Unsupervised Cluster Analysis of Patients With Aortic Stenosis Reveals Distinct Population With Different Phenotypes and Outcomes

Machine Learning Prediction for Prognosis of Patients With Aortic Stenosis

Background

Aortic valve stenosis of any degree is associated with poor outcomes.

Objectives

The authors aimed to develop a risk prediction model for aortic stenosis (AS) prognosis using machine learning techniques.

Methods

A prognostic algorithm was developed using an AS registry of 10,407 patients undergoing echocardiography between 2008 and 2020. Clinical, echocardiographic, laboratory, and medication data were used to train and test a time-to-event model, the random survival forest (RSF), for AS patient's prognosis. The composite outcome included aortic valve replacement or mortality. The SHapley Additive exPlanations method attributed the importance of variables and provided personalized risk assessment. The algorithm was validated in 2 external cohorts of 11,738 and 954 patients with AS.

Results

The median follow-up of the primary cohort was 48 (21-87) months. In this period, 1,116 patients underwent aortic valve replacement, and 5,069 patients died. RSF had an area under the curve (AUC) of 0.83 (95% CI: 0.80-0.86) and 0.83 (95% CI: 0.81-0.84) for outcomes prediction at 1 and 5 years, respectively. Using a cut-off of 50%, the RSF sensitivity and specificity for the composite outcome, were 0.80 and 0.73, respectively. Validation performance in the 2 external cohorts was similar, with AUCs of 0.73 (95% CI: 0.72-0.74) and 0.74 (95% CI: 0.72-0.76), respectively. AS severity, age, serum albumin, pulmonary artery pressure, and chronic kidney disease emerged as the top significant variables in the model.

Conclusions

In patients with AS, a machine learning algorithm predicts outcomes with good accuracy, and prognostic characteristics were identified. The model can potentially guide risk factor modification and clinical decisions to improve patient prognosis.

Novel Classification of Cardiovascular Disease Subtypes Reveals Associations Between Mortality and Polyunsaturated Fatty Acids: Insights from the United Kingdom Biobank Study

Background

Traditional association studies of cardiovascular disease (CVD) categorizations and polyunsaturated fatty acids (PUFAs) yielded conflicting findings. We propose a novel classification system based on fundamental characteristics of cardiovascular patients, such as age, body mass index, waist-hip ratio, to more accurately assess the impact of PUFAs (plasma measures) such as omega (ω)-3 (n-3) and ω-6 on mortality in cardiovascular patients.

Methods

Principal component analysis and k-means clustering were used to determine the CVD subtype. Variables included age, body mass index, waist-hip ratio, diastolic blood pressure, systolic blood pressure, total cholesterol, total triglycerides, high-density lipoprotein-cholesterol, apolipoprotein B:apolipoprotein A1, glycated hemoglobin, creatinine, albumin, C-reactive protein, white blood cell count, platelet count, and hemoglobin concentration. The association of PUFAs with all-cause, cardiovascular, and ischemic heart disease (IHD) mortality in patients with CVD was prospectively evaluated using restricted cubic splines and Cox proportional risk models.

Results

Among the 35,096 participants, 3,786 fatalities occurred. Three distinct CVD subtypes were identified, with cluster 3 characterized by older age, male gender, and low high-density lipoprotein-cholesterol, having the highest risk of mortality. Clusters 2 and 3 had the highest DHA and ω-6/ω-3 ratios, respectively, compared with Cluster 1. The protective effects of total PUFAs, ω-3, and DHA were mainly reflected in all-cause mortality and were more significant in clusters 2 and 3. Furthermore, the ω-6/ω-3 ratio of the highest quartile increased risk of all-cause [Q3: hazard ratio (HR): 1.14, 95% confidence interval [CI]: 1.00, 1.29; Q4: HR: 1.41, 95% CI: 1.24, 1.61], CVD (Q4: HR: 1.36, 95% CI: 1.07, 1.75), and IHD mortality (Q4: HR: 1.17, 95% CI: 1.12, 2.03) in cluster 3 compared with the first quartile.

Conclusions

Our findings highlight the heterogeneity of associations observed for the same type of PUFAs across distinct clusters. This association may be elucidated by the intricate interplay of various factors, encompassing inflammation, lipid metabolism, and cardiovascular health.

Definition and Validation of Prognostic Phenotypes in Moderate Aortic Stenosis

BACKGROUND

Adverse outcomes from moderate aortic stenosis (AS) may be caused by progression to severe AS or by the effects of comorbidities. In the absence of randomized trial evidence favoring aortic valve replacement (AVR) in patients with moderate AS, phenotyping patients according to risk may assist decision making.

OBJECTIVES

This study sought to identify and validate clusters of moderate AS that may be used to guide patient management.

METHODS

Unsupervised clustering algorithms were applied to demographics, comorbidities, and echocardiographic parameters in a training data set in patients with moderate AS (n = 2,469). External validation was obtained by assigning the defined clusters to an independent group with moderate AS (n = 1,358). The primary outcome, a composite of cardiac death, heart failure hospitalization, or aortic valve (AV) intervention after 5 years, was assessed between clusters in both data sets.

