1
|
Namasivayam M, Meredith T, Muller DWM, Roy DA, Roy AK, Kovacic JC, Hayward CS, Feneley MP. Machine learning prediction of progressive subclinical myocardial dysfunction in moderate aortic stenosis. Front Cardiovasc Med 2023; 10:1153814. [PMID: 37324638 PMCID: PMC10266266 DOI: 10.3389/fcvm.2023.1153814] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [Abstract] [Key Words] [Grants] [Track Full Text] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 01/30/2023] [Accepted: 05/15/2023] [Indexed: 06/17/2023] Open
Abstract
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.
Collapse
Affiliation(s)
- Mayooran Namasivayam
- Department of Cardiology, St Vincent’s Hospital, Sydney, NSW, Australia
- Faculty of Medicine and Health, University of New South Wales, Sydney, NSW, Australia
- Heart Valve Disease and Artificial Intelligence Laboratory, Victor Chang Cardiac Research Institute, Sydney, NSW, Australia
| | - Thomas Meredith
- Department of Cardiology, St Vincent’s Hospital, Sydney, NSW, Australia
- Faculty of Medicine and Health, University of New South Wales, Sydney, NSW, Australia
- Heart Valve Disease and Artificial Intelligence Laboratory, Victor Chang Cardiac Research Institute, Sydney, NSW, Australia
| | - David W. M. Muller
- Department of Cardiology, St Vincent’s Hospital, Sydney, NSW, Australia
- Faculty of Medicine and Health, University of New South Wales, Sydney, NSW, Australia
| | - David A. Roy
- Department of Cardiology, St Vincent’s Hospital, Sydney, NSW, Australia
- Faculty of Medicine and Health, University of New South Wales, Sydney, NSW, Australia
| | - Andrew K. Roy
- Department of Cardiology, St Vincent’s Hospital, Sydney, NSW, Australia
| | - Jason C. Kovacic
- Department of Cardiology, St Vincent’s Hospital, Sydney, NSW, Australia
- Faculty of Medicine and Health, University of New South Wales, Sydney, NSW, Australia
- Vascular Biology Laboratory, Victor Chang Cardiac Research Institute, Sydney, NSW, Australia
- Icahn School of Medicine at Mount Sinai, Cardiovascular Research Institute, New York, NY, United States
| | - Christopher S. Hayward
- Department of Cardiology, St Vincent’s Hospital, Sydney, NSW, Australia
- Faculty of Medicine and Health, University of New South Wales, Sydney, NSW, Australia
- Cardiac Mechanics Laboratory, Victor Chang Cardiac Research Institute, Sydney, NSW, Australia
| | - Michael P. Feneley
- Department of Cardiology, St Vincent’s Hospital, Sydney, NSW, Australia
- Faculty of Medicine and Health, University of New South Wales, Sydney, NSW, Australia
- Cardiac Mechanics Laboratory, Victor Chang Cardiac Research Institute, Sydney, NSW, Australia
| |
Collapse
|
2
|
Namasivayam M, Myers PD, Guttag JV, Capoulade R, Pibarot P, Picard MH, Hung J, Stultz CM. Predicting outcomes in patients with aortic stenosis using machine learning: the Aortic Stenosis Risk (ASteRisk) score. Open Heart 2022; 9:openhrt-2022-001990. [PMID: 35641101 PMCID: PMC9157386 DOI: 10.1136/openhrt-2022-001990] [Citation(s) in RCA: 1] [Impact Index Per Article: 0.5] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Submit a Manuscript] [Subscribe] [Scholar Register] [Received: 02/11/2022] [Accepted: 05/04/2022] [Indexed: 12/11/2022] Open
Abstract
Objective To use echocardiographic and clinical features to develop an explainable clinical risk prediction model in patients with aortic stenosis (AS), including those with low-gradient AS (LGAS), using machine learning (ML). Methods In 1130 patients with moderate or severe AS, we used bootstrap lasso regression (BLR), an ML method, to identify echocardiographic and clinical features important for predicting the combined outcome of all-cause mortality or aortic valve replacement (AVR) within 5 years after the initial echocardiogram. A separate hold out set, from a different centre (n=540), was used to test the generality of the model. We also evaluated model performance with respect to each outcome separately and in different subgroups, including patients with LGAS. Results Out of 69 available variables, 26 features were identified as predictive by BLR and expert knowledge was used to further reduce this set to 9 easily available and input features without loss of efficacy. A ridge logistic regression model constructed using these features had an area under the receiver operating characteristic curve (AUC) of 0.74 for the combined outcome of mortality/AVR. The model reliably identified patients at high risk of death in years 2–5 (HRs ≥2.0, upper vs other quartiles, for years 2–5, p<0.05, p=not significant in year 1) and was also predictive in the cohort with LGAS (n=383, HRs≥3.3, p<0.05). The model performed similarly well in the independent hold out set (AUC 0.78, HR ≥2.5 in years 1–5, p<0.05). Conclusion In two separate longitudinal databases, ML identified prognostic features and produced an algorithm that predicts outcome for up to 5 years of follow-up in patients with AS, including patients with LGAS. Our algorithm, the Aortic Stenosis Risk (ASteRisk) score, is available online for public use.
