Trustworthy Clinical Decision Pipeline for Autism Diagnosis

Abstract

Trustworthy Clinical Decision Pipeline has led to the growing popularity of Autism Spectrum Disorder (ASD) which has increased the demand of statistically justifiable and interpretable diagnostic pipelines capable of improving clinical decision-making without undermining transparency, or ethical responsibility. The study presents a Trustworthy Clinical Decision Pipeline (TCDP) which is a combination of machine learning, IoT-based behavioral monitoring, and the principles of NIST AI Risk Management Framework (AI RMF) governance. The system focuses on tracing the data, model interpretability, and clinician supervision on every decision level. The pipeline was evaluated on a multimodal pediatric dataset and achieved a diagnostic accuracy of 93 percent, as well as high explainability and data privacy standards. Findings indicate that the integration of ethical AI governance in a diagnostic process positively impacts the reliability of the process, reduces bias, and promotes clinician confidence in automated assessments of autism.