Toward Automated Regulatory Decision-Making: Trustworthy Medical Device Risk Classification with Multimodal Transformers and Self-Training

Toward Automated Regulatory Decision-Making: Trustworthy Medical Device Risk Classification with Multimodal Transformers and Self-Training Yu Han, Aaron Ceross, and Jeroen H.M. Bergmann May 2, 2025 Abstract Accurate classification of medical device risk levels is essential for regulatory oversight and clinical safety. We present a Transformer-based multimodal framework that integrates textual descriptions and visual information to predict device regulatory classification. The model incorporates a cross-attention mechanism to capture intermodal dependencies and employs a self-training strategy for improved generalization under limited supervision. Experiments on a real-world regulatory dataset demonstrate that our approach achieves up to 90.4% accuracy and 97.9% AUROC, significantly outperforming text-only (77.2%) and image-only (54.8%) baselines. Compared to standard multimodal fusion, the self-training mechanism improved SVM performance by 3.3 percentage points in accuracy (from 87.1% to 90.4%) and 1.4 points in macro-F1, suggesting that pseudo-labeling can effectively enhance generalization under limited supervision. Ablation studies further confirm the complementary benefits of both cross-modal attention and self-training. We also evaluate the model's robustness to noisy inputs and modality-specific perturbations, demonstrating resilience in low-resource settings. Beyond technical gains, we highlight applications in automated pre-screening, compliance verification, and support for harmonization frameworks such as UDI and GMDN. Our findings suggest that domain-adapted multimodal models, when designed with transparency and specificity, can serve as trustworthy decision-support tools for regulatory classification tasks. Index terms: Medical device classification; Multimodal learning; Transformer models; Regulatory science; Self-training; Trustworthy AI. 1 Introduction Medical device classification is a foundational task in regulatory compliance, inventory management, and clinical decision-making.