Machine learning algorithms on eye tracking trajectories to classify patients with spatial neglect.

Area of Science:

• Neuroscience
• Computational Neuroscience
• Medical Technology

Background:

• Eye-movement trajectories offer insights into brain information processing.
• Visuo-spatial neglect is a neurological condition affecting spatial awareness.
• Characterizing neglect from eye movements presents a significant challenge.

Purpose of the Study:

• To develop and validate methods for analyzing saccadic eye trajectories to identify visuo-spatial neglect.
• To apply machine learning algorithms for automated analysis of eye movement data in neglect patients and healthy controls.

Main Methods:

• A standardized pre-processing pipeline for eye-tracker measurements was established.
• Traditional machine learning and deep convolutional neural networks (1D and 2D) were employed for trajectory analysis.
• Performance was evaluated using the Area Under the ROC curve (AUC).

Main Results:

• Machine learning models achieved high classification accuracy (AUC 0.83-0.86) between neglect patients and controls.
• 1D convolutional neural network scores correlated with neglect severity and white matter integrity from Diffusion Tensor Imaging (DTI).
• A correlation was found between DTI findings and the superior longitudinal fasciculus (SLF), a tract often damaged in neglect.

Conclusions:

• The study presents novel pre-processing and classification methods for eye movements in neglect syndrome.
• These methods show potential for application in diagnosing other neurological diseases.
• The approach offers a pathway toward precise, sensitive, and non-invasive computer-aided diagnostic tools.