EyePose: Pose-guided Saccadic Eye Movement Video Generation for Deep Learning-Based Neurologic Disease Phenotyping

Eye movements, including saccades, are widely regarded as highly sensitive and objective biomarkers of neurophysiologic states. Detecting saccadic signatures in neurologic diseases offers a promising, rapid, portable, and non-invasive alternative to current diagnostic tools, such as brain imaging, while overcoming access limitations and cost barriers. Currently, no robust video-oculographic solutions exist for localizing brain abnormalities due to privacy concerns and the lack of large datasets required to train accurate, reliable models. In this work, we propose the first fully synthetic, patient data-free, multimodal eye movement generation pipeline for building a generalizable dataset for saccade analysis. Using this synthetic dataset, we trained a deep learning classifier to distinguish between normal and abnormal (hypometria and hypermetria) saccadic accuracies, and evaluated its performance on real-world clinical data. The model achieved an AUROC of 0.76 and sensitivity of 0.71, showing that the synthetic data has strong potential to generalize for clinical applications. This work highlights the potential of synthetic eye movement data to be used to develop screening tools for at-home and emergency room settings.