RESULTS

Four distinct clusters-cardiovascular (CV)-comorbid, low-flow, calcified AV, and low-risk-with significant outcomes (log-rank P < 0.0001 in both data sets) were identified and replicated. The highest risk was in the CV-comorbid cluster (validation HR: 2.00 [95% CI: 1.54-2.59]; P < 0.001). The effect of AVR on cardiac death differed among the clusters. There was a significantly lower rate of outcomes after AVR in the calcified AV cluster (validation HR: 0.21 [95% CI: 0.08-0.57]; P = 0.002), but no significant effect on outcomes in the other 3 clusters. These analyses were limited by the low rate of AVR.

CONCLUSIONS

Moderate AS has several phenotypes, and multiple comorbidities are the key drivers of adverse outcomes in patients with moderate AS. Outcomes of patients with noncalcified moderate AS were not altered by AVR in these groups. Careful attention to subgroups of moderate AS may be important to define treatable risk.

Predictors of Disease Progression and Adverse Clinical Outcomes in Patients With Moderate Aortic Stenosis Using an Artificial Intelligence-Based Software Platform

Patients with moderate aortic stenosis (AS) have a greater risk of adverse clinical outcomes than that of the general population. How this risk compares with those with severe AS, along with factors associated with outcomes and disease progression, is less clear. We analyzed serial echoes (from 2017 to 2019) from a single healthcare system using Tempus Next (Chicago, Illinois) software. AS severity was defined according to American Heart Association/American College of Cardiology guidelines. Outcomes of interest included death or heart failure hospitalization. We used Cox proportional hazards models and logistic regression to identify predictors of clinical outcome and disease progression, respectively. From 82,805 echoes for 61,546 patients, 1,770; 914; 565; and 1,463 patients had no, mild, moderate, or severe AS, respectively. Both patients with moderate and those with severe AS experienced a similar prevalence of adverse clinical outcomes (p = 0.45) that was significantly greater than that of patients without AS (p <0.01). In patients with moderate AS, atrial fibrillation (hazard ratio 3.29, 95% confidence interval 1.79 to 6.02, p <0.001) and end-stage renal disease (hazard ratio 3.34, 95% confidence interval 1.87 to 5.95, p <0.001) were associated with adverse clinical outcomes. One-third of patients with moderate AS with a subsequent echo (139/434) progressed to severe AS within 1 year. In conclusion, patients with moderate AS can progress rapidly to severe AS and experience a similar risk of adverse clinical outcomes; predictors include atrial fibrillation and low left ventricular ejection fraction. Machine learning algorithms may help identify these patients. Whether these patients may warrant earlier intervention merits further study.

Deciphering Popliteal Artery Aneurysm Patient Diversity: Insights From a Cluster Analysis of the POPART Registry

BACKGROUND

Popliteal artery aneurysms (PAAs) are the most common peripheral aneurysm. However, due to its rarity, the cumulative body of evidence regarding patient patterns, treatment strategies, and perioperative outcomes is limited. This analysis aims to investigate distinct phenotypical patient profiles and associated treatment and outcomes in patients with a PAA by performing an unsupervised clustering analysis of the POPART (Practice of Popliteal Artery Aneurysm Repair and Therapy) registry.

METHODS AND RESULTS

A cluster analysis (using k-means clustering) was performed on data obtained from the multicenter POPART registry (42 centers from Germany and Luxembourg). Sensitivity analyses were conducted to explore validity and stability. Using 2 clusters, patients were primarily separated by the absence or presence of clinical symptoms. Within the cluster of symptomatic patients, the main difference between patients with acute limb ischemia presentation and nonemergency symptomatic patients was PAA diameter. When using 6 clusters, patients were primarily grouped by comorbidities, with patients with acute limb ischemia forming a separate cluster. Despite markedly different risk profiles, perioperative complication rates appeared to be positively associated with the proportion of emergency patients. However, clusters with a higher proportion of patients having any symptoms before treatment experienced a lower rate of perioperative complications.

CONCLUSIONS

The conducted analyses revealed both an insight to the public health reality of PAA care as well as patients with PAA at elevated risk for adverse outcomes. This analysis suggests that the preoperative clinic is a far more crucial adjunct to the patient's preoperative risk assessment than the patient's epidemiological profile by itself.

