Identifying patients with atrial fibrillation during sinus rhythm on ECG: confirming the utility of artificial intelligence algorithm in a small-scale cohort without structural heart diseases

Background

Detection of atrial fibrillation (AF) out of electrocardiograph (ECG) on sinus rhythm (SR) using artificial intelligence (AI) algorithm has been widely studied within recent couple of years. Generally, it is believed that a huge number of ECGs are necessary for developing an AI-enabled ECG to be adequate to correspond to a lot of minor variations of ECGs. For example, structural heart diseases have typical ECG characteristics, but they could be a noise for the purpose of detecting the small signs of electrocardiographic signature of AF. We hypothesized that when patients with structural heart diseases are excluded, AI-enabled ECG for identifying patients with AF can be developed with a small number of ECGs.

Methods

We developed an AI-enabled ECG using a convolutional neural network to detect the electrocardiographic signature of AF present during normal sinus rhythm (NSR) using a digital, standard 10-second, 12-lead ECGs. We included all patients who newly visited the Cardiovascular Institute with at least one NSR ECG between Feb 1, 2010, and March 31, 2018. We classified patients with at least one ECG with a rhythm of AF as positive for AF (AF label) and others as negative for AF (SR label). We allocated ECGs to the training, internal validation, and testing datasets in a 7:1:2 ratio. We calculated the area under the curve (AUC) of the receiver operating characteristic curve for the internal validation dataset to select a probability threshold, which we applied to the testing dataset. We evaluated model performance on the testing dataset by calculating the AUC and the sensitivity, specificity, F1 score, and accuracy with two-sided 95% confidence intervals (CIs).

Results

We totally included 19170 patients with 12-lead ECG. After excluding patients with structural heart diseases, 12825 patients with NSR ECGs at the initial visit were identified (1262 were clinically diagnosed as AF anytime during the time course and 11563 were never diagnosed as AF). Of 11563 non-AF patients, 1818 patients who were followed over 1095 days were selected for the analysis with the SR label, to secure the robustness for maintaining SR. Of 1262 AF patients, 251 patients were selected for the analysis with the AF label, of whom a NSR ECG within 31 days before or after the index AF ECG (the first AF ECG during the time course) could be obtained. In the patients with AF label, the NSR ECG of which the date was the nearest to the index AF ECG was selected for the analysis. The AI-enabled ECG showed an AUC of 0.88 (0.84–0.92) with sensitivity 81% (72–88), specificity 80% (77–83), F1 score 50% (43–57), and overall accuracy 80% (78–83).

Conclusion

An AI-enabled ECG acquired during NSR allowed identification of patients with AF in a small population without structural heart diseases.

Funding Acknowledgement

Type of funding sources: None.

Prediction of atrial fibrillation by 12-lead electrocardiogram parameters in patients without structural heart disease

Background

Recently, the analysis of electrocardiogram (ECG) waveform by artificial intelligence has been reported to pick out those who have atrial fibrillation (AF) or have a high potential of developing AF, which, however, cannot explain the mechanisms or algorisms for the prediction from its nature.

Purpose

The purpose of this study is to conduct a comprehensive analysis to investigate the difference of weighting in predicting capability for AF among hundreds of automatically-measured ECG parameters using a single ECG at sinus rhythm.

Methods and results

Out of Shinken Database 2010–2017 (n=19170), 12825 patients were extracted, where those with ECG showing AF rhythm at the initial visit (including all persistent/permanent AF and a part of paroxysmal AF) and those with structural heart diseases were excluded. Out of 639 automatically-measured ECG parameters in MUSE data management system (GE Healthcare, USA), 438 were used. [Analysis 1] A predicting model for paroxysmal AF were determined by logistic regression analysis (Total, n=12825; paroxysmal AF, n=1138), showing a high predictive capability (AUC = 0.780, p<0.001). In this model, the relative contribution of ECG parameters (by coefficient of determination) according to the time phase were P:72.4%, QRS:32.7%, and ST-T:13.7%, respectively (Figure A). [Analysis 2] Excluding AF at baseline, a predicting model for new-developed AF were determined by Cox regression analysis (Total, n=11687; new-developed AF, n=87), showing a high predictive capability (AUC = 0.887, p<0.001). In this model, the relative contribution of parameters (by log likelihood) according to the time phase were P:40.8%, QRS:42.5%, and ST-T:24.9%, respectively (Figure B).