Collapse
Affiliation(s)
- Mayooran Namasivayam
- Division of Cardiology, Massachusetts General Hospital, Harvard Medical School, Boston, Massachusetts, USA
| | - Paul D Myers
- Department of Electrical Engineering and Computer Science, Massachusetts Institute of Technology, Cambridge, Massachusetts, USA
| | - John V Guttag
- Computer Science and Artificial Intelligence Laboratory, Massachusetts Institute of Technology, Cambridge, Massachusetts, USA
| | - Romain Capoulade
- l'institut du thorax, CHU Nantes, CNRS, INSERM, University of Nantes, Nantes, France
| | - Philippe Pibarot
- Cardiology, Quebec Heart and Lung Institute, Laval University, Quebec City, Quebec, Canada
| | - Michael H Picard
- Division of Cardiology, Massachusetts General Hospital, Harvard Medical School, Boston, Massachusetts, USA
| | - Judy Hung
- Division of Cardiology, Massachusetts General Hospital, Harvard Medical School, Boston, Massachusetts, USA
| | - Collin M Stultz
- Division of Cardiology, Massachusetts General Hospital, Harvard Medical School, Boston, Massachusetts, USA.,Department of Electrical Engineering and Computer Science, Massachusetts Institute of Technology, Cambridge, Massachusetts, USA.,Computer Science and Artificial Intelligence Laboratory, Massachusetts Institute of Technology, Cambridge, Massachusetts, USA.,Research Laboratory of Electronics, Massachusetts Institute of Technology, Cambridge, Massachusetts, USA
| |
Collapse
|
3
|
Excess Mortality Associated with Progression Rate in Asymptomatic Aortic Valve Stenosis. J Am Soc Echocardiogr 2020; 34:237-244. [PMID: 33253813 DOI: 10.1016/j.echo.2020.11.015] [Citation(s) in RCA: 13] [Impact Index Per Article: 3.3] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Journal Information] [Submit a Manuscript] [Subscribe] [Scholar Register] [Received: 04/08/2020] [Revised: 11/23/2020] [Accepted: 11/23/2020] [Indexed: 12/11/2022]
Abstract
BACKGROUND Aortic valve stenosis (AS) is a progressive condition characterized by gradual calcification of the aortic cusps. Progression rate evaluated using echocardiography has been associated with survival. However, data from routine practice covering the whole spectrum of AS severity and the rate of symptom onset are sparse. The aim of this study was to assess outcomes under medical management related to disease progression in asymptomatic patients with a wide range of AS severity. METHODS Two hundred twenty-nine consecutive asymptomatic patients (mean age, 77 ± 10 years; 55% men) with AS, preserved left ventricular ejection fraction, and two or more echocardiographic examinations performed from 2004 to 2014 were retrospectively included. The median time between the two echocardiographic examinations was 24 months (interquartile range, 15-46 months). Patients were identified as rapid progressors if the annualized difference in peak aortic velocity between two echocardiographic examinations was ≥0.3 m/sec/y; others were labeled as slow progressors. The primary end point was mortality during medical follow-up (censoring on aortic valve interventions). The secondary end point was overall mortality. RESULTS Rapid progressors accounted for 67 of the 229 patients (29%), and this feature was not associated with baseline characteristics. During a median of 5.8 years (interquartile range, 3.4-8.3 years) of follow-up from the first echocardiographic examination, 102 patients (45%) died, 86 (84%) during medical follow-up. Rapid progression rate predicted excess mortality (vs slow progression rate) after adjustment for age, sex, symptoms, baseline left ventricular ejection fraction, and baseline aortic valve area (hazard ratio, 2.50; 95% CI, 1.48-4.21; P = .0006) and after adjusting for peak aortic velocity and left ventricular ejection fraction obtained at the last echocardiographic examination (hazard ratio, 2.07; 95% CI, 1.25-3.46; P = .005). Among patients with baseline peak aortic velocity < 4 m/sec (nonsevere AS), rapid progression rate was associated with higher 5-year mortality compared with slow progression (57% vs 22% [P < .0001] under medical management and 44% vs 18% [P = .005] overall). Outcomes were comparable between nonsevere AS rapid progressors and baseline severe AS. Progression rate showed incremental prognostic value on receiver operating characteristic curve analysis versus AS severity. Of note, among slow progressors, 11 patients (5%) presented with high rates of symptom development and poor outcomes related to ventricular dysfunction or other advanced AS features. CONCLUSIONS Progression rate is an individual, almost unpredictable feature among patients with AS. Rapid progression is an incremental marker of excess mortality in asymptomatic patients with AS, independent of clinical and hemodynamic characteristics. Rapid progression rate may identify patients with nonsevere AS at higher risk for events.
Collapse
|
5
|
Impact of Arterial Blood Pressure on Ultrasound Hemodynamic Assessment of Aortic Valve Stenosis Severity. J Am Soc Echocardiogr 2020; 33:1324-1333. [PMID: 32868157 DOI: 10.1016/j.echo.2020.06.013] [Citation(s) in RCA: 6] [Impact Index Per Article: 1.5] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Journal Information] [Submit a Manuscript] [Subscribe] [Scholar Register] [Received: 09/12/2019] [Revised: 05/05/2020] [Accepted: 06/14/2020] [Indexed: 01/13/2023]
Abstract
BACKGROUND Aortic stenosis (AS) severity assessment is based on several indices. Aortic valve area (AVA) is subject to inaccuracies inherent to the measurement method, while velocities and gradients depend on hemodynamic status. There is controversy as to whether blood pressure directly affects common indices of AS severity. OBJECTIVES The study objective was to assess the effect of systolic blood pressure (SBP) variation on AS indices, in a clinical setting. METHODS A prospective, single-center study included 100 patients with at least moderately severe AS with preserved left ventricle ejection fraction. Patients underwent ultrasound examination during which AS severity indices were collected, with three hemodynamic conditions: (1) low SBP: <120 mm Hg; (2) intermediate SBP: between 120 and 150 mm Hg; (3) high SBP: ≥150 mm Hg. For each patient, SBP profiles were obtained by injection of isosorbide dinitrate or phenylephrine. RESULTS At baseline state, 59% presented a mean gradient (Gmean) ≥ 40 mm Hg, 44% a peak aortic jet velocity (Vpeak) ≥4 m/sec, 66% a dimensionless index (DI) ≤0.25, and 87% an indexed AVA (AVAi) ≤ 0.6 cm2/m2. Compared with intermediate and low SBP, high SBP induced a significant decrease in Gmean (39 ± 12 vs 43 ± 12 and 47 ± 12 mm Hg, respectively; P < .05) and in Vpeak (3.8 ± 0.6 vs 4.0 ± 0.6 and 4.2 ± 0.6 mm Hg; P < .05). Compared with the baseline measures, in 16% of patients with an initial Gmean< 40 mm Hg, gradient rose above 40 mm Hg after optimization of the afterload (low SBP; P < .05). Conversely, DI and AVAi did not vary with changes in hemodynamic conditions. Flow rate, not stroke volume was found to impact Gmean and Vpeak but not AVA and DI (P < .05). CONCLUSIONS Hemodynamic conditions may affect the AS ultrasound assessment. High SBP, or afterload, leads to an underestimation of AS severity when based on gradients and velocities. Systolic blood pressure monitoring and control are crucial during AS ultrasound assessment.
Collapse
|