A Network-Based "Phenomics" Approach for Discovering Patient Subtypes From High-Throughput Cardiac Imaging Data

Differences in the Impact of Left Ventricular Outflow Tract Obstruction on Intraventricular Pressure Gradient in Feline Hypertrophic Cardiomyopathy

Hypertrophic cardiomyopathy (HCM) is a common form of cardiomyopathy in cats, and heart failure occurs as diastolic dysfunction progresses. HCM in cats is broadly classified as non-obstructive and obstructive hypertrophic cardiomyopathy, depending on the presence or absence of outflow tract obstruction. Measurement of the intraventricular pressure differences (IVPD) using color M-mode (CMM) has attracted attention as a reliable diastolic index as it correlates with catheterization, the gold standard for the assessment of diastolic performance. Because IVPD is affected by the size of the heart, the intraventricular pressure gradient (IVPG) index, which is unaffected by heart size, is by calculated by dividing IVPD by LV length. In the present study, CMM IVPG was used to non-invasively assess diastolic impairment in cats with obstructive hypertrophic cardiomyopathy. This study was conducted on 10 control cats and 18 cats in the HCM group. Although no severe left atrial enlargement was observed in the HCM group, the basal IVPG was significantly increased in the HOCM group compared to the control group. Although IVPD typically suggests impaired diastolic function and reduced ventricular compliance, the significant increase observed in the HOCM group compared to controls may suggest an indirect elevation in left atrial pressure, likely secondary to left ventricular outflow tract obstruction. The increase in IVPG in HOCM, as shown in this study, is a pathological effect of left ventricular outflow tract obstruction that cannot be detected by conventional echocardiographic indices, and evaluating IVPG is useful to evaluate cardiac function from a perspective that differs from conventional methods.

Cardiac ultrasomics for acute myocardial infarction risk stratification and prediction of all-cause mortality: a feasibility study

BACKGROUND

Current risk stratification tools for acute myocardial infarction (AMI) have limitations, particularly in predicting mortality. This study utilizes cardiac ultrasound radiomics (i.e., ultrasomics) to risk stratify AMI patients when predicting all-cause mortality.

RESULTS

The study included 197 patients: (a) retrospective internal cohort (n = 155) of non-ST-elevation myocardial infarction (n = 63) and ST-elevation myocardial infarction (n = 92) patients, and (b) external cohort from the multicenter Door-To-Unload in ST-segment-elevation myocardial infarction [DTU-STEMI] Pilot Trial (n = 42). Echocardiography images of apical 2, 3, and 4-chamber were processed through an automated deep-learning pipeline to extract ultrasomic features. Unsupervised machine learning (topological data analysis) generated AMI clusters followed by a supervised classifier to generate individual predicted probabilities. Validation included assessing the incremental value of predicted probabilities over the Global Registry of Acute Coronary Events (GRACE) risk score 2.0 to predict 1-year all-cause mortality in the internal cohort and infarct size in the external cohort. Three phenogroups were identified: Cluster A (high-risk), Cluster B (intermediate-risk), and Cluster C (low-risk). Cluster A patients had decreased LV ejection fraction (P < 0.01) and global longitudinal strain (P = 0.03) and increased mortality at 1-year (log rank P = 0.05). Ultrasomics features alone (C-Index: 0.74 vs. 0.70, P = 0.04) and combined with global longitudinal strain (C-Index: 0.81 vs. 0.70, P < 0.01) increased prediction of mortality beyond the GRACE 2.0 score. In the DTU-STEMI clinical trial, Cluster A was associated with larger infarct size (> 10% LV mass, P < 0.01), compared to remaining clusters.

CONCLUSIONS

Ultrasomics-based phenogroup clustering, augmented by TDA and supervised machine learning, provides a novel approach for AMI risk stratification.

Broadening Perspectives of Artificial Intelligence in Echocardiography

Echocardiography frequently serves as the first-line treatment of diagnostic imaging for several pathological entities in cardiology. Artificial intelligence (AI) has been growing substantially in information technology and various commercial industries. Machine learning (ML), a branch of AI, has been shown to expand the capabilities and potential of echocardiography. ML algorithms expand the field of echocardiography by automated assessment of the ejection fraction and left ventricular function, integrating novel approaches such as speckle tracking or tissue Doppler echocardiography or vector flow mapping, improved phenotyping, distinguishing between cardiac conditions, and incorporating information from mobile health and genomics. In this review article, we assess the impact of AI and ML in echocardiography.