The future of valvular heart disease assessment and therapy

Valvular heart disease (VHD) is becoming more prevalent in an ageing population, leading to challenges in diagnosis and management. This two-part Series offers a comprehensive review of changing concepts in VHD, covering diagnosis, intervention timing, novel management strategies, and the current state of research. The first paper highlights the remarkable progress made in imaging and transcatheter techniques, effectively addressing the treatment paradox wherein populations at the highest risk of VHD often receive the least treatment. These advances have attracted the attention of clinicians, researchers, engineers, device manufacturers, and investors, leading to the exploration and proposal of treatment approaches grounded in pathophysiology and multidisciplinary strategies for VHD management. This Series paper focuses on innovations involving computational, pharmacological, and bioengineering approaches that are transforming the diagnosis and management of patients with VHD. Artificial intelligence and digital methods are enhancing screening, diagnosis, and planning procedures, and the integration of imaging and clinical data is improving the classification of VHD severity. The emergence of artificial intelligence techniques, including so-called digital twins-eg, computer-generated replicas of the heart-is aiding the development of new strategies for enhanced risk stratification, prognostication, and individualised therapeutic targeting. Various new molecular targets and novel pharmacological strategies are being developed, including multiomics-ie, analytical methods used to integrate complex biological big data to find novel pathways to halt the progression of VHD. In addition, efforts have been undertaken to engineer heart valve tissue and provide a living valve conduit capable of growth and biological integration. Overall, these advances emphasise the importance of early detection, personalised management, and cutting-edge interventions to optimise outcomes amid the evolving landscape of VHD. Although several challenges must be overcome, these breakthroughs represent opportunities to advance patient-centred investigations.

Unsupervised cluster analysis reveals different phenotypes in patients after transcatheter aortic valve replacement

Aims

The aim of this study was to identify phenotypes with potential prognostic significance in aortic stenosis (AS) patients after transcatheter aortic valve replacement (TAVR) through a clustering approach.

Methods and results

This multi-centre retrospective study included 1365 patients with severe AS who underwent TAVR between January 2015 and March 2019. Among demographics, laboratory, and echocardiography parameters, 20 variables were selected through dimension reduction and used for unsupervised clustering. Phenotypes and outcomes were compared between clusters. Patients were randomly divided into a derivation cohort (n = 1092: 80%) and a validation cohort (n = 273: 20%). Three clusters with markedly different features were identified. Cluster 1 was associated predominantly with elderly age, a high aortic valve gradient, and left ventricular (LV) hypertrophy; Cluster 2 consisted of preserved LV ejection fraction, larger aortic valve area, and high blood pressure; and Cluster 3 demonstrated tachycardia and low flow/low gradient AS. Adverse outcomes differed significantly among clusters during a median of 2.2 years of follow-up (P < 0.001). After adjustment for clinical and echocardiographic data in a Cox proportional hazards model, Cluster 3 (hazard ratio, 4.18; 95% confidence interval, 1.76-9.94; P = 0.001) was associated with increased risk of adverse outcomes. In sequential Cox models, a model based on clinical data and echocardiographic variables (χ2 = 18.4) was improved by Cluster 3 (χ2 = 31.5; P = 0.001) in the validation cohort.

Conclusion

Unsupervised cluster analysis of patients after TAVR revealed three different groups for assessment of prognosis. This provides a new perspective in the categorization of patients after TAVR that considers comorbidities and extravalvular cardiac dysfunction.

Data-driven mortality risk prediction of severe degenerative mitral regurgitation patients undergoing mitral valve surgery

AIMS

The outcomes of mitral valve replacement/repair (MVR) in severe degenerative mitral regurgitation (MR) patients depend on various risk factors. We aimed to develop a risk prediction model for post-MVR mortality in severe degenerative MR patients using machine learning.

METHODS AND RESULTS

Consecutive severe degenerative MR patients undergoing MVR were analysed (n = 1521; 70% training/30% test sets). A random survival forest (RSF) model was constructed, with 3-year post-MVR all-cause mortality as the outcome. Partial dependency plots were used to define the thresholds of each risk factor. A simple scoring system (MVR-score) was developed to stratify post-MVR mortality risk. At 3 years following MVR, 90 patients (5.9%) died in the entire cohort (59 and 31 deaths in the training and test sets). The most important predictors of mortality in order of importance were age, haemoglobin, valve replacement, glomerular filtration rate, left atrial dimension, and left ventricular (LV) end-systolic diameter. The final RSF model with these six variables demonstrated high predictive performance in the test set (3-year C-index 0.880, 95% confidence interval 0.834-0.925), with mortality risk increased strongly with left atrial dimension >55 mm, and LV end-systolic diameter >45 mm. MVR-score demonstrated effective risk stratification and had significantly higher predictability compared to the modified Mitral Regurgitation International Database score (3-year C-index 0.803 vs. 0.750, P = 0.034).

CONCLUSION

A data-driven machine learning model provided accurate post-MVR mortality prediction in severe degenerative MR patients. The outcome following MVR in severe degenerative MR patients is governed by both clinical and echocardiographic factors.

Unsupervised learning to characterize patients with known coronary artery disease undergoing myocardial perfusion imaging

PURPOSE

Patients with known coronary artery disease (CAD) comprise a heterogenous population with varied clinical and imaging characteristics. Unsupervised machine learning can identify new risk phenotypes in an unbiased fashion. We use cluster analysis to risk-stratify patients with known CAD undergoing single-photon emission computed tomography (SPECT) myocardial perfusion imaging (MPI).

