Deep Learning-Driven 3D Eye Tracking as a Biomarker for Alzheimer's Disease Detection and Assessment.

ADEM_TEST is a binocular eye-tracking diagnostic test dataset released in association with the study “Deep Learning-Driven 3D Eye Tracking as a Biomarker for Alzheimer’s Disease Detection and Assessment.” The dataset is intended to support research on Alzheimer’s disease (AD) detection and cognitive assessment using eye-movement behavior under structured visual stimulation paradigms. This repository contains data from 20 participants, including 10 patients with Alzheimer’s disease (AD) and 10 normal controls (NC). The dataset is organized into three top-level components: AD/: subject folders for participants diagnosed with Alzheimer’s disease NC/: subject folders for cognitively normal control participants MMSE_MOCA.xlsx: subject-level clinical and demographic metadata Each subject folder contains binocular eye-movement recordings collected under 12 stimulus conditions. These conditions are designed to probe three major task paradigms relevant to cognitive function: saccade-related tasks visual search tasks visual attention tasks For each stimulus condition, the dataset provides: left-eye and right-eye eye-movement sequence files in .txt format corresponding gaze heatmaps in .png format, stored separately for the left and right eyes The metadata spreadsheet MMSE_MOCA.xlsx includes the following subject-level variables: MMSE: Mini-Mental State Examination score MoCA: Montreal Cognitive Assessment score age: age in years edu: years of education sex: biological sex (1 = male, 2 = female) disease: diagnostic group (1 = NC, 2 = AD) Dataset Summary Total participants: 20 AD participants: 10 NC participants: 10 Data type: binocular eye-tracking sequences and gaze heatmaps Task paradigms: saccade, visual search, visual attention Metadata: MMSE, MoCA, age, education, sex, disease group File Organization The dataset is structured by diagnostic group, with one folder per participant. Each participant folder contains 12 eye-movement sequence files corresponding to the 12 stimulus conditions, together with a gazeheat/ subfolder containing the associated left-eye and right-eye gaze heatmaps. The stimulus files include task names such as: antisaccade1, antisaccade2 presaccade1, presaccade2 visual_attention1 to visual_attention5 visual_search1 to visual_search3 Each heatmap image is named using the corresponding stimulus condition and eye laterality: *_l.png for the left eye *_r.png for the right eye Relation to the Associated Manuscript This dataset is provided as a research data resource associated with the manuscript “Deep Learning-Driven 3D Eye Tracking as a Biomarker for Alzheimer’s Disease Detection and Assessment.” The ADEM_TEST dataset represents a publicly shared diagnostic test dataset containing binocular eye-movement recordings, gaze heatmaps, and clinical cognitive scores for AD and NC participants. It is intended to support transparency, reproducibility, and secondary analysis of eye-tracking biomarkers for Alzheimer’s disease. Potential Applications This dataset may be used for: Alzheimer’s disease screening research digital biomarker development eye-movement analysis visual attention and visual search studies machine learning and deep learning for cognitive impairment detection multimodal or interpretable analysis of gaze behavior in neurodegenerative disease Citation If you use this dataset, please cite the associated publication and the dataset DOI.