A novel deep learning approach for diagnosing Alzheimer's disease based on eye-tracking data

Abstract

Eye-tracking technology has become a powerful tool for biomedical-related applications due to its simplicity of operation and low requirements on patient language skills. This study aims to use the machine-learning models and deep-learning networks to identify key features of eye movements in Alzheimer's Disease (AD) under specific visual tasks, thereby facilitating computer-aided diagnosis of AD. Firstly, a three-dimensional (3D) visuospatial memory task is designed to provide participants with visual stimuli while their eye-movement data are recorded and used to build an eye-tracking dataset. Then, we propose a novel deep-learning-based model for identifying patients with Alzheimer's Disease (PwAD) and healthy controls (HCs) based on the collected eye-movement data. The proposed model utilizes a nested autoencoder network to extract the eye-movement features from the generated fixation heatmaps and a weight adaptive network layer for the feature fusion, which can preserve as much useful information as possible for the final binary classification. To fully verify the performance of the proposed model, we also design two types of models based on traditional machine-learning and typical deep-learning for comparison. Furthermore, we have also done ablation experiments to verify the effectiveness of each module of the proposed network. Finally, these models are evaluated by four-fold cross-validation on the built eye-tracking dataset. The proposed model shows 85% average accuracy in AD recognition, outperforming machine-learning methods and other typical deep-learning networks.

Eye-tracking paradigms for the assessment of mild cognitive impairment: a systematic review

“…A growing body of evidence suggests that gaze metrics are useful in the screening of individuals at risk of diseases, including AD, Parkinson’s, Autism spectrum disorder, and nystagmus syndrome ( Rosengren et al, 2020 ; Kong et al, 2022 ; Sun et al, 2022 ). Further investigations of specific eye movement biomarkers and neuropsychological criteria that precisely separate MCI subtypes (aMCI and naMCI) may assist in the forecasting of dementia progression.…”

“…However, the lack of large-scale eye-tracking datasets is a limiting factor for using deep-learning models for the recognition or classification of AD-related MCI based on eye movement data. Therefore, it is important for the scientific community to establish access to such databases in order to advance the development of machine-learning and deep-learning-based models for identifying cognitive function impairment with higher sensitivity ( Fabrizio et al, 2021 ; Haque et al, 2021 ; Miltiadous et al, 2021 ; Rutkowski et al, 2021 ; Rizzo et al, 2022 ; Sun et al, 2022 ). While remarkable advancements are occurring in the field of digital health sector ( Dagum, 2018 ; Kourtis et al, 2019 ; Topol, 2019 ; Khan and Javaid, 2021 ), the challenge is for healthcare system leaders to stay abreast of the latest findings and information about gaze metrics as an emerging option for cognitive screening.…”

“…Hence, although the presented studies ( Oyama et al, 2019 ; Tadokoro et al, 2021 ) demonstrate practical eye-tracking tests for grading the cognitive state of older adults, their scientific conclusions clearly show a direction for further improvements (for example, modifying/replacing some blocks). Finally, in order to transform valuable observations related to atypical gaze patterns into applicable cognitive scoring tests, it is essential to carry out longitudinal studies in laboratory- and home-based settings, and report the results in equally accessible publications ( Sbordone, 1996 ; Tarnanas et al, 2013 ; Sun et al, 2022 ).…”

Neural Computation-Based Methods for the Early Diagnosis and Prognosis of Alzheimer’s Disease Not Using Neuroimaging Biomarkers: A Systematic Review

“…Proposed in [ 140 ], the “NeAE-Eye” model consisted of three main modules: an “inner autoencoder” (built with a shallow convolutional network with 8-layers), an “outer autoencoder” (with a fully connected neural network, FCN), and a “classifier” (3-layers fully connected network). Using eye-tracking data, this nested AE model surpassed the other methods on a 3D Visual Paired Comparison task.…”

Quantification of identifying cognitive impairment using olfactory-stimulated functional near-infrared spectroscopy with machine learning: a post hoc analysis of a diagnostic trial and validation of an external additional trial

“…Previous studies classifying AD stages using a novel diagnostic method and machine learning investigated wearable EEG ( n = 26) [ 22 ], eye-tracking ( n = 210) [ 23 ], and various genetic or serum biomarkers [ 24 ]. However, previous studies have provided little evidence due to the small sample size, lack of an extra-validation dataset, lack of reported feature importance, and use of an observation study dataset [ 22 – 24 ].…”