Automatic measurements of left ventricular volumes and ejection fraction by artificial intelligence: clinical validation in real time and large databases

AIMS

Echocardiography is a cornerstone in cardiac imaging, and left ventricular (LV) ejection fraction (EF) is a key parameter for patient management. Recent advances in artificial intelligence (AI) have enabled fully automatic measurements of LV volumes and EF both during scanning and in stored recordings. The aim of this study was to evaluate the impact of implementing AI measurements on acquisition and processing time and test-retest reproducibility compared with standard clinical workflow, as well as to study the agreement with reference in large internal and external databases.

METHODS AND RESULTS

Fully automatic measurements of LV volumes and EF by a novel AI software were compared with manual measurements in the following clinical scenarios: (i) in real time use during scanning of 50 consecutive patients, (ii) in 40 subjects with repeated echocardiographic examinations and manual measurements by 4 readers, and (iii) in large internal and external research databases of 1881 and 849 subjects, respectively. Real-time AI measurements significantly reduced the total acquisition and processing time by 77% (median 5.3 min, P < 0.001) compared with standard clinical workflow. Test-retest reproducibility of AI measurements was superior in inter-observer scenarios and non-inferior in intra-observer scenarios. AI measurements showed good agreement with reference measurements both in real time and in large research databases.

CONCLUSION

The software reduced the time taken to perform and volumetrically analyse routine echocardiograms without a decrease in accuracy compared with experts.

Algorithm for predicting valvular heart disease from heart sounds in an unselected cohort

Objective

This study aims to assess the ability of state-of-the-art machine learning algorithms to detect valvular heart disease (VHD) from digital heart sound recordings in a general population that includes asymptomatic cases and intermediate stages of disease progression.

Methods

We trained a recurrent neural network to predict murmurs from heart sound audio using annotated recordings collected with digital stethoscopes from four auscultation positions in 2,124 participants from the Tromsø7 study. The predicted murmurs were used to predict VHD as determined by echocardiography.

Results

The presence of aortic stenosis (AS) was detected with a sensitivity of 90.9%, a specificity of 94.5%, and an area under the curve (AUC) of 0.979 (CI: 0.963-0.995). At least moderate AS was detected with an AUC of 0.993 (CI: 0.989-0.997). Moderate or greater aortic and mitral regurgitation (AR and MR) were predicted with AUC values of 0.634 (CI: 0.565-703) and 0.549 (CI: 0.506-0.593), respectively, which increased to 0.766 and 0.677 when clinical variables were added as predictors. The AUC for predicting symptomatic cases was higher for AR and MR, 0.756 and 0.711, respectively. Screening jointly for symptomatic regurgitation or presence of stenosis resulted in an AUC of 0.86, with 97.7% of AS cases (n = 44) and all 12 MS cases detected.

Conclusions

The algorithm demonstrated excellent performance in detecting AS in a general cohort, surpassing observations from similar studies on selected cohorts. The detection of AR and MR based on HS audio was poor, but accuracy was considerably higher for symptomatic cases, and the inclusion of clinical variables improved the performance of the model significantly.

Novel approach to artefact detection and the definition of normal ranges of segmental strain and strain-rate values

AIMS

Strain artefacts are known to hamper the correct interpretation of segmental strain and strain-rate (S/SR). Defining the normal ranges of myocardial segmental deformation is important in clinical studies and routine echocardiographic practice. In order to define artefact-free normal ranges for segmental longitudinal S/SR parameters, we investigated the extent to which different types of artefacts and their segmental localisation in the three different myocardial layers created a bias in the results of echocardiographic strain measurements.

METHODS

The study included echocardiograms from men and women aged 40-69 years from two population-based studies, namely the Know Your Heart study (Russia) and the Tromsø Study (Norway). Of the 2207 individuals from these studies, 840 had normal results, defined as the absence of hypertension or indicators of any cardiovascular disease. Two-dimensional (2D) global and segmental S/SR of the three myocardial layers were analysed using speckle tracking echocardiography. Artefacts were assessed with two different methods: visual identification of image-artefacts and a novel conceptual approach of 'curve-artefacts' or unphysiological strain-curve formation.

RESULTS

Segmental strain values were found to have significantly reduced in the presence of strain-curve artefacts (14.9%±5.8% towards -20.7%±4.9%), and increased with the foreshortening of the 2D image. However, the individual global strain values were not substantially altered by discarding segmental artefacts. Reduction due to artefacts was observed in all segments, layers, systolic and diastolic strain, and SR. Thus, we presented normal ranges for basal-septal, basal, medial and apical segment groups after excluding artefacts.

CONCLUSION

Strain-curve artefacts introduce systematic errors, resulting in reduced segmental S/SR values. In terms of artefact-robust global longitudinal strain, the detection of curve-artefacts is crucial for the correct interpretation of segmental S/SR patterns. Intersegmental S/SR gradients and artefacts need to be considered for the correct definition of normalcy and pathology.