METHODS

From 37,298 patients in the REFINE SPECT registry, we identified 9221 patients with known coronary artery disease. Unsupervised machine learning was performed using clinical (23), acquisition (17), and image analysis (24) parameters from 4774 patients (internal cohort) and validated with 4447 patients (external cohort). Risk stratification for all-cause mortality was compared to stress total perfusion deficit (< 5%, 5-10%, ≥10%).

RESULTS

Three clusters were identified, with patients in Cluster 3 having a higher body mass index, more diabetes mellitus and hypertension, and less likely to be male, have dyslipidemia, or undergo exercise stress imaging (p < 0.001 for all). In the external cohort, during median follow-up of 2.6 [0.14, 3.3] years, all-cause mortality occurred in 312 patients (7%). Cluster analysis provided better risk stratification for all-cause mortality (Cluster 3: hazard ratio (HR) 5.9, 95% confidence interval (CI) 4.0, 8.6, p < 0.001; Cluster 2: HR 3.3, 95% CI 2.5, 4.5, p < 0.001; Cluster 1, reference) compared to stress total perfusion deficit (≥10%: HR 1.9, 95% CI 1.5, 2.5 p < 0.001; < 5%: reference).

CONCLUSIONS

Our unsupervised cluster analysis in patients with known CAD undergoing SPECT MPI identified three distinct phenotypic clusters and predicted all-cause mortality better than ischemia alone.

The Emerging Role of Artificial Intelligence in Valvular Heart Disease

Valvular heart disease (VHD) is a morbid condition in which timely identification and evidence-based treatments can lead to improved outcomes. Artificial intelligence broadly refers to the ability for computers to perform tasks and problem solve like the human mind. Studies applying AI to VHD have used a variety of structured (eg, sociodemographic, clinical) and unstructured (eg, electrocardiogram, phonocardiogram, and echocardiograms) and machine learning modeling approaches. Additional researches in diverse populations, including prospective clinical trials, are needed to evaluate the effectiveness and value of AI-enabled medical technologies in clinical care for patients with VHD.

Aortic Stenosis Phenotypes and Precision Transcatheter Aortic Valve Implantation

Patients with a clinical indication for aortic valve replacement can either undergo surgical aortic valve replacement (SAVR) or Transcatheter Aortic Valve Implantation (TAVI). There are many different factors that go into determining which type of replacement to undergo, including age, life expectancy, comorbidities, frailty, and patient preference. While both options offer significant benefits to patients in terms of clinical outcomes and quality of life, there is growing interest in expanding the indications for TAVI due to its minimally invasive approach. However, it is worth noting that there are several discrepancies in TAVI outcomes in regards to various endpoints, including death, stroke, and major cardiovascular events. It is unclear why these discrepancies exist, but potential explanations include the diversity of etiologies for aortic stenosis, complex patient comorbidities, and ongoing advancements in both medical therapies and devices. Of these possibilities, we propose that phenotypic variation of aortic stenosis has the most significant impact on post-TAVI clinical outcomes. Such variability in phenotypes is often due to a complex interplay between underlying comorbidities and environmental and inherent patient risk factors. However, there is growing evidence to suggest that patient genetics may also play a role in aortic stenosis pathology. As such, we propose that the selection and management of TAVI patients should emphasize a precision medicine approach.

Machine learning prediction of progressive subclinical myocardial dysfunction in moderate aortic stenosis

Background

Moderate severity aortic stenosis (AS) is poorly understood, is associated with subclinical myocardial dysfunction, and can lead to adverse outcome rates that are comparable to severe AS. Factors associated with progressive myocardial dysfunction in moderate AS are not well described. Artificial neural networks (ANNs) can identify patterns, inform clinical risk, and identify features of importance in clinical datasets.

Methods

We conducted ANN analyses on longitudinal echocardiographic data collected from 66 individuals with moderate AS who underwent serial echocardiography at our institution. Image phenotyping involved left ventricular global longitudinal strain (GLS) and valve stenosis severity (including energetics) analysis. ANNs were constructed using two multilayer perceptron models. The first model was developed to predict change in GLS from baseline echocardiography alone and the second to predict change in GLS using data from baseline and serial echocardiography. ANNs used a single hidden layer architecture and a 70%:30% training/testing split.

Results

Over a median follow-up interval of 1.3 years, change in GLS (≤ or >median change) could be predicted with accuracy rates of 95% in training and 93% in testing using ANN with inputs from baseline echocardiogram data alone (AUC: 0.997). The four most important predictive baseline features (reported as normalized % importance relative to most important feature) were peak gradient (100%), energy loss (93%), GLS (80%), and DI < 0.25 (50%). When a further model was run including inputs from both baseline and serial echocardiography (AUC 0.844), the top four features of importance were change in dimensionless index between index and follow-up studies (100%), baseline peak gradient (79%), baseline energy loss (72%), and baseline GLS (63%).

Conclusions

Artificial neural networks can predict progressive subclinical myocardial dysfunction with high accuracy in moderate AS and identify features of importance. Key features associated with classifying progression in subclinical myocardial dysfunction included peak gradient, dimensionless index, GLS, and hydraulic load (energy loss), suggesting that these features should be closely evaluated and monitored in AS.

