Leveraging Lay and Clinical Narratives for Enhanced Transparency in Autism Diagnosis with BioBERT Deep Learning Techniques
-
By
-
Gondy Leroy
-
Himanshu Nimbarte
-
Madhuri Sai Kandula
-
Prosanta Barai
-
Sumi Lee
-
Nell Maltman
-
Sydney Rice
-
July 17, 2026
Clinical Scorecard: Leveraging Lay and Clinical Narratives for Enhanced Transparency in Autism Diagnosis with BioBERT Deep Learning Techniques
At a Glance
| Category | Detail |
|---|
| Condition | Autism Spectrum Disorder (ASD) |
| Key Mechanisms | BioBERT machine learning model for labeling autism behavioral descriptions based on DSM-5 criteria. |
| Target Population | Children suspected of having autism spectrum disorder. |
| Care Setting | Clinical and digital health environments. |
Key Highlights
- BioBERT achieved higher precision on clinical descriptions (69%) and higher recall on lay descriptions (83%).
- Lay behavioral descriptions provide diagnostically valuable information comparable to clinical observations.
- Training on clinical data yielded the best-performing diagnostic models.
- Integration of lay information could accelerate autism diagnosis.
- AI-generated summaries were moderately representative of behavioral examples.
Guideline-Based Recommendations
Diagnosis
- Utilize both lay and clinical behavioral descriptions for a comprehensive assessment.
Management
- Incorporate digital health approaches to enhance decision-making in autism diagnosis.
Monitoring & Follow-up
- Evaluate the diagnostic utility of both lay and clinical examples across multiple dimensions.
Risks
- Potential performance drop when transferring models between lay and clinical data types.
Patient & Prescribing Data
Children with suspected autism spectrum disorder.
Early intervention is crucial for improving long-term outcomes.
Clinical Best Practices
- Combine parental input with clinical observations to improve diagnostic accuracy.
- Consider cultural factors in behavioral assessments to avoid biases.
Related Resources & Content