Longitudinal Associations Between Cumulative Physical Activity and Change in Structure and Function of the Left Side of the Heart: The Tromsø Study 2007–2016

Background

Current knowledge about the relationship between physical activity (PA) and cardiac remodeling is mainly derived from cross-sectional studies of athletes, and there is a knowledge gap of this association in the general adult and elderly population. Therefore, we aimed to explore the longitudinal association between cumulative PA and change in cardiac structure and function in a general adult and elderly population.

Methods

This longitudinal study includes 594 participants from the sixth (Tromsø6, 2007–08) and seventh (Tromsø7, 2015–16) survey of the Tromsø Study. Cardiac structure and function were assessed by echocardiography at two time points, and PA was self-reported by questionnaire at both time points. PA volume was expressed as cumulative PA (Low, Moderate, and Hard) and the association with left atrial (LA) and left ventricular (LV) structure and function was assessed using ANCOVA.

Results

Overall, LA diameter index (LADi) increased significantly more in Hard compared to Moderate PA (+0.08 cm/m², 95% CI 0.01–0.15, p = 0.020) from Tromsø6 to Tromsø7. When stratified by sex or age, higher levels of cumulative PA were associated with increased LADi in males and in participants <65 years only. Indexed LV mass (LVMi) increased significantly more in Moderate than in Low PA (+3.9 g/m^2.7, 95% CI 0.23–7.57, p = 0.037). When stratified by sex or age, these changes in LVMi and indexed LV diameter (LVDi) were only significant in females. No significant associations were observed between cumulative PA and change in relative wall thickness, E/e' ratio, e' velocity, LV ejection fraction, and LADi/LVDi ratio.

Conclusion

Higher levels of cumulative PA were associated with increased LADi in males and participants <65 years, and with increased LVMi and LVDi in females. Despite cardiac chamber enlargement, the pump function of the heart did not change with higher levels of PA, and the atrioventricular ratio was unchanged. Our results indicate that cardiac chamber enlargement is a physiological response to PA.

NT-ProBNP and high-sensitivity troponin T as screening tests for subclinical chronic heart failure in a general population

Aims

The aim of this study was to establish age‐specific and sex‐specific cut‐off values for N‐terminal pro‐brain natriuretic peptide (NT‐proBNP) and high‐sensitivity troponin T (hs‐troponin T) in healthy subjects and assess cardiac biomarkers as screening tools for subclinical heart failure (HF) in a general population.

Methods and results

Altogether, 1936 participants were randomly selected from the general population Tromsø 7 study in Northern Norway. Diagnostic accuracy (sensitivity, specificity, and negative and positive predictive value) of cardiac markers for echocardiographically defined subclinical HF was evaluated. The receiver‐operating characteristic analysis showed that areas under the curve were relatively low (under 0.75) for both NT‐proBNP and hs‐troponin T, suggesting that the diagnostic accuracy of these biomarkers for subclinical HF was not excellent, especially for mild forms of HF and younger age group 40–49 years. Sex‐specific and age‐specific cut‐offs for hs‐troponin T (99th percentiles) and NT‐proBNP (97.5th percentiles) were established in healthy subjects from the same general population. The sex‐specific and age‐specific cut‐offs for NT‐proBNP had higher specificity for subclinical HF compared with the previously established single cut‐off 125 pg/mL. Age‐specific cut‐off for hs‐troponin T (18 ng/L) for men ≥60 years had also higher specificity than the single cut‐off 14 ng/L. These cut‐offs had high specificity, but low sensitivity, that makes hs‐troponin T and NT‐proBNP good biomarkers to rule in HF in case of a positive test, but not good enough to rule out all unrecognized HF due to false negative results.

Conclusions

N‐terminal pro‐brain natriuretic peptide and hs‐troponin T are suboptimal screening tools for subclinical HF in a general population due to low sensitivity.

Associations between physical activity, left atrial size and incident atrial fibrillation: the Tromsø Study 1994-2016

AIMS

Left atrial (LA) enlargement is an independent risk factor for atrial fibrillation (AF). Interestingly, some athletes have increased risk of AF, which may be linked to LA enlargement; however, little is known about the relationship between LA enlargement and AF risk at moderate-level physical activity (PA). We aimed to explore the associations between PA, LA size and risk of incident AF, and if PA can attenuate the risk of AF with LA enlargement.

METHODS

This prospective study followed 2479 participants (52.4% female), free from known cardiac pathology, for median 20.2 years. Participants were followed up for hospital-diagnosed AF, confirmed by electrocardiography, from 1994-95 through 2016. At baseline, LA size was evaluated by anteroposterior LA diameter, and PA was self-reported by questionnaire.