A streamlined, machine learning-derived approach to risk-stratification in heart failure patients with secondary tricuspid regurgitation

AIMS

Secondary tricuspid regurgitation (sTR) is the most frequent valvular heart disease and has a significant impact on mortality. A high burden of comorbidities often worsens the already dismal prognosis of sTR, while tricuspid interventions remain underused and initiated too late. The aim was to examine the most powerful predictors of all-cause mortality in moderate and severe sTR using machine learning techniques and to provide a streamlined approach to risk-stratification using readily available clinical, echocardiographic and laboratory parameters.

METHODS AND RESULTS

This large-scale, long-term observational study included 3359 moderate and 1509 severe sTR patients encompassing the entire heart failure spectrum (preserved, mid-range and reduced ejection fraction). A random survival forest was applied to investigate the most important predictors and group patients according to their number of adverse features.The identified predictors and thresholds, that were associated with significantly worse mortality were lower glomerular filtration rate (<60 mL/min/1.73m2), higher NT-proBNP, increased high sensitivity C-reactive protein, serum albumin < 40 g/L and hemoglobin < 13 g/dL. Additionally, grouping patients according to the number of adverse features yielded important prognostic information, as patients with 4 or 5 adverse features had a fourfold risk increase in moderate sTR [4.81(3.56-6.50) HR 95%CI, P < 0.001] and fivefold risk increase in severe sTR [5.33 (3.28-8.66) HR 95%CI, P < 0.001].

CONCLUSION

This study presents a streamlined, machine learning-derived and internally validated approach to risk-stratification in patients with moderate and severe sTR, that adds important prognostic information to aid clinical-decision-making.

Long-term outcomes in distinct phenogroups of patients with primary mitral regurgitation undergoing valve surgery

OBJECTIVES

Patients with mitral regurgitation (MR) may be heterogeneous with different risk profiles. We aimed to identify distinct phenogroups of patients with severe primary MR and investigate their long-term prognosis after mitral valve (MV) surgery.

METHODS

The retrospective cohort of patients with severe primary MR undergoing MV surgery (derivation, n=1629; validation, n=692) was analysed. Latent class analysis was used to classify patients into subgroups using 15 variables. The primary outcome was all-cause mortality after MV surgery.

RESULTS

During follow-up (median 6.0 years), 149 patients (9.1%) died in the derivation cohort. In the univariable Cox analysis, age, female, atrial fibrillation, left ventricular (LV) end-systolic dimension/volumes, LV ejection fraction, left atrial dimension and tricuspid regurgitation peak velocity were significant predictors of mortality following MV surgery. Five distinct phenogroups were identified, three younger groups (group 1-3) and two older groups (group 4-5): group 1, least comorbidities; group 2, men with LV enlargement; group 3, predominantly women with rheumatic MR; group 4, low-risk older patients; and group 5, high-risk older patients. Cumulative survival was the lowest in group 5, followed by groups 3 and 4 (5-year survival for groups 1-5: 98.5%, 96.0%, 91.7%, 95.6% and 83.4%; p<0.001). Phenogroups had similar predictive performance compared with the Mitral Regurgitation International Database score in patients with degenerative MR (3-year C-index, 0.763 vs 0.750, p=0.602). These findings were reproduced in the validation cohort.

CONCLUSION

Five phenogroups of patients with severe primary MR with different risk profiles and outcomes were identified. This phenogrouping strategy may improve risk stratification when optimising the timing and type of interventions for severe MR.

Unsupervised cluster analysis reveals distinct subgroups in healthy population with different exercise responses of cardiorespiratory fitness

Background

Considerable attention has been paid to interindividual differences in the cardiorespiratory fitness (CRF) response to exercise. However, the complex multifactorial nature of CRF response variability poses a significant challenge to our understanding of this issue. We aimed to explore whether unsupervised clustering can take advantage of large amounts of clinical data and identify latent subgroups with different CRF exercise responses within a healthy population.

Methods

252 healthy participants (99 men, 153 women; 36.8 ± 13.4 yr) completed moderate endurance training on 3 days/week for 4 months, with exercise intensity prescribed based on anaerobic threshold (AT). Detailed clinical measures, including resting vital signs, ECG, cardiorespiratory parameters, echocardiography, heart rate variability, spirometry and laboratory data, were obtained before and after the exercise intervention. Baseline phenotypic variables that were significantly correlated with CRF exercise response were identified and subjected to selection steps, leaving 10 minimally redundant variables, including age, BMI, maximal oxygen uptake (VO_2max), maximal heart rate, VO₂ at AT as a percentage of VO_2max, minute ventilation at AT, interventricular septal thickness of end-systole, E velocity, root mean square of heart rate variability, and hematocrit. Agglomerative hierarchical clustering was performed on these variables to detect latent subgroups that may be associated with different CRF exercise responses.