Distinguishing heart failure with reduced ejection fraction from heart failure with preserved ejection fraction: A phenomics approach

AIM

Pathophysiological differences between patients with heart failure with preserved (HFpEF) and reduced (HFrEF) ejection fraction (EF) remain unclear. Therefore we used a phenomics approach, integrating selected proteomics data with patient characteristics and cardiac structural and functional parameters, to get insight into differential pathophysiological mechanisms and identify potential treatment targets.

METHODS AND RESULTS

We report data from a representative subcohort of the prospective Singapore Heart Failure Outcomes and Phenotypes (SHOP), including patients with HFrEF (EF <40%, n = 217), HFpEF (EF ≥50%, n = 213), and age- and sex-matched controls without HF (n = 216). We measured 92 biomarkers using a proximity extension assay and assessed cardiac structure and function in all participants using echocardiography. We used multi-block projection to latent structure analysis to integrate clinical, echocardiographic, and biomarker variables. Candidate biomarker targets were cross-referenced with small-molecule and drug databases. The total cohort had a median age of 65 years (interquartile range 60-71), and 50% were women. Protein profiles strongly discriminated patients with HFrEF (area under the curve [AUC] = 0.89) and HFpEF (AUC = 0.94) from controls. Phenomics analyses identified unique druggable inflammatory markers in HFpEF from the tumour necrosis factor receptor superfamily (TNFRSF), which were positively associated with hypertension, diabetes, and increased posterior and relative wall thickness. In HFrEF, interleukin (IL)-8 and IL-6 were possible targets related to lower EF and worsening renal function.

CONCLUSION

We identified pathophysiological mechanisms related to increased cardiac wall thickness parameters and potentially druggable inflammatory markers from the TNFRSF in HFpEF.

Serum apolipoprotein B to apolipoprotein A-I ratio predicts mortality in patients with heart failure

AIMS

Apolipoproteins have been reported to be involved in many cardiovascular diseases. The aim of our study was to investigate the prognostic value of apolipoprotein B (ApoB) to apolipoprotein A-I (ApoA-I) ratio (ApoB/ApoA-I) in patients with heart failure (HF).

METHODS AND RESULTS

We randomly assigned 2400 HF patients into the training cohort (n = 1400) and the validation cohort (n = 1000). Using a receiver operating characteristic curve, we identified the optimal cut-off value of the ApoB/ApoA-I in the training cohort as 0.69, which was further validated in the validation cohort. A propensity score matching (PSM) analysis was conducted to eliminate the imbalance in the baseline characteristics of the high and low ApoB/ApoA-I group. A total of 2242 HF patients were generated in the PSM cohort. We also validated our results with an independent cohort (n = 838). Univariate and multivariate analyses were conducted to explore the independent prognostic value of ApoB/ApoA-I in the training cohort (n = 1400), the validation cohort (n = 1000), the PSM cohort (n = 2242), and the independent cohort (n = 838). Patients with high ApoB/ApoA-I ratio had significantly poorer prognosis compared with those with low ApoB/ApoA-I ratio in the training cohort, the validation cohort, the PSM cohort, and the independent cohort (P < 0.05). Multivariate analysis indicated that the ApoB/ApoA-I was an independent prognostic factor for HF in the training cohort [hazard ratio (HR) = 1.637, 95% confidence interval (CI) = 1.201-2.231, P = 0.002], the validation cohort (HR = 1.54, 95% CI = 1.051-2.257, P = 0.027), the PSM cohort (HR = 1.645, 95% CI = 1.273-2.125, P < 0.001), and the independent cohort (HR = 1.987, 95% CI = 1.251-3.155, P = 0.004).

CONCLUSIONS

Serum ApoB/ApoA-I ratio is an independent predictor for the prognosis of HF patients.

Assessment of anthracycline-induced cardiotoxicity in childhood cancer survivors during long-term follow-up using strain analysis and intraventricular pressure gradient measurements

BACKGROUND

Cardiac dysfunction due to cardiotoxicity from anthracycline chemotherapy is a leading cause of morbidity and mortality in childhood cancer survivors (CCS), and the cumulative incidence of cardiac events has continued to increase. This study identifies an adequate indicator of cardiac dysfunction during long-term follow-up.

