A web-based eye-tracking system for preliminary screening for autism spectrum disorders

Abstract

A web-based eye-tracking system for preliminary screening for autism spectrum disorders, consisting of: a user interface module configured to be accessed via a web browser on an internet-enabled computer device with a camera; an eye-tracking module configured to capture real-time eye-tracking data of a user via the camera without physical contact; a stimulus presentation module configured to display structured visual stimuli, comprising at least one social stimulus and at least one non-social stimulus, on a screen; a backend processing module configured to receive the captured gaze data and extract gaze-related features such as fixation duration, gaze distribution, and viewing time proportions; an artificial intelligence analysis module configured to process the extracted gaze-related features using a trained machine learning model and generate a probability score indicating ASD-related visual attentional behavior; and a results visualization module configured to present a categorized risk-based output representing a low, moderate, or high probability of ASD, the system performs a preliminary examination and risk assessment without providing a medical diagnosis or treatment.

Description

[0001] The present invention relates generally to the field of computer-aided health screening systems and, in particular, to a web-based, non-invasive eye-tracking system that utilizes artificial intelligence and computer vision techniques for the preliminary examination and risk assessment of autism spectrum disorders (ASD). The invention lies at the interface between web technologies, eye-tracking, machine learning, and digital health informatics and aims to enable a scalable, objective, and accessible behavioral assessment using commercially available consumer computers with a webcam and internet connection.

[0002] Autism spectrum disorders (ASD) are neurodevelopmental disorders characterized by persistent difficulties in social communication and interaction, as well as restricted or repetitive patterns of behavior. Early identification of individuals at risk for ASD is critical, as timely intervention has been shown to significantly improve developmental outcomes. However, existing approaches to ASD identification have several limitations that affect their accessibility, objectivity, and scalability. Conventional methods for identifying ASD rely primarily on clinical assessments and caregiver-administered questionnaires, such as standardized behavioral observation instruments and screening checklists. These approaches are highly dependent on the availability of trained professionals, require personal judgment, and are often time-consuming and costly.Furthermore, questionnaire-based assessments are inherently subjective, as they rely on the memory, perception, and interpretation of caregivers, which can lead to fluctuations and potential biases in the results. In many regions, particularly resource-poor or rural areas, limited access to specialists results in long waiting times, delayed referrals, or a complete lack of early detection screenings. To overcome this subjectivity, laboratory-based eye-tracking systems have been introduced as an objective means of analyzing visual attention patterns associated with ASD, such as differing gaze behaviors toward social and non-social stimuli. While such systems are scientifically validated, they require specialized, costly hardware and controlled laboratory environments operated by trained personnel.These requirements significantly limit their use outside of research institutions and specialized clinics, rendering them unsuitable for large-scale or home-based screenings. Recent advances in machine learning and image processing have highlighted the potential for analyzing behavioral and visual attention data for mental health screenings. However, existing solutions are typically fragmented, hardware-dependent, or limited to experimental environments, lacking a fully integrated, browser-based platform capable of delivering immediate, automated screening results using common consumer devices. Accordingly, there is a need for a cost-effective, non-invasive, and scalable pre-screening system that objectively analyzes gaze behavior without requiring specialized equipment or expert supervision.There is also a need for a web-enabled solution that can securely capture, process, and analyze eye-tracking data in real time and provide interpretable, risk-based results to enable timely referral for formal clinical evaluation. The present invention addresses these unmet needs by providing an integrated, web-based, artificial intelligence-based eye-tracking system designed to improve accessibility and early detection while maintaining privacy and ease of use.

[0003] To solve this problem, the present invention offers a web-based eye-tracking system for preliminary screening for autism spectrum disorders.

[0004] The system enables early and easily accessible screening at home, in educational institutions and in the community via a scalable, web-based platform.

[0005] The system presents structured visual stimuli, including social and non-social (geometric) content, to elicit distinguishable gaze behaviors that are relevant for ASD risk assessment.

[0006] The system automatically extracts temporal and spatial gaze characteristics, including fixation duration, gaze distribution, and viewing time proportions, for computer-aided analysis.

[0007] The system uses artificial intelligence and deep learning models to analyze the extracted gaze features and generate a risk-based output indicating a low, medium, or high probability of ASD, without making a medical diagnosis.

[0008] The system ensures secure user authentication, data protection and integrity through controlled access and protected storage of user data and screening results.

[0009] The system offers a cloud-based Software-as-a-Service (SaaS) architecture that supports large-scale multi-user screenings across different geographic locations.

[0010] The system offers a cost-effective, privacy-compliant and user-friendly screening solution that democratizes access to preliminary autism screenings and supports early intervention measures.

