SciGroveBeta
Medicine

Tracing the Heart: An Evidence-Linked Pipeline for Heart-Failure Feature Engineering

Soorya Ram Shimgekar, Michelle Hu, Dorisa Shehi, Daniel Kang, Roy Ka-Wei Lee, Koustuv Saha, Christian Poellabauer, Christopher Lee, Sajeev Singh, Piyum Zonooz, Navin Kumar, Zeeshan Ahmed, Priyadarshini Kachroo

Featured August 9, 2026

Original Paper

This analysis was generated by SciGrove. Upload your own PDFs or enter a DOI — and get the same AI breakdown on any paper.

Get started

AI-generated analysis — This is SciGrove's AI interpretation of the paper, not peer-reviewed content. Always refer to the original paper.

Simply

A new AI system automatically turns messy patient records into clear, rule-based health scores for heart failure, making it much easier for doctors to understand and predict heart conditions.

In depth
The paper introduces the Nimblemind Multi-Agent System (nMAS), an automated pipeline that transforms fragmented electronic health record (EHR) data into analysis-ready, patient-level features for heart failure. This system is evidence-linked and rubric-grounded, meaning it uses explicit clinical guidelines to derive features and maintains traceability to source data. A restricted large language model (LLM) acts as an auditor, ensuring the generated features are structurally valid and clinically coherent, significantly improving predictive performance for heart failure phenotyping.

Key Takeaways

  • 1
    The nMAS pipeline automates heart-failure feature engineering from fragmented EHRs, addressing a major bottleneck in clinical AI and research.
  • 2
    It employs an evidence-linked, rubric-grounded approach, using explicit clinical guidelines and LLM auditing to ensure features are traceable, auditable, and clinically coherent.
  • 3
    Adding nMAS-generated features significantly improved AUROC for HFrEF (0.895 to 0.963) and HFpEF (0.870 to 0.910) phenotyping, demonstrating their predictive value.

Conceptual Flow

HIGH LEVEL
1
Methodology: Automated Feature Creation

The system takes messy patient data, cleans it up, and then uses a set of rules to create new, easy-to-understand health scores, checking its work with a smart computer program.

Messy Patient Data
Clean and Score
Clear Health Scores
2
Results: Better Health Predictions

Using these new health scores, computer models became much better at predicting specific heart conditions compared to using only the original, raw patient data.

Old Prediction Score
New Prediction Score
Compare Accuracy
Much Better Prediction