Early Detection of Autism: Researchers Propose AI-Based Framework Model

Strategic Insights -

Researchers at the University of Jordan have made a groundbreaking discovery in the early detection of autism spectrum disorder (ASD). According to a study published in Scientific Reports, the team proposes an AI-based framework model that combines attention layers, residual layers, and bidirectional long short-term memory (BiLSTM) models to improve detection and recognition performance. The model achieved impressive results, with average values for precision, recall, F1 score, and accuracy scores of 87.5%, 87%, 87.5%, and 87.7%, respectively. The study's findings suggest that the proposed model is a promising tool for early detection of ASD, reducing adverse outcomes and improving patient care.

Key Takeaways:

  • The research study was conducted by a team of researchers at the University of Jordan, led by Rami S. Alkhawaldeh.
  • The proposed AI-based framework model combines attention layers, residual layers, and BiLSTM models to improve detection and recognition performance.
  • The model achieved average values for precision, recall, F1 score, and accuracy scores of 87.5%, 87%, 87.5%, and 87.7%, respectively.
  • The study's ROC AUC values indicate that the model is robust in distinguishing autism among images, with class-specific ROC AUC values close to 95%.
  • The proposed model has the potential to reduce adverse outcomes associated with late ASD diagnosis and improve patient care.
  • The study highlights the importance of effective and automated methods for early detection of ASD.
  • The researchers propose a multi-phase pipeline that significantly improves detection and recognition performance.

Statistics:

  • Average precision score: 87.5%
  • Average recall score: 87%
  • Average F1 score: 87.5%
  • Average accuracy score: 87.7%
  • Class-specific ROC AUC values (close to 95%)
  • The proposed model demonstrates balanced performance across multiple metrics.

Sources:

  • An attention-based multi-residual and BiLSTM architecture for early diagnosis of autism spectrum disorder. Scientific Reports, 2025;15(1):33608. Nature Publishing Group - www.nature.com/; Scientific Reports - www.nature.com/srep/
  • Mental Health Weekly Digest. October 13, 2025; p 943.