Conclusions

We determined ECG parameters that potentially contribute to picking up existing AF or predicting future development of AF, where the measurement of P wave strongly contributed in the former whereas all time phases were similarly important in the latter.

Weighting of parameters to predict AF

Funding Acknowledgement

Type of funding source: Private hospital(s). Main funding source(s): Self funding of the institute

Heart rate recovery after exercise as a prognostic predictor in patients with atrial fibrillation

Background

A delayed heart rate recovery (HRR) after exercise is related to mortality in sinus rhythm. This study aimed to investigate this concept can be applied to patients with atrial fibrillation (AF).

Methods

We analyzed 483 patients with AF (mean 65 years, male 74%). HRR integral was calculated by integrating the difference in HR in every 3 second between the end of exercise and the specified time after the exercise (30, 60, 120 and 180 seconds). After evaluating the prognostic power of each HRR integral, we selected HRR integral of 180 seconds (180HRR-integral).

Results

We divided the patients into two groups using median value of 180HRR-integral. All-cause mortality, the incidence of cardiovascular events and heart failure events were higher in the poor 180HRR-integral. After adjustment for covariates, the impact of the high 180HRR-integral for all-cause mortality was 3.15 (p=0.057), 1.77 for cardiovascular events (p=0.067) and 1.28 for heart failure events (p=0.519).

Conclusion

Poor HRR was associated with worse prognosis in patients with AF.

Funding Acknowledgement

Type of funding source: None

153 Multiparametric assessment of diastolic function in heart failure

Background

The assessment of diastolic function is still challenging in the setting of heart failure (HF). We tested the hypothesis that applying a machine learning algorithm would detect heterogeneity in diastolic function and improve risk stratification in HF population.

Methods

This study included consecutive 279 patients with clinically stable HF referred for echocardiographic assessment, for whom diastolic function variables were measured according to the current guidelines. Cluster analysis, an unsupervised machine learning algorithm, was undertaken on these variables to form homogeneous groups of patients with similar profiles of the variables. Sequential Cox models paralleling the clinical sequence of HF assessment were used to elucidate the benefit of cluster-based classification over guidelines-based classification. The primary endpoint was a hospitalization for worsening HF.

Results

Cluster analysis identified 3 clusters with distinct properties of diastolic function that shared similarities with guidelines-based classification. The clusters were associated with brain natriuretic peptide level (p < 0.001, figure A). During follow-up period of 2.6 ± 2.0 years, 62 patients (22%) experienced the primary endpoint. Cluster-based classification exhibited a significant prognostic value (c2 = 20.3, p < 0.001, figure B), independent from and incremental to an established clinical risk score for HF (MAGGIC score) and left ventricular end-diastolic volume (hazard ratio = 1.677, p = 0.017, model c2: from 47.5 to 54.1, p = 0.015, figure D). Although guideline-based classification showed a significant prognostic value (c2 = 13.1, p = 0.001, figure C), it did not significantly improve overall prognostication from the baseline (model c2: from 47.5 to 49.9, p = 0.199, figure D).

Conclusion

Machine learning techniques help grading diastolic function and stratifying the risk for decompensation in HF.

Abstract 153 Figure.

P289 Aortic valve resistance risk-stratifies low-gradient aortic stenosis

Background

Grading the severity of aortic stenosis (AS) is challenging, since there is a discrepancy between aortic valve area (AVA) and mean pressure gradient (mPG). Arotic valve resistance (RES) has been proposed as a usuful descriptor of AS severity, but it is not commonly used for clinical decision-making, because its robust validation of clinical-outcome efficacy is lacking. This study aimed to investigate whether RES holds an incremental value for risk-stratifying AS.

