An LLM-enhanced AI agent for rapid diagnosis of acute aortic dissection in multi-center settings

Abstract

Summary Acute aortic dissection (AAD) is a high-mortality cardiovascular emergency, yet early diagnosis before computed tomography angiography (CTA) remains challenging. We developed AAD-Agent as a pre-imaging triage aid—not as a replacement for CTA—and evaluated it in a retrospective, multi-center study. Using 869 suspected patients from 3 Chinese hospitals for internal validation and 200 from 2 additional hospitals for external validation, we compared AAD-Agent against a machine learning (ML) ensemble model. The ML-ensemble model performed well internally (accuracy 0.948, F1-score 0.964) but deteriorated externally (0.540, 0.600). In contrast, AAD-Agent maintained external stability across LLMs: DeepSeek-R1 (accuracy 0.715, F1-score 0.820), GPT-3.5 (0.715, 0.825), and GPT-4o (0.710, 0.803). Although the case-enriched design inflates these metrics, prevalence-adjusted negative predictive value (≈99.6% at 1% prevalence) supports AAD-Agent as a safe, low-cost rule-out aid to defer CTA in low-risk patients. Its low external specificity precludes confirmatory diagnosis, limiting use to rule out pending prospective validation.