Abstract:Background: Researchers have developed machine learning-based ECG diagnostic algorithms that match or even surpass cardiologist level of performance. However, most of them cannot be used in real-world, as older generation ECG machines do not permit installation of new algorithms. Objective: To develop a smartphone application that automatically extract ECG waveforms from photos and to convert them to voltage-time series for downstream analysis by a variety of diagnostic algorithms built by researchers. Methods: A novel approach of using objective detection and image segmentation models to automatically extract ECG waveforms from photos taken by clinicians was devised. Modular machine learning models were developed to sequentially perform waveform identification, gridline removal, and scale calibration. The extracted data were then analysed using a machine learning-based cardiac rhythm classifier. Results: Waveforms from 40 516 scanned and 444 photographed ECGs were automatically extracted. 12 828 of 13 258 (96.8%) scanned and 5399 of 5743 (94.0%) photographed waveforms were correctly cropped and labelled. 11 604 of 12 735 (91.1%) scanned and 5062 of 5752 (88.0%) photographed waveforms achieved successful voltage-time signal extraction after automatic gridline and background noise removal. In a proof-of-concept demonstration, an atrial fibrillation diagnostic algorithm achieved 91.3% sensitivity, 94.2% specificity, 95.6% positive predictive value, 88.6% negative predictive value and 93.4% F1 score, using photos of ECGs as input. Conclusion: Object detection and image segmentation models allow automatic extraction of ECG signals from photos for downstream diagnostics. This novel pipeline circumvents the need for costly ECG hardware upgrades, thereby paving the way for large-scale implementation of machine learning-based diagnostic algorithms.

Tracing Algorithms of ECG Tracks from Scanned Image

Continuously Tracing Algorithms of ECGtracks from Scanned Image

Solving the Electrocardiography Inverse Problem by Using an Optimal Algorithm Based on the Total Least Squares Theory

The Combination of Self-Organizing Feature Maps and Support Vector Regression for Solving the Inverse ECG Problem

Knowledge Discovery Based Auxiliary Diagnosis Algorithm for Electrocardiograph On-Line Data

Algorithm for Clustering Analysis of ECG Data

The Strategy of T Wave Detection and QT Interval Extraction

Automated Optical Reading of Scanned ECGs

Extraction of ECG ST-T segment and its system implementation

Development and Validation of an Algorithm for the Digitization of ECG Paper Images

Non-Invasive Locating of Premature Ventricular Contraction Origin with Low Rank/Tv Regularization

Paper-Recorded ECG Digitization Method with Automatic Reference Voltage Selection for Telemonitoring and Diagnosis

Research on intelligent algorithm for detecting ECG R waves

Practical Fine-Grained Learning Based Anomaly Classification for ECG Image.

A Crucial Wave Detection and Delineation Method for Twelve-Lead ECG Signals

Research on Representation and Similarity Measurement of ECG Series *

A low-complexity ECG feature extraction algorithm for mobile healthcare applications

A Novel Method Based on Shifted Rank-1 Reconstruction for Removing EMG Artifacts in ECG Signals

Automatic detection of cardiac conditions from photos of electrocardiogram captured by smartphones

A novel abnormal ECG beats detection method

ECG_SegNet: An ECG delineation model based on the encoder-decoder structure