Catalog of Regulatory Science Tools to Help Assess New Medical Devices
This regulatory science tool presents a lab method that estimate noise signals by processing recorded ECGs containing physiologic ECG and noise/artifact components.
Technical Description
This software tool obtains an estimate of a noise signal collected as part of an electrocardiograph (ECG) recording. That is, this is a method to derive realistic noise-only signals by processing recorded ECGs containing physiologic ECG and noise/artifact components. The objective is to obtain representative samples of noise and motion artifacts from ECG devices under test that can then be used for assessing the robustness of ECG analysis algorithms and devices.
This regulatory science tool obtains noise-only signals from ECGs acquired from healthy volunteers using an ECG acquisition device (for example including electrode materials and anatomical sensor locations of the final device). These noise-only signals can be obtained from ECGs collected during various realistic settings, such as, different positions and activity levels. The obtained noise-only signal added to databases of annotated ECG recordings yields an ECG under the various noise conditions that could represent a test signal for devices and analysis algorithms developed for ECG-based applications.
Intended Purpose
The tool is intended as part of a test method for the assessment of ECG-based analysis algorithms (for example, for arrhythmia detection) and devices during noise and motion conditions. It will facilitate performance assessment of such devices and analysis algorithms. For example, the software code provides device-specific noise estimates during movement activities performed during ECG collection that can then be added to pathological ECGs at varying Signal-Noise ratios for evaluating the robustness of ECG analysis algorithms (for example, to create a device-specific noise stress test database similar to the database published in [2]).
Testing
The software tool is initially tested with successful execution over a synthetic ECG [1] by adding noise segments from the MIT-BIH Noise Stress Test (NST) database [2]. The effects of estimated noise by the software are visually assessed against the clean ECGs from select datasets. In addition, the software is assessed by quantitatively comparing the effects of the noise obtained using this software against the effects of the noise obtained by merely removing the QRS complex from the ECG signals. This assessment was separately performed using the simulated data as described in [1].
Limitations
- This software code is not intended to provide noise estimates for noise cancelation efforts/application.
- This software code is intended to estimate noise from ECG signals. Other cardiovascular activity-related signals, such as photoplethysmographic signals, are not tested and may not be suitable input signals for this software.
- The method implemented by the software code primarily relies on accurate R peak locations for proper elimination of QRS effects from the noise samples. Any inaccuracies in the detected R peaks may result in the presence of QRS complex effects in the estimated noise.
- This software code is evaluated under linearly additive noise conditions. The effects of the inherent noise recorded in the ECG have not been evaluated.
The software code has not been tested in the presence of arrhythmias or other pathological ECG conditions. In addition, it has not been tested during conditions of variable ECG morphology. Therefore, it is recommended to be used on short (<60 second) segments of ECGs collected from healthy volunteers.
Supporting Documentation
The underlying algorithm for this software is described in [1].
Galeotti, L. and C.G. Scully, A method to extract realistic artifacts from electrocardiogram recordings for robust algorithm testing. Journal of Electrocardiology, 2018. 51(6, Supplement): p. S56-S60.
Instructions for useExternal Link Disclaimer
The software code packageExternal Link Disclaimer is accessible through GitHub.
Relevant Publications
Galeotti, L. and C.G. Scully, A method to extract realistic artifacts from electrocardiogram recordings for robust algorithm testing. Journal of Electrocardiology, 2018. 51(6, Supplement): p. S56-S60. Moody GB, M.W., Mark RG, A noise stress test for arrhythmia detectors., in Computers in CardiologyExternal Link Disclaimer. 1984. p. 381-384.