Results

Unsupervised clustering revealed two mutually exclusive groups with distinct baseline phenotypes and CRF exercise responses. The two groups differed markedly in baseline characteristics, initial fitness, echocardiographic measurements, laboratory values, and heart rate variability parameters. A significant improvement in CRF following the 16-week endurance training, expressed by the absolute change in VO_2max, was observed only in one of the two groups (3.42 ± 0.4 vs 0.58 ± 0.65 ml⋅kg^-1∙min^-1, P = 0.002). Assuming a minimal clinically important difference of 3.5 ml⋅kg^-1∙min^-1 in VO_2max, the proportion of population response was 56.1% and 13.9% for group 1 and group 2, respectively (P<0.001). Although group 1 exhibited no significant improvement in CRF at group level, a significant decrease in diastolic blood pressure (70.4 ± 7.8 vs 68.7 ± 7.2 mm Hg, P = 0.027) was observed.

Conclusions

Unsupervised learning based on dense phenotypic characteristics identified meaningful subgroups within a healthy population with different CRF responses following standardized aerobic training. Our model could serve as a useful tool for clinicians to develop personalized exercise prescriptions and optimize training effects.

Artificial intelligence-enabled phenotyping of patients with severe aortic stenosis: on the recovery of extra-aortic valve cardiac damage after transcatheter aortic valve replacement

OBJECTIVE

A novel artificial intelligence-based phenotyping approach to stratify patients with severe aortic stenosis (AS) prior to transcatheter aortic valve replacement (TAVR) has been proposed, based on echocardiographic and haemodynamic data. This study aimed to analyse the recovery of extra-aortic valve cardiac damage in accordance with this novel stratification system following TAVR.

METHODS

The proposed phenotyping approach was previously established employing data from 366 patients with severe AS from a bicentric registry. For this consecutive study, echocardiographic follow-up data, obtained on day 147±75.1 after TAVR, were available from 247 patients (67.5%).

RESULTS

Correction of severe AS by TAVR significantly reduced the proportion of patients suffering from concurrent severe mitral regurgitation (from 9.29% to 3.64%, p value: 0.0015). Moreover, pulmonary artery pressures were ameliorated (estimated systolic pulmonary artery pressure: from 47.2±15.8 to 43.3±15.1 mm Hg, p value: 0.0079). However, right heart dysfunction as well as the proportion of patients with severe tricuspid regurgitation remained unchanged. Clusters with persistent right heart dysfunction ultimately displayed 2-year survival rates of 69.2% (95% CI 56.6% to 84.7%) and 74.6% (95% CI 65.9% to 84.4%), which were significantly lower compared with clusters with little or no persistent cardiopulmonary impairment (88.3% (95% CI 83.3% to 93.5%) and 85.5% (95% CI 77.1% to 94.8%)).

CONCLUSIONS

This phenotyping approach preprocedurally identifies patients with severe AS, who will not recover from extra-aortic valve cardiac damage following TAVR and whose survival is therefore significantly reduced. Importantly, not the degree of pulmonary hypertension at initial presentation, but the irreversibility of right heart dysfunction determines prognosis.

Aortic Stenosis: New Insights in Diagnosis, Treatment, and Prevention

Aortic stenosis (AS) is one of the most common valvular heart diseases and the number of patients with AS is expected to increase globally as the older population is growing fast. Since the majority of patients are elderly, AS is no longer a simple valvular heart disease of left ventricular outflow obstruction but is accompanied by other cardiac and comorbid conditions. Because of the significant variations of the disease, identifying patients at high risk and even earlier detection of patients with AS before developing symptomatic severe AS is becoming increasingly important. With the proven of efficacy and safety of transcatheter aortic valve replacement (TAVR) in the severe AS population, there is a growing interest in applying TAVR in those with less than severe AS. A medical therapy to reduce or prevent the progression in AS is actively investigated by several randomized control trials. In this review, we will summarize the most recent findings in AS and discuss potential future management strategies of patients with AS.

Artificial intelligence for the echocardiographic assessment of valvular heart disease

Developments in artificial intelligence (AI) have led to an explosion of studies exploring its application to cardiovascular medicine. Due to the need for training and expertise, one area where AI could be impactful would be in the diagnosis and management of valvular heart disease. This is because AI can be applied to the multitude of data generated from clinical assessments, imaging and biochemical testing during the care of the patient. In the area of valvular heart disease, the focus of AI has been on the echocardiographic assessment and phenotyping of patient populations to identify high-risk groups. AI can assist image acquisition, view identification for review, and segmentation of valve and cardiac structures for automated analysis. Using image recognition algorithms, aortic and mitral valve disease states have been directly detected from the images themselves. Measurements obtained during echocardiographic valvular assessment have been integrated with other clinical data to identify novel aortic valve disease subgroups and describe new predictors of aortic valve disease progression. In the future, AI could integrate echocardiographic parameters with other clinical data for precision medical management of patients with valvular heart disease.

Harnessing feature extraction capacities from a pre-trained convolutional neural network (VGG-16) for the unsupervised distinction of aortic outflow velocity profiles in patients with severe aortic stenosis

Aims

Hypothesizing that aortic outflow velocity profiles contain more valuable information about aortic valve obstruction and left ventricular contractility than can be captured by the human eye, features of the complex geometry of Doppler tracings from patients with severe aortic stenosis (AS) were extracted by a convolutional neural network (CNN).

