QAI/ML-SaMD: A Hybrid Health–Technology Quantifiable Quality Metric for Artificial Intelligence/Machine Learning-Based Software as a Medical Device

Background: The increasing integration of Artificial Intelligence (AI) and Machine Learning (ML) into medical devices necessitates robust quality evaluation methods. However, existing approaches remain qualitative, sector-specific, or focused on isolated attributes, leaving a gap in quantifiable assessment for AI/ML-driven Software as a Medical Device (SaMD). Objective: This study introduces QAI/ML-SaMD, a novel hybrid metric that provides a comprehensive, quantifiable measure of AI/ML-SaMD quality by synthesizing health and information technology (IT) dimensions into a single composite, benchmark-ready score. Methods: The metric integrates key attributes from a systematic literature review, classified into Health and IT domains. Sub-metrics (QHealth and QIT) use weighted sums, while the overall score employs a Weighted Geometric Mean with configurable parameters to penalize domain imbalances. Validation included (a) theoretical validation against four mathematical properties, (b) an illustrative example with sensitivity analysis, (c) expert-based validation with six specialists, and (d) an evidence-based case study on FDA-authorized IDx-DR using public regulatory and clinical documentation. Results: The illustrative example yielded a score of 29.7 (“Unsuitable”). Sensitivity analysis confirmed robustness across weight, score, and combined uncertainty perturbations, with classification unchanged. Expert validation showed 83.3% agreement. The IDx-DR case study produced a score of 82.3 (“Admissible”), correctly aligning with the device’s regulatory status and supporting external validity. Conclusions: The QAI/ML-SaMD metric provides a foundational, quantifiable framework for AI/ML-SaMD quality assessment, bridging qualitative regulatory principles and measurable outcomes. It offers a practical tool for developers, regulators, and clinicians to benchmark and track quality across the SaMD lifecycle.

Preparing paper insights from the available abstract and paper details.

Authors

Shouki A. Ebad

Published

Aug 17, 2026

Journal or venue

Healthcare

DOI

10.3390/healthcare14162587

Paper type

journal article

Source

Crossref