VLM-in-the-Loop: A Plug-In Quality Assurance Module for ECG Digitization Pipelines

TL;DR

This paper presents VLM-in-the-Loop, a plug-in module that enhances ECG digitization quality by integrating visual language models with domain-specific signal analysis, improving accuracy and robustness.

Contribution

It introduces a novel tool-grounding approach that anchors VLM assessments in quantitative signal evidence, applicable without modifying existing digitization systems.

Findings

Tool grounding increased verdict consistency from 71% to 89%.

Doubled fidelity separation ($b4$PCC 0.03 to 0.08).

Improved quality metrics across multiple backends and datasets.

Abstract

ECG digitization could unlock billions of archived clinical records, yet existing methods collapse on real-world images despite strong benchmark numbers. We introduce \textbf{VLM-in-the-Loop}, a plug-in quality assurance module that wraps any digitization backend with closed-loop VLM feedback via a standardized interface, requiring no modification to the underlying digitizer. The core mechanism is \textbf{tool grounding}: anchoring VLM assessment in quantitative evidence from domain-specific signal analysis tools. In a controlled ablation on 200 records with paired ground truth, tool grounding raises verdict consistency from 71\% to 89\% and doubles fidelity separation ($\Delta$PCC 0.03 $\rightarrow$ 0.08), with the effect replicating across three VLMs (Claude Opus~4, GPT-4o, Gemini~2.5 Pro), confirming a pattern-level rather than model-specific gain. Deployed across four backends, the module improves every one: 29.4\% of borderline leads improved on our pipeline; 41.2\% of failed limb leads recovered on ECG-Digitiser; valid leads per image doubled on Open-ECG-Digitizer (2.5 $\rightarrow$ 5.8). On 428 real clinical HCM images, the integrated system reaches 98.0\% Excellent quality. Both the plug-in architecture and tool-grounding mechanism are domain-parametric, suggesting broader applicability wherever quality criteria are objectively measurable.