Methods

This study recuited 565 AS patients (AVA < 1.5cm²) referred to echocardiography for valve assessment. The patients were divided into three different groups, according to the guidelines: high-gradient AS (HG-AS, mPG≥40mmHg, n = 157), low-gradient AS (LG-AS, mPG < 40mmHg + AVA ≤ 1.0cm², n = 155) and moderate AS (Mod-AS, mPG < 40mmHg + AVA > 1.0cm², n = 253). RES was calculated from Doppler measurement of mPG and stoke volume. The diagnositic cutoff point for RES was determined at 190 dynes × s×cm-5 by substituting AVA = 1.0cm² and mPG = 40mmHg into the definition formula of RES and Gorlin formula. The patients were followed up for 2 years. The endpoint was a composite of cardiac death, hospitalization for heart failure and aortic valve replacement necessitated by the development of AS-related symptoms.

Result

Kaplan-Meier analyses showed that LG-AS exhibited an intermediate outcome between HG-AS and Mod-AS (event-free survival at 2 years = 20.9% for HG-AS, 59.7% for LG-AS, 89.9% for Mod-AS, p < 0.001, figure A). When LG-AS was stratified by RES, the survival curves showed a significant separation (event-free survival at 2 years = 35.3% for high RES, 70.7% for low RES, p < 0.001, figure B). This trend persisted even when analysed separately for norml (stroke volume index > 35ml/m²) and low (stroke volume index ≤ 35ml/m²) flow state ((normal flow) event-free survival at 2 years = 38.7% for high RES, 70.4% for low RES, p = 0.023, figure C; (low flow) event-free survival at 2 years = 26.7% for high RES, 74.6% for low RES, p < 0.001, figure D).

Conclusion

This study confirmed the clinical efficacy of RES for risk-stratifying LG-AS patients.

Abstract P289 Figure.

Effect of attention bias modification on event-related potentials in patients with irritable bowel syndrome: A preliminary brain function and psycho-behavioral study

BACKGROUND

Attention bias modification normalizes electroencephalographic abnormalities in alpha and beta power percentages related to attention in patients with irritable bowel syndrome (IBS). Yet, it is unknown whether ABM contributes to the normalization of event-related potentials (ERP) in these patients. We hypothesized that ERP related to attention deficit would be normalized after ABM implementation in individuals with IBS.

METHODS

Thirteen patients with IBS and 10 control subjects completed a 2-month intervention that included five ABM sessions. Each session included 128 trials, resulting in a total of 640 trials during the study period. Event-related potentials were measured at the first and fifth sessions. As per the international 10-20 system for electroencephalographic electrode placement, right parietal P4 was evaluated to measure the attention component of facial expression processing.

KEY RESULTS

A group comparison of P100 latency at P4 revealed that latencies were significantly different between groups in session 1 (IBS vs control, 108 ± 8 vs 97 ± 14; t = -2.51, P = .0203). This difference was absent in session 5 (94 ± 11 vs 93 ± 11, respectively; t = -0.397, P = .6954, r = .09), indicating an effect of ABM in the IBS group.

CONCLUSIONS AND INFERENCES

Attention bias modification may have clinical utility for normalizing brain function and specifically attentional abnormalities in patients with IBS.

Intracoronary electrocardiogram for early detection of myocardial viability during coronary angioplasty in acute myocardial infarction

PURPOSE

The clinical value of the intracoronary electrocardiogram (ECG) for detecting myocardial viability in acute myocardial infarction was evaluated by thallium-201 scintigraphy and left ventriculogram at the chronic stage.

METHODS

Intracoronary ECGs, recorded from the tip of a guidewire during emergency coronary angioplasty, were obtained in 65 patients with reperfused anterior myocardial infarction. Further ST segment elevation of greater than 0.2 mV detected during the balloon inflation was taken as significant. The left ventricular segmental shortening was measured from left ventriculograms recorded at acute and chronic stages. The infarct area was defined as viable when a thallium uptake of more than 50% was detected on thallium-201 myocardial scintigraphy at the chronic stage.

RESULTS

During emergency coronary angioplasty, significant ST segment elevation was noted in 45 patients (Group A); however, the ST segment was not significantly elevated in the other 20 patients (Group B). The infarct area of 42 patients in Group A and three patients in Group B was viable on scintigraphy. Improvement left ventricular wall motion of the infarct area was observed in 39 of the 42 patients in Group A and the three patients in Group B. Therefore, intracoronary ECG can predict reversible dysfunction with excellent sensitivity (92.9%) and specificity (73.9%).

