Failures Reveal What Metrics Miss: An Evidence-Driven Agent for Recursive Refinement of ECG Classifiers

2026-07-27

AI Summary

This work addresses the limitations of existing ECG classification systems, whose optimization typically relies on manual failure analysis and struggles with coarse-grained, aggregate metrics that obscure specific error sources. To overcome this, we propose RecursiveECG, a novel framework that formalizes clinical diagnostic criteria into executable measurement functions. By integrating waveform data, derived measurements, and model predictions, RecursiveECG enables evidence-driven failure auditing. Leveraging a large language model as an offline designer, the framework recursively refines the classifier based on concrete failure cases. Evaluated on PTB-XL, Georgia, and CPSC2018 benchmarks, RecursiveECG achieves an average relative performance gain of 10.0% over strong baselines. Notably, it incurs no LLM inference overhead at deployment and supports auditable, traceable model revisions.

Abstract

Deep models have substantially advanced 12-lead ECG classification, yet their refinement still relies heavily on human experts to inspect failures and iteratively revise classifier designs. Recent LLM-based agents have demonstrated the potential for automated model design, but when guided only by aggregate performance metrics, they lack insight into why individual cases fail and how the classifier should be revised. We present RecursiveECG, an evidence-driven LLM-as-Designer framework in which an LLM serves as an offline model designer that refines ECG classifiers based on concrete failures and objective ECG evidence. To ground failure diagnosis in executable evidence, Criteria-to-Measurement Compilation converts curated ECG criteria into validated deterministic functions that produce reproducible, reference-backed measurements for individual ECGs. Building on these measurements, Evidence-Grounded Failure Review analyzes failed and comparator cases by jointly considering raw waveforms, measurements, and model outputs, enabling the LLM to diagnose classifier limitations and formulate targeted revisions. Candidate revisions are executed and re-evaluated under a fixed problem contract, and only evidence-supported updates are retained. The resulting predictor is frozen after refinement and requires no LLM inference during deployment, while an audit trail links each accepted revision to its supporting evidence. Across PTB-XL, Georgia, and CPSC2018, RecursiveECG consistently outperforms strong baselines, achieving an average relative improvement of 10.0%. Extensive ablation and transfer studies further validate the effectiveness of its evidence-grounded refinement process.

Problem

• ECG classification
• model refinement
• failure analysis
• evidence-driven design
• automated model improvement

Innovation

• Evidence-Driven Refinement
• LLM-as-Designer
• Criteria-to-Measurement Compilation
• ECG Classification
• Failure Diagnosis