PROCEDURE

In total, 116 patients (median age: 15.5 [range: 4.7-40.2] years) with childhood cancer who were treated with anthracycline were divided into three age groups for analysis (C1: 4-12 years of age, C2: 13-18 years of age, C3: 19-40 years of age), and 116 control patients of similar ages were divided into three corresponding groups (N1, N2, and N3). Layer-specific strains were assessed for longitudinal strain (LS) and circumferential strain (CS). The total and segmental intraventricular pressure gradients (IVPG) were also calculated based on Doppler imaging of the mitral inflow using Euler's equation.

RESULTS

Conventional echocardiographic parameters were not significantly different between the patients and controls. All layers of the LS and inner and middle layers of the basal and papillary CS in all ages and all IVPGs in C2 and C3 decreased compared to those of corresponding age groups. Interestingly, basal CS and basal IVPG in CCS showed moderate correlation and both tended to rapidly decrease with aging. Furthermore, basal IVPG and anthracycline dose showed significant correlations.

CONCLUSIONS

Basal CS and total and basal IVPGs may be particularly useful indicators of cardiotoxicity in long-term follow-up.

Clinical phenotypes among patients with normal cardiac perfusion using unsupervised learning: a retrospective observational study

BACKGROUND

Myocardial perfusion imaging (MPI) is one of the most common cardiac scans and is used for diagnosis of coronary artery disease and assessment of cardiovascular risk. However, the large majority of MPI patients have normal results. We evaluated whether unsupervised machine learning could identify unique phenotypes among patients with normal scans and whether those phenotypes were associated with risk of death or myocardial infarction.

METHODS

Patients from a large international multicenter MPI registry (10 sites) with normal perfusion by expert visual interpretation were included in this cohort analysis. The training population included 9849 patients, and external testing population 12,528 patients. Unsupervised cluster analysis was performed, with separate training and external testing cohorts, to identify clusters, with four distinct phenotypes. We evaluated the clinical and imaging features of clusters and their associations with death or myocardial infarction.

FINDINGS

Patients in Clusters 1 and 2 almost exclusively underwent exercise stress, while patients in Clusters 3 and 4 mostly required pharmacologic stress. In external testing, the risk for Cluster 4 patients (20.2% of population, unadjusted hazard ratio [HR] 6.17, 95% confidence interval [CI] 4.64-8.20) was higher than the risk associated with pharmacologic stress (HR 3.03, 95% CI 2.53-3.63), or previous myocardial infarction (HR 1.82, 95% CI 1.40-2.36).

INTERPRETATION

Unsupervised learning identified four distinct phenotypes of patients with normal perfusion scans, with a significant proportion of patients at very high risk of myocardial infarction or death. Our results suggest a potential role for patient phenotyping to improve risk stratification of patients with normal imaging results.

FUNDING

This work was supported by the National Heart, Lung, and Blood Institute at the National Institutes of Health [R35HL161195 to PS]. The REFINE SPECT database was supported by the National Heart, Lung, and Blood Institute at the National Institutes of Health [R01HL089765 to PS]. MCW was supported by the British Heart Foundation [FS/ICRF/20/26002].

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.

Phenomics and Robust Multiomics Data for Cardiovascular Disease Subtyping

The complex landscape of cardiovascular diseases encompasses a wide range of related pathologies arising from diverse molecular mechanisms and exhibiting heterogeneous phenotypes. This variety of manifestations poses significant challenges in the development of treatment strategies. The increasing availability of precise phenotypic and multiomics data of cardiovascular disease patient populations has spurred the development of a variety of computational disease subtyping techniques to identify distinct subgroups with unique underlying pathogeneses. In this review, we outline the essential components of computational approaches to select, integrate, and cluster omics and clinical data in the context of cardiovascular disease research. We delve into the challenges faced during different stages of the analysis, including feature selection and extraction, data integration, and clustering algorithms. Next, we highlight representative applications of subtyping pipelines in heart failure and coronary artery disease. Finally, we discuss the current challenges and future directions in the development of robust subtyping approaches that can be implemented in clinical workflows, ultimately contributing to the ongoing evolution of precision medicine in health care.