Methods and results

After pre-training a CNN (VGG-16) on a large data set (ImageNet data set; 14 million images belonging to 1000 classes), the convolutional part was employed to transform Doppler tracings to 1D arrays. Among 366 eligible patients [age: 79.8 ± 6.77 years; 146 (39.9%) women] with pre-procedural echocardiography and right heart catheterization prior to transcatheter aortic valve replacement (TAVR), good quality Doppler tracings from 101 patients were analysed. The convolutional part of the pre-trained VGG-16 model in conjunction with principal component analysis and k-means clustering distinguished two shapes of aortic outflow velocity profiles. Kaplan-Meier analysis revealed that mortality in patients from Cluster 2 (n = 40, 39.6%) was significantly increased [hazard ratio (HR) for 2-year mortality: 3; 95% confidence interval (CI): 1-8.9]. Apart from reduced cardiac output and mean aortic valve gradient, patients from Cluster 2 were also characterized by signs of pulmonary hypertension, impaired right ventricular function, and right atrial enlargement. After training an extreme gradient boosting algorithm on these 101 patients, validation on the remaining 265 patients confirmed that patients assigned to Cluster 2 show increased mortality (HR for 2-year mortality: 2.6; 95% CI: 1.4-5.1, P-value: 0.004).

Conclusion

Transfer learning enables sophisticated pattern recognition even in clinical data sets of limited size. Importantly, it is the left ventricular compensation capacity in the face of increased afterload, and not so much the actual obstruction of the aortic valve, that determines fate after TAVR.

Subphenotyping of Patients With Aortic Stenosis by Unsupervised Agglomerative Clustering of Echocardiographic and Hemodynamic Data

OBJECTIVES

The aim of this retrospective analysis was to categorize patients with severe aortic stenosis (AS) according to clinical presentation by applying unsupervised machine learning.

BACKGROUND

Patients with severe AS present with heterogeneous clinical phenotypes, depending on disease progression and comorbidities.

METHODS

Unsupervised agglomerative clustering was applied to preprocedural data from echocardiography and right heart catheterization from 366 consecutively enrolled patients undergoing transcatheter aortic valve replacement for severe AS.

RESULTS

Cluster analysis revealed 4 distinct phenotypes. Patients in cluster 1 (n = 164 [44.8%]), serving as a reference, presented with regular cardiac function and without pulmonary hypertension (PH). Accordingly, estimated 2-year survival was 90.6% (95% CI: 85.8%-95.6%). Clusters 2 (n = 66 [18.0%]) and 4 (n = 91 [24.9%]) both comprised patients with postcapillary PH. Yet patients in cluster 2 with preserved left and right ventricular structure and function showed a similar survival as those in cluster 1 (2-year survival 85.8%; 95% CI: 76.9%-95.6%), whereas patients in cluster 4 with dilatation of all heart chambers and a high prevalence of mitral and tricuspid regurgitation (12.5% and 14.8%, respectively) died more often (2-year survival 74.9% [95% CI: 65.9%-85.2%]; HR for 2-year mortality: 2.8 [95% CI: 1.4-5.5]). Patients in cluster 3, the smallest (n = 45 [12.3%]), displayed the most extensive disease characteristics (ie, left and right heart dysfunction together with combined pre- and postcapillary PH), and 2-year survival was accordingly reduced (77.3% [95% CI: 65.2%-91.6%]; HR for 2-year mortality: 2.6 [95% CI: 1.1-6.2]).

CONCLUSIONS

Unsupervised machine learning aids in capturing complex clinical presentations as observed in patients with severe AS. Importantly, structural alterations in left and right heart morphology, possibly due to genetic predisposition, constitute an equally sensitive indicator of poor prognosis compared with high-grade PH.

Markers of Myocardial Damage Predict Mortality in Patients With Aortic Stenosis

BACKGROUND

Cardiovascular magnetic resonance (CMR) is increasingly used for risk stratification in aortic stenosis (AS). However, the relative prognostic power of CMR markers and their respective thresholds remains undefined.

OBJECTIVES

Using machine learning, the study aimed to identify prognostically important CMR markers in AS and their thresholds of mortality.

METHODS

Patients with severe AS undergoing AVR (n = 440, derivation; n = 359, validation cohort) were prospectively enrolled across 13 international sites (median 3.8 years' follow-up). CMR was performed shortly before surgical or transcatheter AVR. A random survival forest model was built using 29 variables (13 CMR) with post-AVR death as the outcome.

RESULTS

There were 52 deaths in the derivation cohort and 51 deaths in the validation cohort. The 4 most predictive CMR markers were extracellular volume fraction, late gadolinium enhancement, indexed left ventricular end-diastolic volume (LVEDVi), and right ventricular ejection fraction. Across the whole cohort and in asymptomatic patients, risk-adjusted predicted mortality increased strongly once extracellular volume fraction exceeded 27%, while late gadolinium enhancement >2% showed persistent high risk. Increased mortality was also observed with both large (LVEDVi >80 mL/m²) and small (LVEDVi ≤55 mL/m²) ventricles, and with high (>80%) and low (≤50%) right ventricular ejection fraction. The predictability was improved when these 4 markers were added to clinical factors (3-year C-index: 0.778 vs 0.739). The prognostic thresholds and risk stratification by CMR variables were reproduced in the validation cohort.