CONCLUSIONS

The myocardium within an infarct area can be regarded as viable when a further ST segment elevation occurs on intracoronary ECG during emergency coronary angioplasty. It is useful, therefore, to monitor the intracoronary ECG during coronary angioplasty balloon inflation to assess the myocardial viability of the infarct area.

Saliva level of free 3-methoxy-4-hydroxyphenylglycol (MHPG) as a biological index of anxiety disorders

To access the saliva level of free 3-methoxy-4-hydroxyphenylglycol (MHPG) as a biological index of anxiety disorders, gender- or age-dependent changes in saliva MHPG level in patients with anxiety disorders were investigated. Saliva MHPG levels in 196 normal volunteers (59 male, 137 female) and 42 outpatients with anxiety disorders (20 male, 22 female) at the initial consultation to the hospital were measured by gas chromatography-mass spectrometry. Saliva MHPG levels in patients were higher than those in normal subjects. The increase in saliva MHPG levels in male patients was greater than that in female patients. Age-associated increase in the saliva MHPG level was greater in patients than in normal subjects. Especially, a significant interaction of age vs. patient effect was found in female subjects (P=0.0005), but not in male subjects (P=0. 174). These data indicate that the measurement of saliva MHPG would be valuable for detecting pathological anxiety in male patients regardless of age and in older female patients, but not in younger female patients.

[Evaluation of coronary flow velocity and coronary flow reserve before and after coronary angioplasty using transthoracic Doppler echocardiography and Doppler guide wire]

Evaluation of left anterior descending coronary (LAD) blood flow before and after coronary angioplasty was carried out non-invasively by ultrasonic Doppler echocardiography with a newly developed digital, high-frequency, high-resolution transthoracic ultrasonic Doppler flowmeter and a 7.5 MHz probe. The results were compared with those obtained using an intracoronary Doppler guide wire. Sixteen patients, 12 males and 4 females (mean age 57 +/- 14 years) with old myocardial infarction (8 patients) and angina pectoris (8 patients) were studied. Coronary flow reserve was compared following intravenous administration of adenosine triphosphate in 12 patients. The LAD blood flow was detected in 15 of 16 patients. There was a significant increase in the diastolic peak velocity from 22.2 +/- 10.6 to 29.4 +/- 14.6 cm/sec (mean +/- SD) and the coronary flow reserve from 1.8 +/- 0.3 to 2.8 +/- 0.6 (mean +/- SD). There was a good correlation between the data obtained using transthoracic flow measurement and intracoronary flow measurement (r = 0.61, p < 0.05). LAD blood flow can be easily detected parasternally using a digital, high frequency, high-resolution ultrasonic Doppler flowmeter. This method may be applicable for judging the efficacy of coronary angioplasty by measuring coronary flow reserve and for observing the clinical course of the patient non-invasively.

[Saliva level of free-3-methoxy-4-hydroxyphenylglycol in patients with anxiety disorders]

The saliva level of free 3-methoxy-4-hydroxyphenylglycol (MHPG) in 16 outpatients with anxiety disorders was determined by gas chromatography-mass spectrometry. The anxiety level of the patients was scored by the Hamilton Anxiety Scale (HAS) before and after a one-week drug treatment with alprazolam. The saliva level of free-MHPG at the subjects' first visit to the hospital was significantly higher than that of the control subjects and was reduced by the one-week alprazolam treatment. The post-treatment level of MHPG in patients who showed good response to the treatment was significantly less than the pretreatment level, but there was no significant difference between pre- and post-treatment levels of MHPG in poor responders. There was no correlation between the MHPG level and the HAS score at the first hospital visit. The MHPG levels after the treatment correlated with the HAS scores. The reduction of the anxiety level as scored by the HAS correlated with the reduction of the MHPG level. These results indicate that the free saliva MHPG level may be a useful indicator for assessing not only the level of anxiety but also the response to drug treatment for anxiety in these patients.

