Application of Pre-Trained Deep Learning Models for Clinical ECGs

Automatic electrocardiogram (ECG) analysis has been one of the very early use cases for computer assisted diagnosis (CAD). Most ECG devices provide some level of automatic ECG analysis. In the recent years, Deep Learning (DL) is increasingly used for this task, with the first models that claim to perform better than human physicians. In this manuscript, a pilot study is conducted to evaluate the added value of such a DL model to existing built-in analysis with respect to clinical relevance. 29 12-lead ECGs have been analyzed with a published DL model and results are compared to build-in analysis and clinical diagnosis. We could not reproduce the results of the test data exactly, presumably due to a different runtime environment. However, the errors were in the order of rounding errors and did not affect the final classification. The excellent performance in detection of left bundle branch block and atrial fibrillation that was reported in the publication could be reproduced. The DL method and the built-in method performed similarly good for the chosen cases regarding clinical relevance. While benefit of the DL method for research can be attested and usage in training can be envisioned, evaluation of added value in clinical practice would require a more comprehensive study with further and more complex cases.

Lead one ratio: A new electrocardiogram marker for cardiac resynchronization therapy response

BACKGROUND

Wider QRS duration and presence of left bundle branch block (LBBB) predict better cardiac resynchronization therapy (CRT) response. Despite strict patient selection, one-third of patients have a sub-optimal response. We aim to evaluate the impact of lead one ratio (LOR) on CRT response.

METHODS

We enrolled 93 patients receiving CRT from August 2016 to August 2019. Pre-implant 12-lead electrocardiogram (ECG) was recorded, and LOR was derived by dividing the maximum positive deflection of QRS complex in ECG lead I by the maximum negative deflection in lead I; cut-off value of 12 was used to divide the cohort into two groups. Patients were followed for 6 months, and outcomes were compared for CRT response, New York Heart Association (NYHA) class improvement, all-cause mortality, and heart failure (HF) hospitalization events.

RESULTS

At the end of 6-month follow-up, LOR ≥ 12 was associated with significantly better CRT response (75.76% vs. 51.85% in LOR < 12, P = 0.02), lower mortality per 100 patient-years (9.09 vs. 14.81 in LOR < 12, P = 0.012), and more improvement in HF symptoms (NYHA improvement) (78.79% vs. 55.56% in LOR < 12, P = 0.02). Patients with LOR < 12 had more HF hospitalization events (2.04 vs. 1.81 episodes in LOR ≥ 12, P = 0.029) and less QRS narrowing (∆5.74 ± 2.09 vs. ∆7.10 ± 3.97 ms in LOR ≥ 12, P = 0.01). QRS duration and LBBB morphology were predictors of response in both groups of patients.

CONCLUSION

LOR ≥ 12 was associated with better response to CRT, less HF hospitalization, and more relief in HF symptoms. This ratio helps to identify possible sub-optimal response among patients with an indication for CRT.

Right Atrial Diverticulum in an Adult Woman with Left Bundle Branch Block

Right atrial diverticulum is a very rare anomaly. It is an outpouching arising from the right atrial free wall. Clinical presentations vary widely but some cases are associated with supraventricular tachycardia and atrial flutter/fibrillation. The incidence/prevalence of this anomaly is not available because only a few cases have been reported. We report a 38-year-old female patient who presented to the Heart Clinic, Tehran, Iran in 2019 with a history of dyspnea and chest pain. Electrocardiography revealed left bundle branch block. Following a magnetic resonance imaging study, the patient was diagnosed with a right atrial diverticulum. She underwent surgical resection of the diverticulum. The post-operative course was uneventful and no recurrence of the arrhythmia was detected during the six months of follow-up. To the best of the authors’ knowledge, this combination has not been described in the literature.

Is exercise stress echocardiography useful in patients with suspected obstructive coronary artery disease who have resting left bundle branch block?

BACKGROUND

Current guidelines support exercise stress echocardiography (ESE) for evaluation of suspected obstructive coronary artery disease (OCAD) in ambulant patients with left bundle branch block (LBBB). Data regarding the diagnostic utility of ESE in patients with LBBB are limited.

HYPOTHESIS

We hypothesized that the diagnostic performance of ESE for the assessment of suspected OCAD is reduced in the context of LBBB.

METHODS

We studied 191 consecutive patients with resting LBBB undergoing ESE for the investigation of suspected OCAD between 2008 and 2015 at our center. The studies were categorized as inconclusive, normal, or abnormal. Patients with an abnormal response were subcategorized as regional ischemic response or globally abnormal.

