{"slug": "medpao-assist-edge-deployable-modality-aware-agentic-ai-for-radiology-report", "title": "MedPAO-Assist: Edge-Deployable, Modality-Aware Agentic AI for Radiology Report Structuring", "summary": "Researchers posted arXiv paper 2610.26913v1 describing MedPAO-Assist, an edge-deployable agentic framework that turns radiologist dictation and voice commands into structured radiology reports using only about 3.3 GB peak RAM and averaging roughly 31 seconds end-to-end on a standard laptop CPU. The system runs its full pipeline on a Small Language Model in the 0.6B-parameter range across an Extract-Ground-Generate architecture covering concept extraction, SNOMED-CT grounding and structured report generation, with a speech-driven human-in-the-loop correction phase, and supports four imaging modalities: Chest X-ray, Chest CT, Abdomen CT and Brain MRI. The authors report per-stage evaluations of concept extraction, ontology grounding and concept categorization plus hardware profiling, arguing the CPU-only, offline design removes the cloud and high-end GPU requirements that have bottlenecked privacy-sensitive radiology adoption.", "body_md": "arXiv:2610.26913v1 Announce Type: new \nAbstract: The adoption of Large Language Models for radiology report structuring has been largely dominated by either proprietary cloud-based models or large open models requiring high-end GPUs for deployment, which is a critical bottleneck for clinical adoption in radiology workflows that demand data privacy and offline operation. To address this, we present MedPAO-Assist, an edge-deployable agentic framework that generates structured radiology reports directly from radiologist dictation and voice commands, requiring only $\\approx3.3$ GB peak RAM and achieving an average end-to-end latency of $\\approx31$ seconds on a standard laptop CPU through an asynchronous, multi-threaded execution pipeline. The system currently supports four major imaging modalities -- Chest X-ray, Chest CT, Abdomen CT, and Brain MRI. To our knowledge, it is among the earliest agentic systems to run its entire pipeline with a Small Language Model (SLM) in the 0.6B-parameter range. This multi-agent architecture, built around an Extract-Ground-Generate paradigm, performs concept extraction, SNOMED-CT grounding, and structured report generation, followed by a speech-driven human-in-the-loop correction phase, together forming a fully \"no-touch\" reporting workflow that requires no keyboard input, external API calls, or internet connectivity. We evaluate each pipeline stage individually -- concept extraction, ontology grounding, and concept categorization and report hardware profiling results demonstrating the system's feasibility for real-time, CPU-only clinical deployment.", "url": "https://wpnews.pro/news/medpao-assist-edge-deployable-modality-aware-agentic-ai-for-radiology-report", "canonical_source": "https://arxiv.org/abs/2610.26913", "published_at": "2026-10-21 04:00:00+00:00", "updated_at": "2026-10-21 04:19:52.943626+00:00", "lang": "en", "topics": ["artificial-intelligence", "large-language-models", "ai-agents", "ai-research", "ai-safety"], "entities": ["MedPAO-Assist", "arXiv", "SNOMED-CT"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/medpao-assist-edge-deployable-modality-aware-agentic-ai-for-radiology-report", "markdown": "https://wpnews.pro/news/medpao-assist-edge-deployable-modality-aware-agentic-ai-for-radiology-report.md", "text": "https://wpnews.pro/news/medpao-assist-edge-deployable-modality-aware-agentic-ai-for-radiology-report.txt", "jsonld": "https://wpnews.pro/news/medpao-assist-edge-deployable-modality-aware-agentic-ai-for-radiology-report.jsonld"}}