Normative change with gestation in fetal intraventricular pressure difference with color M-mode Doppler echocardiography

AIM

The intraventricular pressure difference (IVPD) is the pressure difference in early diastole from the base to the apex of the ventricle. It is a useful marker for evaluating diastolic function because of its role as a suction force. This study investigated the changes in total and segmental IVPDs in normal fetuses throughout gestation to obtain normative data equations.

METHODS

One hundred thirty-seven healthy pregnant women at 12-40 weeks of gestation were prospectively enrolled to evaluate IVPD. The color M mode was performed, and the image was evaluated using our own code to calculate the IVPD. Segmental IVPD was divided into mid to apex and base. Pearson's correlation coefficient was used to evaluate this relationship.

RESULTS

There was a significant, positive relationship between IVPD and gestational age in both ventricles (right ventricle [RV]: r = 0.800, left ventricle [LV]: r = 0.818). As for segmental IVPD, basal and mid-apical IVPD also increased with gestation in both ventricles (RV: basal, r = 0.627; mid-apical, r = 0.705; LV: basal r = 0.758; mid-apical, r = 0.756). IVPG, which was calculated as IVPD/ventricular length, also showed a weak, positive relationship with gestation in both ventricles (RV r = 0.351, p < 0.001; LV r = 0.373, p < 0.001).

CONCLUSION

The total and segmental IVPDs significantly increased linearly through time.

A prediction model for major adverse cardiovascular events in patients with heart failure based on high-throughput echocardiographic data

Background

Heart failure (HF) is a serious end-stage condition of various heart diseases with increasing frequency. Few studies have combined clinical features with high-throughput echocardiographic data to assess the risk of major cardiovascular events (MACE) in patients with heart failure. In this study, we assessed the relationship between these factors and heart failure to develop a practical and accurate prognostic dynamic nomogram model to identify high-risk groups of heart failure and ultimately provide tailored treatment options.

Materials and methods

We conducted a prospective study of 468 patients with heart failure and established a clinical predictive model. Modeling to predict risk of MACE in heart failure patients within 6 months after discharge obtained 320 features including general clinical data, laboratory examination, 2-dimensional and Doppler measurements, left ventricular (LV) and left atrial (LA) speckle tracking echocardiography (STE), and left ventricular vector flow mapping (VFM) data, were obtained by building a model to predict the risk of MACE within 6 months of discharge for patients with heart failure. In addition, the addition of machine learning models also confirmed the necessity of increasing the STE and VFM parameters.

Results

Through regular follow-up 6 months after discharge, MACE occurred in 156 patients (33.3%). The prediction model showed good discrimination C-statistic value, 0.876 (p < 0.05), which indicated good identical calibration and clinical efficacy. In multiple datasets, through machine learning multi-model comparison, we found that the area under curve (AUC) of the model with VFM and STE parameters was higher, which was more significant with the XGboost model.

Conclusion

In this study, we developed a prediction model and nomogram to estimate the risk of MACE within 6 months of discharge among patients with heart failure. The results of this study can provide a reference for clinical physicians for detection of the risk of MACE in terms of clinical characteristics, cardiac structure and function, hemodynamics, and enable its prompt management, which is a convenient, practical and effective clinical decision-making tool for providing accurate prognosis.

Network Analysis of Cardiac Remodeling by Primary Mitral Regurgitation Emphasizes the Role of Diastolic Function

BACKGROUND

Topological data analysis (TDA) can generate patient-patient similarity networks by analyzing large, complex data and derive new insights that may not be possible with standard statistics.

OBJECTIVES

The purpose of this paper was to discover novel phenotypes of chronic primary mitral regurgitation (MR) patients and to analyze their clinical implications using network analysis of echocardiographic data.

METHODS

Patients with chronic moderate to severe primary MR were prospectively enrolled from 11 Asian tertiary hospitals (n = 850; mean age 56.9 ± 14.2 years, 57.9% men). We performed TDA to generate network models using 14 demographic and echocardiographic variables. The patients were grouped by phenotypes in the network, and the prognosis was compared by groups.

