Machine learning models of 6-lead ECGs for the interpretation of left ventricular hypertrophy (LVH).

BACKGROUND: Left Ventricular Hypertrophy (LVH) is closely linked to the cardiovascular disease prognosis, and thus, timely diagnosis improves outcomes. Diagnosis is challenging due to dependency on doctor's visits and a 12­lead ECG. In addition, the interpretation of LVH from ECGs is challenging due to variability of ECG measurements, body habitus, electrode positioning, several LVH ECG criteria and EP mechanisms. The aims of this study are to evaluate different big data-driven machine learning models for ECG LVH interpretation based on limb leads only, and to compare the performance of an ECG parameter-based statistical model with a deep learning-based model. METHODS AND DATA: The first two models are binary class Random Forest (RF) models, an ensemble learning method which constructs many decision trees at training time and predicts the class chosen by the greatest number of trees at inference time. One random forest is trained using the following five features: lead aVL R-wave amplitude, lead I, II, aVL ST segment amplitude, and QRS duration. The second RF model uses 54 features across all limb leads, including the five features used by the smaller model. The second type of model is a multi-class deep neural network (DNN) which takes median beats of 6 limb leads arranged in Cabrera sequence as input. The signal preprocessing included forming median beats, filtering with a 40-Hz lowpass filter, and down-sampling to 125 Hz. The DNN models consist of 1 lead-formation convolutional layer, 5 downsampling convolutional resnet blocks with skip connections, and 3 fully connected layers. The training dataset consisted of 1 million 10-s 12­lead ECGs, and an independent test dataset consisted of 250,000 10-s ECGs from the Mayo Clinic. RESULTS: The five-parameter RF model has the prediction performance of Area Under the Receiver-Operator Curve (AUC) 0.78, and the larger RF model had AUC of 0.83. The DNN model for ECG LVH detection achieves AUC 0.92 using only the limb leads, compared to an AUC of 0.98 for the full 12­lead DNN. CONCLUSION: The study shows that machine learning models trained only on limb leads achieve promising results with potential to add clinical value to early detection mechanisms. We also observe that the RF model splits parameters by thresholds known to be characteristic of LVH, and that the DNN model can automatically detect morphology differences from 6 limb lead ECGs. This will be meaningful for expanding the capabilities of potential electrical LVH detection in mobile 6­lead ECG devices.

Atrial fibrillation detection in outpatient electrocardiogram monitoring: An algorithmic crowdsourcing approach.

BACKGROUND: Atrial fibrillation (AFib) is the most common cardiac arrhythmia associated with stroke, blood clots, heart failure, coronary artery disease, and/or death. Multiple methods have been proposed for AFib detection, with varying performances, but no single approach appears to be optimal. We hypothesized that each state-of-the-art algorithm is appropriate for different subsets of patients and provides some independent information. Therefore, a set of suitably chosen algorithms, combined in a weighted voting framework, will provide a superior performance to any single algorithm. METHODS: We investigate and modify 38 state-of-the-art AFib classification algorithms for a single-lead ambulatory electrocardiogram (ECG) monitoring device. All algorithms are ranked using a random forest classifier and an expert-labeled training dataset of 2,532 recordings. The seven top-ranked algorithms are combined by using an optimized weighting approach. RESULTS: The proposed fusion algorithm, when validated on a separate test dataset consisting of 4,644 recordings, resulted in an area under the receiver operating characteristic (ROC) curve of 0.99. The sensitivity, specificity, positive-predictive-value (PPV), negative-predictive-value (NPV), and F1-score of the proposed algorithm were 0.93, 0.97, 0.87, 0.99, and 0.90, respectively, which were all superior to any single algorithm or any previously published. CONCLUSION: This study demonstrates how a set of well-chosen independent algorithms and a voting mechanism to fuse the outputs of the algorithms, outperforms any single state-of-the-art algorithm for AFib detection. The proposed framework is a case study for the general notion of crowdsourcing between open-source algorithms in healthcare applications. The extension of this framework to similar applications may significantly save time, effort, and resources, by combining readily existing algorithms. It is also a step toward the democratization of artificial intelligence and its application in healthcare.

Development and Validation of a Deep-Learning Model to Screen for Hyperkalemia From the Electrocardiogram.

Importance: For patients with chronic kidney disease (CKD), hyperkalemia is common, associated with fatal arrhythmias, and often asymptomatic, while guideline-directed monitoring of serum potassium is underused. A deep-learning model that enables noninvasive hyperkalemia screening from the electrocardiogram (ECG) may improve detection of this life-threatening condition. Objective: To evaluate the performance of a deep-learning model in detection of hyperkalemia from the ECG in patients with CKD. Design, Setting, and Participants: A deep convolutional neural network (DNN) was trained using 1 576 581 ECGs from 449 380 patients seen at Mayo Clinic, Rochester, Minnesota, from 1994 to 2017. The DNN was trained using 2 (leads I and II) or 4 (leads I, II, V3, and V5) ECG leads to detect serum potassium levels of 5.5 mEq/L or less (to convert to millimoles per liter, multiply by 1) and was validated using retrospective data from the Mayo Clinic in Minnesota, Florida, and Arizona. The validation included 61 965 patients with stage 3 or greater CKD. Each patient had a serum potassium count drawn within 4 hours after their ECG was recorded. Data were analyzed between April 12, 2018, and June 25, 2018. Exposures: Use of a deep-learning model. Main Outcomes and Measures: Area under the receiver operating characteristic curve (AUC) and sensitivity and specificity, with serum potassium level as the reference standard. The model was evaluated at 2 operating points, 1 for equal specificity and sensitivity and another for high (90%) sensitivity. Results: Of the total 1 638 546 ECGs, 908 000 (55%) were from men. The prevalence of hyperkalemia in the 3 validation data sets ranged from 2.6% (n = 1282 of 50 099; Minnesota) to 4.8% (n = 287 of 6011; Florida). Using ECG leads I and II, the AUC of the deep-learning model was 0.883 (95% CI, 0.873-0.893) for Minnesota, 0.860 (95% CI, 0.837-0.883) for Florida, and 0.853 (95% CI, 0.830-0.877) for Arizona. Using a 90% sensitivity operating point, the sensitivity was 90.2% (95% CI, 88.4%-91.7%) and specificity was 63.2% (95% CI, 62.7%-63.6%) for Minnesota; the sensitivity was 91.3% (95% CI, 87.4%-94.3%) and specificity was 54.7% (95% CI, 53.4%-56.0%) for Florida; and the sensitivity was 88.9% (95% CI, 84.5%-92.4%) and specificity was 55.0% (95% CI, 53.7%-56.3%) for Arizona. Conclusions and Relevance: In this study, using only 2 ECG leads, a deep-learning model detected hyperkalemia in patients with renal disease with an AUC of 0.853 to 0.883. The application of artificial intelligence to the ECG may enable screening for hyperkalemia. Prospective studies are warranted.