A Hierarchical Computer Vision Pipeline for Physiological Data Extraction from Bedside Monitors

TL;DR

The paper tackles the interoperability gap in low-resource ICUs by digitizing vital signs from legacy bedside monitors using a lightweight, edge-friendly hierarchical computer vision pipeline. It localizes the monitor with YOLOv11, detects vital-sign ROIs, rectifies perspective, and performs OCR with PaddleOCR, achieving high end-to-end accuracy (≥98.9%) and real-time performance on modest hardware. Key contributions include a dual-YOLOv11 architecture for localization and ROI detection, a geometry-based rectification module, and a robust PP-OCRv5 extraction stage, validated on open datasets and 2,098 real-world ICU images. The approach demonstrates strong potential for practical deployment in LMICs, enabling rapid digitization of otherwise closed, non-networked monitor data without hardware replacement. Future work targets waveform digitization, edge-device optimization, and transformer-based detectors to further improve robustness and autonomy.

Abstract

In many low-resource healthcare settings, bedside monitors remain standalone legacy devices without network connectivity, creating a persistent interoperability gap that prevents seamless integration of physiological data into electronic health record (EHR) systems. To address this challenge without requiring costly hardware replacement, we present a computer vision-based pipeline for the automated capture and digitisation of vital sign data directly from bedside monitor screens. Our method employs a hierarchical detection framework combining YOLOv11 for accurate monitor and region of interest (ROI) localisation with PaddleOCR for robust text extraction. To enhance reliability across variable camera angles and lighting conditions, a geometric rectification module standardizes the screen perspective before character recognition. We evaluated the system on a dataset of 6,498 images collected from open-source corpora and real-world intensive care units in Vietnam. The model achieved a mean Average Precision (mAP@50-95) of 99.5% for monitor detection and 91.5% for vital sign ROI localisation. The end-to-end extraction accuracy exceeded 98.9% for core physiological parameters, including heart rate, oxygen saturation SpO2, and arterial blood pressure. These results demonstrate that a lightweight, camera-based approach can reliably transform unstructured information from screen captures into structured digital data, providing a practical and scalable pathway to improve information accessibility and clinical documentation in low-resource settings.

Figures (3)

• Figure 1: Confusion matrix of the ROI detection model evaluated on the OUCRU Clinical Dataset ($N=12,002$ instances).
• Figure S1: Training and validation loss curves for the Monitor Localization model (YOLOv11n-seg)
• Figure S2: Training dynamics of the YOLOv11s architecture for Vital Sign ROI detection.