RESULTS

The network model by TDA revealed 3 distinct phenogroups. Group A was the youngest with fewer comorbidities but increased left ventricular (LV) end-systolic volume, representing compensatory LV dilation commonly seen in chronic primary MR. Group B was the oldest with high blood pressure and a predominant diastolic dysfunction but relatively preserved LV size, an unnoticed phenotype in chronic primary MR. Group C showed advanced LV remodeling with impaired systolic, diastolic function, and LV dilation, indicating advanced chronic primary MR. During follow-up (median 3.5 years), 60 patients received surgery for symptomatic MR or died of cardiovascular causes. Kaplan-Meier curves demonstrated that although group C had the worst clinical outcome (P < 0.001), group B, characterized by diastolic dysfunction, had an event-free survival comparable to group A despite preserved LV chamber size. The grouping information by the network model was an independent predictor for the composite of MR surgery or cardiovascular death (adjusted HR: 1.918; 95% CI: 1.257-2.927; P = 0.003).

CONCLUSIONS

The patient-patient similarity network by TDA visualized diverse remodeling patterns in chronic primary MR and revealed distinct phenotypes not emphasized currently. Importantly, diastolic dysfunction deserves equal attention when understanding the clinical presentation of chronic primary MR.

Exploring and Identifying Prognostic Phenotypes of Patients with Heart Failure Guided by Explainable Machine Learning

Identifying patient prognostic phenotypes facilitates precision medicine. This study aimed to explore phenotypes of patients with heart failure (HF) corresponding to prognostic condition (risk of mortality) and identify the phenotype of new patients by machine learning (ML). A unsupervised ML was applied to explore phenotypes of patients in a derivation dataset (n = 562) based on their medical records. Thereafter, supervised ML models were trained on the derivation dataset to classify these identified phenotypes. Then, the trained classifiers were further validated on an independent validation dataset (n = 168). Finally, Shapley additive explanations were used to interpret decision making of phenotype classification. Three patient phenotypes corresponding to stratified mortality risk (high, low, and intermediate) were identified. Kaplan−Meier survival curves among the three phenotypes had significant difference (pairwise comparison p < 0.05). Hazard ratio of all-cause mortality between patients in phenotype 1 (n = 91; high risk) and phenotype 3 (n = 329; intermediate risk) was 2.08 (95%CI 1.29−3.37, p = 0.003), and 0.26 (95%CI 0.11−0.61, p = 0.002) between phenotype 2 (n = 142; low risk) and phenotype 3. For phenotypes classification by random forest, AUCs of phenotypes 1, 2, and 3 were 0.736 ± 0.038, 0.815 ± 0.035, and 0.721 ± 0.03, respectively, slightly better than the decision tree. Then, the classifier effectively identified the phenotypes for new patients in the validation dataset with significant difference on survival curves and hazard ratios. Finally, age and creatinine clearance rate were identified as the top two most important predictors. ML could effectively identify patient prognostic phenotypes, facilitating reasonable management and treatment considering prognostic condition.

Machine Learning Augmented Echocardiography for Diastolic Function Assessment

Cardiac diastolic dysfunction is prevalent and is a diagnostic criterion for heart failure with preserved ejection fraction—a burgeoning global health issue. As gold-standard invasive haemodynamic assessment of diastolic function is not routinely performed, clinical guidelines advise using echocardiography measures to determine the grade of diastolic function. However, the current process has suboptimal accuracy, regular indeterminate classifications and is susceptible to confounding from comorbidities. Advances in artificial intelligence in recent years have created revolutionary ways to evaluate and integrate large quantities of cardiology data. Imaging is an area of particular strength for the sub-field of machine-learning, with evidence that trained algorithms can accurately discern cardiac structures, reliably estimate chamber volumes, and output systolic function metrics from echocardiographic images. In this review, we present the emerging field of machine-learning based echocardiographic diastolic function assessment. We summarise how machine-learning has made use of diastolic parameters to accurately differentiate pathology, to identify novel phenotypes within diastolic disease, and to grade diastolic function. Perspectives are given about how these innovations could be used to augment clinical practice, whilst areas for future investigation are identified.

Advanced Cardiovascular Imaging in Clinical Heart Failure

Cardiovascular imaging is the cornerstone of the assessment of patients with heart failure. Although noninvasive volumetric estimation of the cardiac function is an essential and indisputably useful clinical tool, cardiac imaging has evolved and matured to offer detailed functional, hemodynamic, and tissue characterization. The adoption of a new framework to diagnose and phenotype heart failure that incorporates comprehensive imaging assessment has been lacking in clinical trials. The present review offers a general overview of available imaging strategies for patients with heart failure.

