Clinical artificial intelligence applications of vision-language foundation models

Journal article : Review

Vision-language models (VLMs) represent a transformative advance in generative artificial intelligence (AI), using multimodal data processing to enhance clinical decision-making and workflow efficiency. Built on transformer architectures, VLMs excel in tasks like image interpretation, report generation, and visual question-answering, with emerging applications in radiology, pathology, and broader clinical practice. Their potential extends to automating documentation, improving medical education, and assisting with clinical decision-making in real-time. However, successful integration requires rigorous validation to address challenges such as bias, interpretability, and safety concerns. Prospective clinical trials, health economic evaluations, and stakeholder engagement are essential to ensure equitable and effective deployment. Regulatory frameworks must evolve to accommodate VLM functionality while maintaining accountability and protecting patient safety. By balancing innovation with robust oversight, VLMs hold promise in reducing clinician workload, expanding access to expert care, and advancing precision medicine—ushering in a new era of AI-augmented healthcare.

Authors: Thirunavukarasu, AJ; Li, S; Qin, P; Nie, D; Sanghera, R

• Publication status: Published
• Peer review status: Peer reviewed
• Publisher: Public Library of Science
• Journal: PLOS Digital Health
• Volume: 5
• Issue: 6
• Pages: e0001453
• Article number: e0001453
• Publication date: 2026-06-11
• DOI: 10.1371/journal.pdig.0001453
• EISSN: 2767-3170
• ISSN: 2767-3170
• Language: English
• Subtype: Review