Artificial Intelligence for Detection of Ventricular Oversensing Machine Learning Approaches for Noise Detection Within Non-Sustained Ventricular Tachycardia Episodes Remotely Transmitted by Pacemakers and Implantable Cardioverter Defibrillators
Résumé
Background:
Pacemakers (PMs) and implantable cardioverter defibrillators (ICDs) increasingly automatically record and remotely transmit non-sustained ventricular tachycardia (NSVT) episodes which may reveal ventricular oversensing.
Objectives:
We aimed to develop and validate a machine learning algorithm which accurately classifies NSVT episodes transmitted by PMs and ICDs in order to lighten healthcare workload burden and improve patient safety.
Methods:
PMs or ICDs (Boston Scientific) from four French hospitals with ≥1 transmitted NSVT episode were split into three subgroups: training set, validation set, and test set. Each NSVT episode was labelled as either physiological or non-physiological. Four machine learning algorithms (2DTF-CNN, 2D-DenseNet, 2DTF-VGG, and 1D-AgResNet) were developed using a training and validation dataset. Accuracies of the classifiers were compared with an analysis of the remote monitoring team of the Bordeaux University Hospital using F2 scores (favoring sensitivity over predictive positive value) using an independent test set.
Results:
807 devices transmitted 10.471 NSVT recordings (82% ICD, 18% PM), of which 87 devices (10.8%) transmitted 544 NSVT recordings with non-physiological signals. The classification by the remote monitoring team resulted in an F2 score of 0,932 (sensitivity of 95%, specificity of 99%) The four machine learning algorithms showed high and comparable F2 scores (2DTF-CNN: 0,914, 2D-DenseNet: 0,906, 2DTF-VGG: 0,863, 1D-AgResNet: 0,791) and only 1D-AgResNet had significantly different labeling as compared with the remote monitoring team.
Conclusion:
Machine learning algorithms were accurate in detecting non-physiological signals within EGMs transmitted by pacemaker and ICDs. An artificial intelligence approach may render remote monitoring less resourceful and improve patient safety.
