BAAI Cardiac Agent: An intelligent multimodal agent for automated reasoning and diagnosis of cardiovascular diseases from cardiac magnetic resonance imaging

A research team from the Beijing Academy of Artificial Intelligence (BAAI) has introduced the BAAI Cardiac Agent, an intelligent multimodal agent designed to automate the complex interpretation of cardiac magnetic resonance imaging (CMR). Cardiac MRI is a gold-standard diagnostic tool but is notoriously difficult to analyze, requiring specialists to manually interpret multi-sequence, multi-phase images and calculate quantitative measures like ejection fraction. The new system tackles this by integrating specialized expert models into a unified workflow that performs automated segmentation of heart structures, functional quantification, tissue characterization, and disease diagnosis, ultimately generating a structured clinical report.

In a rigorous evaluation on CMR datasets from two hospitals involving 2,413 patients and spanning seven major cardiovascular diseases, the agent demonstrated strong performance. Internally, it achieved an area under the receiver-operating-characteristic curve (AUC) exceeding 0.93, indicating high diagnostic accuracy. Its external validation score was 0.81. For core functional parameters like left ventricular ejection fraction, stroke volume, and mass, its outputs showed a Pearson correlation coefficient greater than 0.90 when compared to clinical reports, signifying excellent agreement. The system outperformed state-of-the-art models in segmentation and diagnostic tasks, and the clinical reports it generated showed high concordance with assessments from six expert radiologists of varying experience levels.

The framework's key innovation is its agent-based architecture, which dynamically orchestrates multiple specialized models to perform a coordinated, multimodal analysis of the imaging data. This approach mirrors the step-by-step reasoning a cardiologist might use, moving from identifying structures to measuring function and finally making a diagnostic judgment. By automating this labor-intensive process, the BAAI Cardiac Agent has the potential to make advanced cardiac MRI analysis more accessible, efficient, and consistent, addressing a significant bottleneck in cardiovascular care. The code for the project has been made publicly available.