Method and a system for detection of eye gaze-pattern abnormalities and related neurological diseases
Inventors
De Villers-Sidani, Etienne • DROUIN-PICARO, Paul Alexandre • Desgagne, Yves
Assignees
Interested in licensing this patent?
MTEC can help explore whether this patent might be available for licensing for your application.
Publication Number
US-12396641-B2
Publication Date
2025-08-26
Expiration Date
Abstract
The present disclosure relates to a method and a system for detecting a neurological disease and an eye gaze-pattern abnormality related to the neurological disease of a user. The method comprises displaying stimulus videos on a screen of an electronic device and simultaneously filming with a camera of the electronic device to generate a video of the user's face for each one of the stimulus videos, each one of the stimulus videos corresponding to a task. The method further comprises providing a machine learning model for gaze predictions, generating the gaze predictions for each video frame of the recorded video, and determining features for each task to detect the neurological disease using a pre-trained machine learning model.
Core Innovation
The invention provides a method for detecting a neurological disease using device-camera-based eye gaze-pattern testing performed with an electronic device. A set of tasks is performed, including a calibration task and at least one of a smooth pursuit task, a fixation task, a pro-saccade task and an anti-saccade task, and each task corresponds to a distinct set of features. For each task, stimulus videos are displayed on a screen while a camera of the electronic device in proximity to the screen simultaneously films a video of a user's face.
Each stimulus video follows a predetermined continuous or disconnected path with a target appearing moving at a pre-determined speed to prompt deliberate following of the target movement during displaying. From the generated face videos, a machine learning model generates gaze predictions for each video frame, based on calibration data obtained during the calibration task and using an internal representation of the machine learning model to perform the gaze predictions. Based on the gaze predictions, values of the set of features are determined for each task.
The neurological disease is detected using a pre-trained machine learning model based on the values of the set of features. A related implementation detects neurological disease by detecting movement of the user's eye by measuring movement of areas of interest on the video of the user's face for each stimulus video and determining features for each task using a first pre-trained machine learning model before detecting the neurological disease using a second pre-trained machine learning model based on the features determined for each task.
Claims Coverage
Two independent claims define the scope, centered on an end-to-end pipeline that uses device video to generate gaze-related information from task stimuli and then applies pre-trained machine learning models for neurological disease detection. The claims cover four inventive features across the independent claims.
Multi-task stimulus video with simultaneous face video acquisition
Performing a set of tasks, each task distinct and corresponding to a distinct set of features, including a calibration task and at least one of a smooth pursuit task, a fixation task, a pro-saccade task and an anti-saccade task; displaying stimulus videos on a screen of an electronic device and simultaneously filming with a camera located in proximity to the screen to generate a video of a user's face for each stimulus video.
Calibration task-enabled gaze prediction for video frames
Using a pre-trained gaze prediction pipeline in which calibration data obtained during the calibration task is fed to perform gaze predictions, and the pre-trained model uses an internal representation to perform the gaze predictions.
Task-specific gaze feature determination followed by pre-trained neurological disease detection
Generating gaze predictions for each video frame using a machine learning model for gaze predictions, determining values of the set of features for each task based on the generated gaze predictions, and detecting the neurological disease using a pre-trained machine learning model based on the values of the set of features.
Areas of interest movement measurement and first/second pre-trained model disease detection
Detecting movement of the user's eye by measuring movement of areas of interest on the video of the user's face for each stimulus video and determining features for each task using a first pre-trained machine learning model, then detecting the neurological disease using a second pre-trained machine learning model based on the features determined for each task.
Across the independent claims, neurological disease detection is performed using pre-trained machine learning models applied to task-specific features extracted from gaze predictions or measured eye movement from device-displayed stimulus videos and simultaneously filmed user face video. The first independent claim further emphasizes calibration data being fed into another pre-trained gaze prediction model using an internal representation to generate frame-level gaze predictions prior to feature value determination.
Stated Advantages
Not explicitly described in patent.
Documented Applications
Not explicitly described in patent.
Interested in licensing this patent?