RESULTS

We observed a U-shaped relationship between PA and AF, and moderately active had 32% lower AF risk than inactive (HRadjusted 0.68, 95% CI 0.50 to 0.93). Participants with LA enlargement had 38% higher AF risk compared with participants with normal LA size (HRadjusted 1.38, 95% CI 1.12 to 1.69). However, the increased AF risk with LA enlargement was attenuated by PA; compared with inactive participants with LA enlargement, the AF risk was 45% lower among active with LA enlargement (HRadjusted 0.55, 95% CI 0.39 to 0.79). AF risk in active participants with LA enlargement did not differ from active with normal LA size. These patterns were observed in both men and women, and in participants over/under 65 years.

CONCLUSION

Moderate PA was associated with reduced AF risk, and PA attenuated the increased risk of AF with LA enlargement in both men and women and all age groups.

Cross-sectional associations between accelerometry-measured physical activity, left atrial size, and indices of left ventricular diastolic dysfunction: The Tromsø Study

•

Examination of how objectively PA and LAVi associate in a general population represents novelty.

•

PA is associated with greater LA size in participants <70 years with normal diastolic function.

•

LA enlargement is only associated with diastolic dysfunction in the most inactive participants.

•

We suggest that LA enlargement in active individuals is not an expression of cardiac dysfunction.

Whereas left atrial (LA) enlargement is an independent predictor for adverse cardiovascular events and all-cause mortality, this is regarded a physiological adaption of exercise. Paradoxically, LA size in athletes may overlap the enlargement observed in patients with cardiac pathology. Current knowledge is mainly derived from studies of athletes, and little is known about cardiac adaptations to physical activity (PA) in the general population. We explored the association between objectively measured PA and LA volume index (LAVi), and between LAVi enlargement and indices of diastolic dysfunction stratified by PA-level.

Our study included 1573 participants from the population-based Tromsø Study (2015–16). PA was assessed with an ActiGraph wGT3X-BT accelerometer. Echocardiography was performed according to current guidelines. The associations between PA and LAVi, and between LAVi enlargement and indices of diastolic dysfunction were estimated by univariable and multivariable linear regression analyses, adjusted for sex, age, and cardiovascular risk factors.

Our multiple adjusted analyses showed significant linear associations between PA and LAVi in ages < 70 years, and between PA and LAVi in participants with normal diastolic function. No associations were seen in ages ≥ 70 years or for participants with abnormal diastolic function. In those 40–54 years, the most active participants had larger LAVi (4.45 mL/m², p = 0.016) than the least active. LAVi enlargement was only associated with indices of diastolic dysfunction in the most inactive participants.

In conclusion, higher levels of PA associate with greater LAVi in participants < 70 years with normal diastolic function. LAVi enlargement is only associated with diastolic dysfunction in the most inactive participants.

Prediction of chronic heart failure and chronic obstructive pulmonary disease in a general population: the Tromsø study

Aims

Heart failure (HF) and chronic obstructive pulmonary disease (COPD) are main causes of dyspnoea, and echocardiography and spirometry are essential investigations for these diagnoses. Our aim was to determine the prevalence of HF and COPD in a general population, also how the diseases may be identified, and to what extent their clinical characteristics differ.

Methods and results

In the seventh survey of Tromsø study (2015–16), subjects aged 40 years or more were examined with echocardiography, spirometry, lung sound recordings, questionnaires, including the modified Medical Research Council (mMRC) questionnaire on dyspnoea, and N‐terminal pro‐brain natriuretic peptide analysis. A diagnosis of HF (HF with reduced ejection fraction, HF with mid‐range ejection fraction, or HF with preserved ejection fraction) or COPD was established according to current guidelines. Predictors of HF and COPD were evaluated by logistic regression and receiver operating characteristic curve analysis. A total of 7110 participants could be evaluated for COPD, 1624 for HF, and 1538 for both diseases. Age‐standardized prevalence of HF was 6.8% for women and 6.1% for men; the respective figures for COPD were 5.2% and 5.1%. Among the 1538 evaluated for both diseases, 139 subjects fulfilled the HF criteria, but only 17.1% reported to have the disease. Of those fulfilling the COPD criteria, 31.6% reported to have the disease. Shortness of breath at exertion was a frequent finding in HF; 59% of those with mMRC ≥2 had HF, while such shortness of breath was found in 24% among those with COPD. Reporting mMRC ≥2 had an odds ratio for HF of 19.5 (95% confidence interval 11.3–33.7), whereas the odds ratio for COPD was 6.3 (95% confidence interval 3.5–11.6). Current smoking was the strongest predictor of COPD but did not predict HF. Basal inspiratory crackles were significant predictors of HF in multivariable analysis. Among the subtypes of HF, an age <70 years was most frequently found in HF with reduced ejection fraction, in 51.7%. Clinical scores based on the predictive value in multivariable analysis of history, symptoms, and signs predicted HF and COPD with areas under the curve of 0.833 and 0.829, respectively.

