characterized by motor symptoms such as resting tremors, muscle rigidity, bradykinesia (slowness of movement), and postural instability. Patients may notice hand tremors at rest, stiffness during movement, or a shuffling gait. In addition to motor features, non-motor symptoms include sleep disturbances, mood and behavioral changes, constipation, and cognitive impairment, all of which...
01:24Parkinson's Disease: Treatment
Neurodegenerative disorders, such as Parkinson's Disease (PD), involve the gradual and irreversible destruction of neurons in particular brain areas. These disorders exhibit standard features like proteinopathies, selective vulnerability of some neurons, and an interaction of intrinsic properties, genetics, and environmental influences in neural injury.
Parkinson's Disease is primarily a result of the loss of dopaminergic neurons in the substantia nigra pars compacta. The cornerstone of its...
Parkinson's Disease is primarily a result of the loss of dopaminergic neurons in the substantia nigra pars compacta. The cornerstone of its...
01:37Neural Regulation
Digestion begins with a cephalic phase that prepares the digestive system to receive food. When our brain processes visual or olfactory information about food, it triggers impulses in the cranial nerves innervating the salivary glands and stomach to prepare for food.
01:15Parkinson's Disease: Overview
Neurodegenerative disorders are progressive diseases that cause irreversible damage and loss to neurons in specific brain areas. Examples of these disorders include Parkinson's disease, Alzheimer's disease, Multiple Sclerosis (MS), and Amyotrophic Lateral Sclerosis (ALS). These disorders share characteristics such as proteinopathies, selective neuronal vulnerability, and a complex interplay between genetic and environmental factors. The primary therapeutic goal for these conditions is to...
01:24Parkinson Disease ll: Pathophysiology
Parkinson disease (PD) is a progressive neurodegenerative disorder primarily affecting movement, with additional non-motor features. Its pathophysiology involves complex interactions among genetic susceptibility, environmental exposures, and cellular dysfunction, including dopaminergic neuron loss, protein aggregation, and mitochondrial impairment.Selective NeurodegenerationA key feature is the degeneration of dopaminergic neurons in the substantia nigra pars compacta, leading to reduced...