CONCLUSIONS

Machine learning identified myocardial fibrosis and biventricular remodeling markers as the top predictors of survival in AS and highlighted their nonlinear association with mortality. These markers may have potential in optimizing the decision of AVR.

Clinical Phenotype Classification of Atrial Fibrillation Patients Using Cluster Analysis and Associations with Trial-Adjudicated Outcomes

BACKGROUND AND PURPOSE

Given the great clinical heterogeneity of atrial fibrillation (AF) patients, conventional classification only based on disease subtype or arrhythmia patterns may not adequately characterize this population. We aimed to identify different groups of AF patients who shared common clinical phenotypes using cluster analysis and evaluate the association between identified clusters and clinical outcomes.

METHODS

We performed a hierarchical cluster analysis in AF patients from AMADEUS and BOREALIS trials. The primary outcome was a composite of stroke/thromboembolism (TE), cardiovascular (CV) death, myocardial infarction, and/or all-cause death. Individual components of the primary outcome and major bleeding were also assessed.

RESULTS

We included 3980 AF patients treated with the Vitamin-K Antagonist from the AMADEUS and BOREALIS studies. The analysis identified four clusters in which patients varied significantly among clinical characteristics. Cluster 1 was characterized by patients with low rates of CV risk factors and comorbidities; Cluster 2 was characterized by patients with a high burden of CV risk factors; Cluster 3 consisted of patients with a high burden of CV comorbidities; Cluster 4 was characterized by the highest rates of non-CV comorbidities. After a mean follow-up of 365 (standard deviation 187) days, Cluster 4 had the highest cumulative risk of outcomes. Compared with Cluster 1, Cluster 4 was independently associated with an increased risk for the composite outcome (hazard ratio (HR) 2.43, 95% confidence interval (CI) 1.70-3.46), all-cause death (HR 2.35, 95% CI 1.58-3.49) and major bleeding (HR 2.18, 95% CI 1.19-3.96).

CONCLUSIONS

Cluster analysis identified four different clinically relevant phenotypes of AF patients that had unique clinical characteristics and different outcomes. Cluster analysis highlights the high degree of heterogeneity in patients with AF, suggesting the need for a phenotype-driven approach to comorbidities, which could provide a more holistic approach to management aimed to improve patients' outcomes.

Artificial Intelligence and Echocardiography

Artificial intelligence (AI) is evolving in the field of diagnostic medical imaging, including echocardiography. Although the dynamic nature of echocardiography presents challenges beyond those of static images from X-ray, computed tomography, magnetic resonance, and radioisotope imaging, AI has influenced all steps of echocardiography, from image acquisition to automatic measurement and interpretation. Considering that echocardiography often is affected by inter-observer variability and shows a strong dependence on the level of experience, AI could be extremely advantageous in minimizing observer variation and providing reproducible measures, enabling accurate diagnosis. Currently, most reported AI applications in echocardiographic measurement have focused on improved image acquisition and automation of repetitive and tedious tasks; however, the role of AI applications should not be limited to conventional processes. Rather, AI could provide clinically important insights from subtle and non-specific data, such as changes in myocardial texture in patients with myocardial disease. Recent initiatives to develop large echocardiographic databases can facilitate development of AI applications. The ultimate goal of applying AI to echocardiography is automation of the entire process of echocardiogram analysis. Once automatic analysis becomes reliable, workflows in clinical echocardiographic will change radically. The human expert will remain the master controlling the overall diagnostic process, will not be replaced by AI, and will obtain significant support from AI systems to guide acquisition, perform measurements, and integrate and compare data on request.

Artificial Intelligence and Transcatheter Interventions for Structural Heart Disease: A glance at the (near) future

With innovations in therapeutic technologies and changes in population demographics, transcatheter interventions for structural heart disease have become the preferred treatment and will keep growing. Yet, a thorough clinical selection and efficient pathway from diagnosis to treatment and follow-up are mandatory. In this review we reflect on how artificial intelligence may help to improve patient selection, pre-procedural planning, procedure execution and follow-up so to establish efficient and high quality health care in an increasing number of patients.

The Role of Artificial Intelligence in Cardiovascular Imaging: State of the Art Review

In this current digital landscape, artificial intelligence (AI) has established itself as a powerful tool in the commercial industry and is an evolving technology in healthcare. Cutting-edge imaging modalities outputting multi-dimensional data are becoming increasingly complex. In this era of data explosion, the field of cardiovascular imaging is undergoing a paradigm shift toward machine learning (ML) driven platforms. These diverse algorithms can seamlessly analyze information and automate a range of tasks. In this review article, we explore the role of ML in the field of cardiovascular imaging.