A novel graph-based AI system, RSA-KG, significantly improves clinical decision support for diagnosing recurrent spontaneous abortion (RSA) by integrating multimodal clinical data and expert guidelines.
Key Details
- 1RSA-KG integrates structured and unstructured clinical data, including imaging reports, lab values, and molecular biomarkers.
- 2The knowledge graph is built from five major international RSA guidelines and extensive literature using NLP and multimodal AI models.
- 3A rigorous evaluation with 3,000 clinician-validated questions showed RSA-KG-enhanced LLMs outperform naive RAG and raw models (e.g., 86.5% vs. 76.5% accuracy for DeepSeek-R1).
- 4Qualitative expert scoring from 10 clinicians confirmed higher clinical usefulness of RSA-KG outputs over standard LLMs and other medical models.
- 5Key limitations include limited biomarker recency, single-discipline expert evaluation, and need for multicenter clinical validation.
Why It Matters
RSA-KG demonstrates a state-of-the-art integration of guidelines, imaging, and clinical data via AI, highlighting pathways to more adaptable and clinically relevant decision support systems in complex diagnostic areas. Radiology-AI professionals can leverage such frameworks for robust, multimodal reasoning and model validation.
Source
EurekAlert
Related News
•EurekAlert
Holographic Neural Network Achieves Multi-Task AI via Light-Based Computing
Researchers have developed a holographic neural network using optics to perform multiple AI classification tasks simultaneously without hardware reconfiguration.
•EurekAlert
Rethinking Multimodal Cardiovascular AI: When Models Should Defer Predictions
A commentary in Medicine Discovery argues cardiovascular AI models must learn to defer predictions when key modalities are missing, as missingness and asynchrony are clinical realities.
•EurekAlert
MSKAI Unveils AI Tool for Embryo Cell Analysis and Imaging-Driven Cancer Insights
MSK researchers employ AI and advanced imaging to reveal cancer mechanisms and track embryonic development at the single-cell level.