Conclusions

Study participants with HF and COPD were in most cases not aware of their condition. In general practice, when an elderly patient present with shortness of breath, both diseases should be considered. Previous cardiovascular disease points at HF, while a history of smoking points at COPD. The threshold should be low for ordering echocardiography or spirometry for verifying the suspected cause of dyspnoea.

Global myocardial longitudinal strain in a general population-associations with blood pressure and subclinical heart failure: The Tromsø Study

The early detection of subclinical myocardial dysfunction can contribute to the treatment and prevention of heart failure (HF). The aim of the study was to (i) describe myocardial global longitudinal strain (GLS) patterns in a large general population sample from Norway and their relation to established cardiovascular disease (CVD) risk factors; (ii) to determine its normal thresholds in healthy individuals and (iii) ascertain the relation of myocardial GLS to stage A subclinical heart failure (SAHF). Participants (n = 1855) of the 7th survey of the population-based Tromsø Study of Norway (2015-2016) with GLS measurements were studied. Linear and logistic regression models were used for assessment of the associations between CVD risk factors and GLS. Mean GLS (SD) in healthy participants was - 15.9 (2.7) % in men and - 17.8 (3.1) % in women. Among healthy subjects, defined as those without known cardiovascular diseases and comorbidities, GLS declined with age. An increase of systolic blood pressure (SBP) of 10 mm Hg was associated with a 0.2% GLS reduction. Myocardial GLS in individuals with SAHF was 1.2% lower than in participants without SAHF (p < 0.001). Mean myocardial GLS declines with age in both sexes, both in a general population and in the healthy subsample. SBP increase associated with GLS decline in women. Our findings indicate high sensitivity of GLS for early subclinical stages of HF.

Know Your Heart: Rationale, design and conduct of a cross-sectional study of cardiovascular structure, function and risk factors in 4500 men and women aged 35-69 years from two Russian cities, 2015-18

Russia has one of the highest rates of cardiovascular disease in the world. The International Project on Cardiovascular Disease in Russia (IPCDR) was set up to understand the reasons for this. A substantial component of this study was the Know Your Heart Study devoted to characterising the nature and causes of cardiovascular disease in Russia by conducting large cross-sectional surveys in two Russian cities Novosibirsk and Arkhangelsk. The study population was 4542 men and women aged 35-69 years recruited from the general population. Fieldwork took place between 2015-18. There were two study components: 1) a baseline interview to collect information on socio-demographic characteristics and cardiovascular risk factors, usually conducted at home, and 2) a comprehensive health check at a primary care clinic which included detailed examination of the cardiovascular system. In this paper we describe in detail the rationale for, design and conduct of these studies.

Know Your Heart: Rationale, design and conduct of a cross-sectional study of cardiovascular structure, function and risk factors in 4500 men and women aged 35-69 years from two Russian cities, 2015-18

Russia has one of the highest rates of cardiovascular disease in the world. The International Project on Cardiovascular Disease in Russia (IPCDR) was set up to understand the reasons for this. A substantial component of this study was the Know Your Heart Study devoted to characterising the nature and causes of cardiovascular disease in Russia by conducting large cross-sectional surveys in two Russian cities Novosibirsk and Arkhangelsk. The study population was 4542 men and women aged 35-69 years recruited from the general population. Fieldwork took place between 2015-18. There were two study components: 1) a baseline interview to collect information on socio-demographic characteristics and cardiovascular risk factors, usually conducted at home, and 2) a comprehensive health check at a primary care clinic which included detailed examination of the cardiovascular system. In this paper we describe in detail the rationale for, design and conduct of these studies.

Know Your Heart: Rationale, design and conduct of a cross-sectional study of cardiovascular structure, function and risk factors in 4500 men and women aged 35-69 years from two Russian cities, 2015-18

Russia has one of the highest rates of cardiovascular disease in the world. The International Project on Cardiovascular Disease in Russia (IPCDR) was set up to understand the reasons for this. A substantial component of this study was the Know Your Heart Study devoted to characterising the nature and causes of cardiovascular disease in Russia by conducting large cross-sectional surveys in two Russian cities Novosibirsk and Arkhangelsk. The study population was 4542 men and women aged 35-69 years recruited from the general population. Fieldwork took place between 2015-18. There were two study components: 1) a baseline interview to collect information on socio-demographic characteristics and cardiovascular risk factors, usually conducted at home, and 2) a comprehensive health check at a primary care clinic which included detailed examination of the cardiovascular system. In this paper we describe in detail the rationale for, design and conduct of these studies.