[0011] The present invention relates to a web-based, non-invasive system and method for the preliminary screening and risk assessment of autism spectrum disorders (ASD) using gaze direction analysis and artificial intelligence. The invention serves to objectively evaluate visual attention patterns associated with ASD using a standard webcam and an internet-enabled computer device, thereby eliminating the need for specialized eye-tracking hardware or personal clinical monitoring. According to one embodiment of the present invention, the system presents a user with a series of structured visual stimuli consisting of paired social and non-social (geometric) images, displayed over several screening rounds.While the user views the stimuli, a browser-based eye-tracking module uses the webcam to wirelessly capture gaze coordinates, fixation durations, and saccadic movements in real time. The captured gaze data is continuously logged and transmitted to a backend processing module. This module preprocesses the raw gaze data to perform automated feature extraction and generates numerical feature vectors that include total viewing time, gaze time ratios between stimulus categories, temporal differences, and other derived metrics indicative of visual attentional behavior. These extracted features are then analyzed using a trained artificial intelligence model based on deep learning techniques to calculate a probability score indicating the likelihood of ASD-related behavioral patterns.Based on the probability score, the system generates a categorized, risk-based output that classifies the screening result into predefined levels such as low, medium, or high probability of ASD, along with the associated confidence metrics. The results are presented to the user immediately after completion of the screening process via an intuitive web interface, enabling timely referral for further professional evaluation. The system also includes secure user authentication, controlled data storage, and privacy protection mechanisms to ensure the confidentiality and integrity of user information. The invention features a scalable, cloud-based Software-as-a-Service (SaaS) architecture, allowing for widespread use in private, educational, and clinical settings.By integrating browser-based eye-gaze tracking, automated feature engineering, and artificial intelligence into a single end-to-end platform, the present invention offers an accessible, objective, and cost-effective solution for preliminary autism screening, representing a significant technological advance over existing subjective and hardware-dependent screening methods.

[0012] The present invention relates to a system for AI-supported multi-disaster prediction and security management in smart homes, designed for proactive, intelligent, and automated disaster preparedness and response in residential environments. The system integrates IoT-based sensors, AI-supported analytics, and smart home automation to enable early prediction, continuous monitoring, and real-time mitigation of multi-disasters such as earthquakes, floods, fires, gas leaks, and extreme environmental conditions.

[0013] According to the invention, a distributed network of IoT sensors is installed within and around a residential structure to continuously collect environmental and structural data. These sensors include seismic sensors for detecting ground vibrations, temperature and smoke sensors for fire detection, gas sensors for identifying hazardous leaks, and water level or humidity sensors for monitoring flooding and leaks. Additional parameters such as humidity, air quality, air pressure, and structural load can also be monitored. Furthermore, the system can receive external data input from weather forecasting services, seismic monitoring stations, satellite imagery, and historical disaster databases to enhance situational awareness and forecast accuracy.The collected data from various sources is processed by an AI processing layer that includes several machine learning and deep learning models. These models are configured to perform disaster-specific predictions and risk analyses using time-series forecasting, pattern recognition, and anomaly detection techniques. For example, seismic data can be analyzed using recurrent neural networks to identify early tremors, while flood risk can be predicted by correlating rainfall data, water level measurements, and historical flood patterns. The system uses data fusion techniques to combine heterogeneous data streams and generate localized, real-time risk assessments for each type of disaster.Based on the predicted risk levels and severity generated by AI models, an intelligent decision-making engine determines appropriate safety measures. The system is capable of autonomously triggering alarms, sending emergency notifications, and controlling connected smart home components to mitigate potential damage. These measures can include shutting off the gas or electricity supply, activating ventilation systems, controlling drainage or pumping mechanisms, activating fire suppression systems, and adjusting lighting or access controls to facilitate safe evacuation. These responses are dynamically adapted to the type of disaster, its intensity, and the specific configuration of the residential environment. Furthermore, the invention provides a user-oriented interaction interface that delivers important information to residents in real time.The interface can be accessed via mobile applications, smart displays, or voice-controlled devices and presents a live dashboard showing system status, risk levels, and recommended actions. Personalized emergency instructions and evacuation directions are generated based on the apartment's floor plan and the residents' location. In certain configurations, augmented reality-based navigation can be used to visually guide residents along the safest evacuation routes in emergency situations.

Claims

[1] A web-based eye-tracking system for pre-screening for autism spectrum disorders, consisting of: a user interface module configured to be accessed via a web browser on an internet-enabled computer device with a camera; an eye-tracking module configured to capture real-time eye-tracking data of a user via the camera without physical contact; a stimulus presentation module configured to display structured visual stimuli, comprising at least one social stimulus and at least one non-social stimulus, on a screen; a backend processing module configured to receive the captured gaze data and extract gaze-related features such as fixation duration, gaze distribution, and viewing time proportions; an artificial intelligence analysis module configured to process the extracted gaze-related features using a trained machine learning model and generate a probability score indicating ASD-related visual attentional behavior; and a results visualization module configured to present a categorized risk-based output representing a low, moderate, or high probability of ASD, the system performs a preliminary examination and risk assessment without providing a medical diagnosis or treatment. [2] System according to claim 1, wherein the eye-tracking module operates entirely within a web browser using computer vision techniques. [3] System according to claim 1, wherein the structured visual stimuli are presented in several screening rounds. [4] System according to claim 1, wherein the extracted gaze-related features include the total viewing time per stimulus, the viewing time ratios between social and non-social stimuli, the number of fixations and temporal differences. [5] System according to claim 1, wherein the artificial intelligence analysis module comprises a deep learning model trained on identified gaze data sets. [6] System according to claim 5, wherein the deep learning model comprises a convolutional neural network architecture. [7] System according to claim 1, wherein the result visualization module additionally displays confidence metrics associated with the generated probability value. [8] System according to claim 1, further comprising a user authentication module configured to securely manage user access and screening sessions. [9] System according to claim 1, wherein user data and screening results are stored in an encrypted database with controlled access.