Automated AI based identification of autism spectrum disorder from home videos

Enterprise AI Analysis: Automated AI based identification of autism spectrum disorder from home videos

Automated AI based identification of autism spectrum disorder from home videos

This study developed an AI-based screening system for early autism spectrum disorder (ASD) detection using short home-recorded videos (name-response, imitation, ball-playing). The system extracts task-specific and common behavioral features using deep learning and combines them with demographic data via machine learning classifiers. The ensemble model achieved an AUROC of 0.83 and an accuracy of 0.75, demonstrating a scalable, objective, and ecologically valid approach to complement clinical evaluations and enable earlier intervention.

Schedule a Live Demo

Executive Impact

Our AI-driven solution delivers measurable improvements in early detection and operational efficiency, translating directly to enhanced patient outcomes and resource utilization.

0.83AUROC (Ensemble Model)

0.75Accuracy (Ensemble Model)

14.2sInference Time per Video (seconds)

510Children in Dataset

Deep Analysis & Enterprise Applications

Select a topic to dive deeper, then explore the specific findings from the research, rebuilt as interactive, enterprise-focused modules.

Current Challenges

Methodological Advantage

AI System Architecture

Clinical Alignment

Critical Gap in Early ASD Diagnosis

3.5-4 yearsAverage age of ASD diagnosis worldwide

This is considerably later than the ideal window for early intervention (before age 2), highlighting the need for earlier screening.

AI vs. Traditional ASD Assessment
Feature	Traditional Methods	AI-based Home Videos
Administration	In-person, clinic-based	Home-based, parent-recorded
Time/Cost	Time-consuming, high cost	Rapid (14.2s/video), cost-free (open-source models)
Expertise	Trained professionals required	Automated, no specialized training needed
Objectivity	Prone to observer bias, caregiver recall variability	Objective, AI-extracted features
Ecological Validity	Atypical behaviors in clinic	Natural behaviors in familiar settings
Scalability	Limited by resources and access	Highly scalable, supports early identification

Enterprise Process Flow

Home Videos (Name-response, Imitation, Ball-playing)

→

Video Screening & Quality Control

→

Deep Learning Modules (STT, Key-point, Ball Detector)

→

Feature Extraction (Gaze, Motion, Parental Attempts)

→

Machine Learning Classifiers

→

Ensemble Model & ASD Probability

Clinical Relevance of AI-Extracted Features

The AI model identified several key behavioral indicators consistent with clinical understanding of ASD.

For the name-response task, longer response latency and elevated variability in parental calling attempts were strongly associated with ASD predictions. This reflects difficulties in social orienting and responsiveness.

In the imitation task, reduced eye contact duration, diminished physical engagement, and delayed imitation responses were key drivers of ASD classification, consistent with impairments in motor imitation and joint attention.

The ball-playing task showed prolonged turn-taking durations and reduced eye contact contributing to ASD predictions, indicating challenges in reciprocal social engagement and coordination.

These findings underscore the model's ability to capture the multi-faceted nature of ASD symptomatology objectively.

Quantify Your ROI

Our AI-powered video analysis system can significantly reduce the time and cost associated with initial ASD screenings, allowing healthcare providers to reallocate resources and streamline patient pathways. By automating the preliminary assessment, clinics can decrease waiting times, improve early intervention rates, and ultimately enhance patient outcomes and operational efficiency.

Industry Sector

Number of Employees (or cases per year)

Average Hours Spent per Case (manual processing)

Average Hourly Cost per Employee ($)

Projected Annual Savings$0

Hours Reclaimed Annually0

Your AI Implementation Roadmap

A phased approach ensures seamless integration and maximum impact for your enterprise.

Phase 1: Pilot & Integration (2-4 Weeks)

Initial setup of the AI screening platform within a pilot clinical setting. Integrate with existing EMR/patient management systems. Train clinical staff on video collection protocols and AI report interpretation.

Phase 2: Data Validation & Refinement (4-8 Weeks)

Collect and analyze initial screening data from a larger cohort. Compare AI predictions with traditional diagnostic outcomes to refine model performance and ensure clinical accuracy. Gather feedback from parents and clinicians.

Phase 3: Scalable Rollout & Monitoring (8-12 Weeks)

Expand the AI screening system across multiple clinics or healthcare networks. Implement continuous monitoring of model performance and user feedback. Develop advanced features, such as personalized intervention recommendations based on AI insights.

Ready to Transform Your Operations with AI?

Book a complimentary strategy session to explore how our advanced AI solutions can drive efficiency and innovation in your enterprise.

Enterprise AI Analysis: Automated AI based identification of autism spectrum disorder from home videos