Clinical utility of negative contrast intravascular ultrasound to evaluate plaque morphology before and after coronary interventions

Although intravascular ultrasound (IVUS) is used for evaluation of plaque volume and lumen size as well as detection of vessel wall structures after catheter-based interventions, differentiation between the lumen and plaque structures can be difficult. This study attempted to evaluate the efficacy of negative contrast IVUS imaging for assessment of vessel wall morphology after coronary interventions. IVUS studies were performed in 67 lesions in 66 patients before and after coronary interventions. After the baseline ultrasound imaging run, warm 5% glucose solution was injected manually through the guiding catheter into the coronary artery to washout blood from the lumen to avoid speckled reflections from red blood cells (negative contrast). Quantitative measurements were obtained and plaque morphology was assessed for the presence and extent of medial dissections and intimal flaps. There was no difference in each quantitative parameter between baseline images and negative contrast images. The vessel wall boundary was clearly delineated from the lumen, which was defined as effective negative contrast in 51 of 67 lesions (76%). The baseline images revealed plaque dissection in 9 lesions (18%) and an intimal flap in 13 lesions (25%). In addition, 4 dissections (8%) and 16 intimal flaps (31%) were visualized during the infusion of negative contrast. Additional treatment was performed in 4 lesions (8%) based on the images with negative contrast. Negative contrast IVUS was more sensitive in demonstrating a plaque fracture than were baseline images. This method is useful for enhancing the diagnostic capability of IVUS imaging and may influence the decision-making process during interventional procedures.

Saliva level of free 3-methoxy-4-hydroxyphenylglycol in psychiatric outpatients with anxiety

As a measurement of the level of anxiety in psychiatric outpatients with anxiety, we determined the saliva level of free 3-methoxy-4-hydroxyphenyleglycol (MHPG) using gas chromatography- mass spectrometry and scored the levels of anxiety with the Hamilton Anxiety Scale (HAS) in patients, before and after drug treatment with alprazolam for 1 week. The saliva level of free-MHPG at first visit to hospital was significantly higher than that of control individuals and disease control individuals and was reduced by alprazolam treatment for 1 week. There was no correlation between MHPG level and the HAS score at the first hospital visit. The MHPG levels after treatment correlated with the HAS scores. The reduction of the anxiety level as scored by the HAS correlated with the reduction of MHPG level. These results indicate that the free saliva MHPG level may be a useful indicator for assessing not only the level of anxiety, but also the response to drug treatment for anxiety in these patients.

Assessment of plaque rupture by intravascular ultrasound

To investigate the mechanism of plaque rupture, we carried out morphologic and quantitative assessments of ruptured plaque with a 30-MHz intravascular ultrasound catheter before coronary intervention. Plaque rupture, defined as an echolucent intraplaque area communicating with arterial lumen, was noted in all of 22 patients with coronary artery disease examined in the study. The plaque was eccentric in 16 patients (73%) and in 10 of these 16 patients there was a tear at the margin of the plaque (63%). Calcification was noted in only 6 patients (22%). Plaque area ranged from 8.7 to 44.1 mm2. The ruptured area ranged widely, from 1.0 to 14.0 mm2, and the percent ruptured area also ranged widely, from 3.9% to 50.9%. Fibrous cap was recognized in 20 patients. The thickness of the fibrous cap ranged from 0.1 to 0.6 mm (mean, 0.36 mm) and thinning of the fibrous cap was noted in 16 patients (80%). From these results, we conclude that intravascular ultrasound can provide detailed observations of plaque rupture consistent with pathologic studies. Accordingly, it is a suitable method for evaluation of plaque rupture in vivo. Plaque weakness, due to the presence of atheromatous core, thin fibrous cap, and no calcification, seems to be the major determinant of plaque rupture. In addition, the great variation in cap thickness and ruptured area suggests the role of other triggers, such as mechanical and hemodynamic stress, in plaque rupture.

Plaque rupture as a cause of apparent coronary aneurysm formation following directional coronary atherectomy

Late coronary aneurysm formation was observed following treatment by directional coronary atherectomy. Intravascular ultrasound disclosed that the mechanism involved was plaque rupture. The cause of coronary aneurysm may be multifactorial after coronary interventions. Intravascular ultrasound imaging is useful for understanding the pathologic mechanism of coronary aneurysm production.