The 43rd International Society for Computerized Electrocardiology ECG initiative for the automated detection of strict left bundle branch block

Detection of strict left bundle branch block by neural network and a method to test detection consistency

OBJECTIVE

To develop an automatic algorithm to detect strict left bundle branch block (LBBB) on electrocardiograms (ECG) and propose a procedure to test the consistency of neural network detections.

APPROACH

The database for the classification of strict LBBB was provided by Telemetric and Holter ECG Warehouse. It contained 10 s ECGs taken from the MADIT-CRT clinical trial. The database was divided into a training dataset (N  =  300, strict LBBB  =  174, non-strict LBBB  =  126) and a test dataset (N  =  302, strict LBBB  =  156, non-strict LBBB  =  146). LBBB-related features were extracted by Philips DXL^™ algorithm, selected by a random forest classifier, and fed into a 5-layer neural network (NN) for the classification of strict LBBB on the training dataset. The performance of NN on the test dataset was compared to two random forest classifiers, an algorithm applying strict LBBB criteria, a wavelet-based approach, and a support-vector-machine approach. The consistency of NN's detection was tested on 549 2 min recordings of the PTB diagnostic ECG database. LBBB annotations are not required to measure consistency.

MAIN RESULTS

The performance of NN on the test dataset were sensitivity  =  91. 7%, specificity  =  85.6% and accuracy  =  88.7% (PPV  =  87.2%, NPV  =  90.6%). The consistency score of strict-LBBB and non-strict-LBBB detection was 0.9341 and 0.9973 respectively.

CONCLUSION

NN achieved the highest specificity, accuracy, and PPV. Using random forest for feature selection and NN for classification increased interpretability and reduced computational cost. The consistency test showed that NN achieved high consistency scores in the detection of strict LBBB.

SIGNIFICANCE

This work proposed an approach for selecting features and training NN for the detection of strict LBBB as well as a consistency test for black-box algorithms.

The relationship between ECG predictors of cardiac resynchronization therapy benefit

INTRODUCTION

Cardiac resynchronization therapy (CRT) is an effective treatment that reduces mortality and improves cardiac function in patients with left bundle branch block (LBBB). However, about 30% of patients passing the current criteria do not benefit or benefit only a little from CRT. Three predictors of benefit based on different ECG properties were compared: 1) "strict" left bundle branch block classification (SLBBB); 2) QRS area; 3) ventricular electrical delay (VED) which defines the septal-lateral conduction delay. These predictors have never been analyzed concurrently. We analyzed the relationship between them on a subset of 602 records from the MADIT-CRT trial.

METHODS & RESULTS

SLBBB classification was performed by two experts; QRS area and VED were computed fully automatically. High-frequency QRS (HFQRS) maps were used to inspect conduction abnormalities. The correlation between SLBBB and other predictors was R = 0.613, 0.523 and 0.390 for VED, QRS area in Z lead, and QRS duration, respectively. Scatter plots were used to pick up disagreement between the predictors. The majority of SLBBB subjects- 295 of 330 (89%)-are supposed to respond positively to CRT according to the VED and QRS area, though 93 of 272 (34%) non-SLBBB should also benefit from CRT according to the VED and QRS area.

CONCLUSION

SLBBB classification is limited by the proper setting of cut-off values. In addition, it is too "strict" and excludes patients that may benefit from CRT therapy. QRS area and VED are clearly defined parameters. They may be used to optimize biventricular stimulation. Detailed analysis of conduction irregularities with CRT optimization should be based on HFQRS maps.

Automatic diagnosis of strict left bundle branch block using a wavelet-based approach

Patients with left bundle branch block (LBBB) are known to have a good clinical response to cardiac resynchronization therapy. However, the high number of false positive diagnosis obtained with the conventional LBBB criteria limits the effectiveness of this therapy, which has yielded to the definition of new stricter criteria. They require prolonged QRS duration, a QS or rS pattern in the QRS complexes at leads V1 and V2 and the presence of mid-QRS notch/slurs in 2 leads within V1, V2, V5, V6, I and aVL. The aim of this work was to develop and assess a fully-automatic algorithm for strict LBBB diagnosis based on the wavelet transform. Twelve-lead, high-resolution, 10-second ECGs from 602 patients enrolled in the MADIT-CRT trial were available. Data were labelled for strict LBBB by 2 independent experts and divided into training (n = 300) and validation sets (n = 302) for assessing algorithm performance. After QRS detection, a wavelet-based delineator was used to detect individual QRS waves (Q, R, S), QRS onsets and ends, and to identify the morphological QRS pattern on each standard lead. Then, multilead QRS boundaries were defined in order to compute the global QRS duration. Finally, an automatic algorithm for notch/slur detection within the QRS complex was applied based on the same wavelet approach used for delineation. In the validation set, LBBB was diagnosed with a sensitivity and specificity of Se = 92.9% and Sp = 65.1% (Acc = 79.5%, PPV = 74% and NPV = 89.6%). The results confirmed that diagnosis of strict LBBB can be done based on a fully automatic extraction of temporal and morphological QRS features. However, it became evident that consensus in the definition of QRS duration as well as notch and slurs definitions is necessary in order to guarantee accurate and repeatable diagnosis of complete LBBB.