2026-10-06T16:28:11Z https://api.figshare.com/v2/oai

oai:figshare.com:article/34005774 2026-09-30T00:37:23Z category_28807 category_29161 category_29170 portal_12 item_type_5 month_year_09_2026

Rabia Rao: EAIP-DARV: Ensemble AI for Early Autism Screening & Referral Support Rabia Rao (17746017) Hiran Thabrew (1195092) Reza Shahamiri (8483712) Applications in health Deep learning Semi- and unsupervised learning Autism screening artificial intelligence ensemble learning calibration referral support machine learning early identification EAIP-DARV <p dir="ltr">Autism screening can be challenging when families face limited access, long pathways, cost barriers, and uncertainty about next steps. This research presents EAIP-DARV, an ensemble AI framework designed to support early autism screening and referral. The system integrates complementary AI models for screening, assessment, and correlation analysis, combining their predictions through an adaptive voting mechanism. A disagreement-aware refinement and calibration stage further improves the reliability of prediction probabilities. In evaluation, EAIP-DARV achieved 82% accuracy and 88% sensitivity, with a Brier score of 0.138 and Expected Calibration Error (ECE) of 0.031. Its UAR reached 86.0%, representing improvements over conventional screening and existing AI approaches. The framework demonstrates the potential of accessible, data-driven AI to support earlier screening and clearer referral pathways, while clinical assessment remains central to diagnosis.</p> 2026-09-30T00:37:23Z Image Poster 10.17608/k6.auckland.34005774.v2 https://figshare.com/articles/poster/EAIP-DARV_Ensemble_AI_for_Early_Autism_Screening_Referral_Support/34005774 CC BY 4.0