RESULTS

Eighty-two patients (43%) demonstrated a normal left ventricular contractile response (LVCR) to exercise; 92 (48%) developed an abnormal LVCR to exercise, including 70 patients with globally abnormal and 22 patients with regional ischemic responses. Of the patients with abnormal responses, 62 patients had anatomic imaging, only 29 of whom had significant OCAD, conferring an overall specificity of ESE for significant OCAD of 21% and accuracy of 52%. Of patients who developed a regionally abnormal response, 89% had significant OCAD.

CONCLUSIONS

For patients with LBBB who develop a globally abnormal LVCR during ESE, the specificity of ESE for reliably excluding significant OCAD is significantly reduced. ESE appears to be a suboptimal test for the evaluation of OCAD in patients with resting LBBB, as about 50% of patients will have an abnormal response, the majority due to globally abnormal contraction where OCAD cannot be reliably diagnosed. Alternative testing should be considered for the investigation of suspected OCAD in patients with resting LBBB.

Prevalence and clinical significance of left bundle branch block according to classical or strict definition criteria in permanent pacemaker patients

BACKGROUND

Previous studies have shown that the presence of left bundle branch block (LBBB) is associated with an increased risk of cardiac mortality and heart failure (HF). Recently, new criteria to define strict LBBB have been proposed: QRS duration ≥140 ms for men and ≥130 ms for women, along with mid-QRS notching or slurring in ≥2 contiguous leads.

HYPOTHESIS

We assessed the prevalence and prognostic significance of LBBB according to classical (QRS duration ≥120ms) and strict criteria in permanent pacemaker patients.

METHODS

We retrospectively enrolled 723 consecutive patients who had undergone single- or dual-chamber pacemaker implantation at the study center from July 2002 to December 2014. Patients with a left ventricular ejection fraction ≤35% or a prior diagnosis of HF were excluded.

RESULTS

LBBB was reported in 54 (7%) patients, and strict-LBBB in 15 (2%) patients. During a median follow-up of 48 months (range, 18-92 months), 147 (20%) patients reached the combined endpoint of death or HF hospitalization. Patients with LBBB and those with strict-LBBB displayed significantly higher rates of death or HF hospitalization (log-rank test, all P < 0.0001). In particular, strict-LBBB was associated with the worst outcome. The presence of LBBB according to classical definition criteria (hazard ratio [HR] = 1.98, confidence interval [CI]: 1.23-3.19, P = 0.005) and to strict criteria (HR = 2.20; CI: 1.04-4.65; P = 0.039) were both confirmed as independent predictors of death or HF hospitalization after adjustment for relevant clinical covariates.

CONCLUSIONS

Among patients who had undergone standard pacemaker implantation, the prevalence of LBBB was 7% according to classical definition criteria and 2% according to strict criteria. The presence of LBBB, and particularly of strict-LBBB, at the baseline predicted a poor outcome in terms of death or HF hospitalization.

Assessment of QT and JT Intervals in Patients With Left Bundle Branch Block

Background

Prolongation of the QT interval is considered a risk factor for cardiac adverse events and mortality. Left bundle branch block (LBBB) lengthens the QT interval. The corrected QT interval (QTc) is most likely overestimated because its prolongation is caused by increases in depolarization duration and not in repolarization.

Objectives

In this study, we aimed to apply corrected JT interval (JTc) as an appropriate measure of ventricular repolarization for predicting QTc in a formula.

Patients and Methods

The study population consisted of 101 patients with sinus rhythm (SR) and narrow QRS complexes (< 120 milliseconds). All patients underwent electrophysiology studies or ablation. A diagnostic catheter was positioned in the right ventricular apex (RVA) to induce LBBB at two different cycle lengths (CLs; 600 and 700 mv). The intrinsic QRS complex, QT time, and JT time were measured during SR and subsequent RVA pacing. The JTc was derived simply by subtracting the QRS duration from the QTc.

Results

Stimulation from the RVA increased the QTc from 456.20 ± 38.63 ms to 530.67 ± 47.73 ms at a CL of 600 (P < 0.0001) and to 502.32 ± 47.26 ms at 700 CL (P < 0.0001). JTc showed no significant changes with stimulation from the RVA (102.97 ± 11.35 ms vs. 103.59 ± 10.67 ms, P = 0.24). There was no significant correlation between JTc and QRS complex duration. A significant correlation was seen between QRS and QTc at both CLs. The ROC curve indicated that sensitivity of 80% and specificity of 67% were obtained with JTc duration of 92.6 ms.

Conclusions

Right ventricular pacing increases the QT interval without increasing the JT interval. Our results confirm that JTc, as an index of repolarization, is independent of ventricular depolarization. Therefore, it can be applied for predicting QTc in patients with LBBB.