Automated Pattern Recognition in Whole-Cardiac Cycle Echocardiographic Data: Capturing Functional Phenotypes with Machine Learning

BACKGROUND

Echocardiography provides complex data on cardiac function that can be integrated into patterns of dysfunction related to the severity of cardiac disease. The aim of this study was to demonstrate the feasibility of applying machine learning (ML) to automate the integration of echocardiographic data from the whole cardiac cycle and to automatically recognize patterns in velocity profiles and deformation curves, allowing the identification of functional phenotypes.

METHODS

Echocardiography was performed in 189 clinically managed patients with hypertension and 97 healthy individuals without hypertension. Speckle-tracking analysis of the left ventricle and atrium was performed, and deformation curves were extracted. Aortic and mitral blood pool pulsed-wave Doppler and mitral annular tissue pulsed-wave Doppler velocity profiles were obtained. These whole-cardiac cycle deformation and velocity curves were used as ML input. Unsupervised ML was used to create a representation of patients with hypertension in a virtual space in which patients are positioned on the basis of the similarity of their integrated whole-cardiac cycle echocardiography data. Regression methods were used to explore patterns of echocardiographic traces within this virtual ML-derived space, while clustering was used to define phenogroups.

RESULTS

The algorithm captured different patterns in tissue and blood-pool velocity and deformation profiles and integrated the findings, yielding phenotypes related to normal cardiac function and others to advanced remodeling associated with pressure overload in hypertension. The addition of individuals without hypertension into the ML-derived space confirmed the interpretation of normal and remodeled phenotypes.

CONCLUSIONS

ML-based pattern recognition is feasible from echocardiographic data obtained during the whole cardiac cycle. Automated algorithms can consistently capture patterns in velocity and deformation data and, on the basis of these patterns, group patients into interpretable, clinically comprehensive phenogroups that describe structural and functional remodeling. Automated pattern recognition may potentially aid interpretation of imaging data and diagnostic accuracy.

A Machine-Learning Framework to Identify Distinct Phenotypes of Aortic Stenosis Severity

OBJECTIVES

The authors explored the development and validation of machine-learning models for augmenting the echocardiographic grading of aortic stenosis (AS) severity.

BACKGROUND

In AS, symptoms and adverse events develop secondarily to valvular obstruction and left ventricular decompensation. The current echocardiographic grading of AS severity focuses on the valve and is limited by diagnostic uncertainty.

METHODS

Using echocardiography (ECHO) measurements (ECHO cohort, n = 1,052), we performed patient similarity analysis to derive high-severity and low-severity phenogroups of AS. We subsequently developed a supervised machine-learning classifier and validated its performance with independent markers of disease severity obtained using computed tomography (CT) (CT cohort, n = 752) and cardiovascular magnetic resonance (CMR) imaging (CMR cohort, n = 160). The classifier's prognostic value was further validated using clinical outcomes (aortic valve replacement [AVR] and death) observed in the ECHO and CMR cohorts.

RESULTS

In 1,964 patients from the 3 multi-institutional cohorts, 1,346 (68%) subjects had either nonsevere or discordant AS severity. Machine learning identified 1,117 (57%) patients as having high-severity and 847 (43%) as having low-severity AS. High-severity patients in CT and CMR cohorts had higher valve calcium scores and left ventricular mass and fibrosis, respectively than the low-severity group. In the ECHO cohort, progression to AVR and progression to death in patients who did not receive AVR was faster in the high-severity group. Compared with the conventional classification of disease severity, machine-learning-based severity classification improved discrimination (integrated discrimination improvement: 0.07; 95% confidence interval: 0.02 to 0.12) and reclassification (net reclassification improvement: 0.17; 95% confidence interval: 0.11 to 0.23) for the outcome of AVR at 5 years. For both ECHO and CMR cohorts, we observed prognostic value of the machine-learning classifications for subgroups with asymptomatic, nonsevere or discordant AS.

CONCLUSIONS

Machine learning can integrate ECHO measurements to augment the classification of disease severity in most patients with AS, with major potential to optimize the timing of AVR.

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.