Hardway, Heather

Abstract Background Screening for left ventricular (LV) systolic dysfunction (defined as ejection fraction ≤35%) based on data from a standard 12-lead electrocardiogram (ECG) has become well established when standard digital ECGs are available–8 independent leads sampled at least 250 hertz for 10 seconds. As the algorithm has been incorporated into various clinical scenarios and ancillary research projects, a limitation of the binary classification at 35% has become apparent. Purpose The objective of this study was to develop and validate a deep learning-based algorithm that would classify LVEF into three categories based on only the digital ECG input. Methods After IRB approval, native digital resting ECGs acquired between 1/1/2010 and 12/31/2021 on patients seen in Mayo Clinic in Jacksonville were extracted from the institutional electronic ECG database management system (MUSE, GE Healthcare). These ECGs were matched with transthoracic echocardiograms obtained up to four days prior or 30 days after the ECGs acquisition. A convolutional neural network consisting of 8 layers of convolutions, batch normalization and pooling was trained using Keras and Tensorflow with hyper-parameter optimization for L1 and L2 regularization, learning rate adjustments, and class weights to predict three classes of LVEF: ≤35%, 36–51%, and ≥52% based on clinical relevance. The primary measure of overall performance was the detection of LVEF ≤35%; however, the triad of model predictions was also considered in translating the model output to human interpretable findings. Results A total of 30,153 patients (60,169 ECG pairings; mean age 63 years; 48% male) were randomly split at the patient level into training (24,172 patients), validation (2,973 patients) and testing (3,008 patients). The trained model provided robust discrimination in the withheld testing data – AUROC of 0.941 (95% CI: 0.931 to 0.950). Using the optimal model threshold based on Youden's index from the validation data (0.186), sensitivity and specificity were estimated to be 87.9% (95% CI: 83.8% to 91.2%) and 86.3% (95% CI: 85.4% to 87.2%) in the testing data. In instances where discordant predictions were observed, the posterior distribution of model probabilities provide additional insights into the possible underlying value of LVEF (Figure 1). Conclusions The utilization of a multi-category deep learning classification model for the detection of reduced ejection fraction adds new dimensions to the use of AI technologies on digital ECGs. This work shows high discrimination can still be obtained when using three classes of LVEF. Funding Acknowledgement Type of funding sources: None.

Abstract Background Current approaches utilizing non-invasive methods to screen for cardiac allograft rejection (gene expression profiling and cell free DNA) have yet to be broadly integrated into heart transplant management and have shown limited discrimination (AUCs of 0.68 and 0.77, respectively). Changes in the electrocardiogram (ECG) have been reported at the time of severe cardiac rejection, including low voltages and conduction abnormalities. It remains unknown if subtle ECG changes correlating with cardiac allograft rejection can be detected earlier using machine learning methods. Purpose We sought to develop an artificial intelligence (AI) model to detect cardiac allograft rejection based on the 12 lead ECG. Methods We identified all patients who underwent a heart transplant at 3 hospital sites within a single health system from Jan 1998 through Apr 2021 and extracted digital 12-lead ECG data as well as endomyocardial biopsy pathology results from the electronic medical record. We partitioned our data into a training (80%), validation (10%), and test set (10%) based on a group stratification sampling. Each patient was present in only one set and each set had a positivity rate of 2.6% with 6,074/758/758 ECGs belonging to 1,146/140/141 unique patients in each set respectively. Cardiac allograft rejection was defined as moderate or severe acute cellular rejection based on International Society for Heart and Lung Transplantation (ISHLT) guidelines. A convolutional neural network, using the 12-lead ECG data as input, was trained with hyperparameter optimization for regularization, learning rate adjustments, and class weights. Model performance metrics were based on the test data and estimated using the final model architecture. Results 1,587 heart transplant recipients who had at least one endomyocardial biopsy were evaluated for inclusion. We limited our sample to ECGs performed within 30 days of the biopsy date (7,590 ECGs, representing 1,425 unique patients). Our study population had a median age of 55.8 years and 28.7% were female. The median number of ECG-biopsy pairs per patient was 5. The majority of endomyocardial biopsy results were classified as none or mild rejection (97.1%), and 2.9% had moderate/severe rejection. The ECG-based AI model detected cardiac allograft rejection with an area under the receiver operative curve (AUC) of 0.84 in the test set. The sensitivity, specificity, positive and negative predictive values were 95%, 52.6%. 5.2% and 99.7% respectively. Conclusions An AI-ECG model appears to outperform novel non-invasive laboratory tests (gene expression profiling and cell free DNA) for detecting cardiac allograft rejection and does not require a blood draw or the additional complexities surrounding sample processing. This model relies on a readily available and relatively inexpensive test, the ECG. In addition, AI predictions can be made available within a few minutes following ECG acquisition. Funding Acknowledgement Type of funding sources: Private hospital(s). Main funding source(s): Mayo Clinic