A machine learning approach for the prediction of pulmonary hypertension.

BACKGROUND: Machine learning (ML) is a powerful tool for identifying and structuring several informative variables for predictive tasks. Here, we investigated how ML algorithms may assist in echocardiographic pulmonary hypertension (PH) prediction, where current guidelines recommend integrating several echocardiographic parameters. METHODS: In our database of 90 patients with invasively determined pulmonary artery pressure (PAP) with corresponding echocardiographic estimations of PAP obtained within 24 hours, we trained and applied five ML algorithms (random forest of classification trees, random forest of regression trees, lasso penalized logistic regression, boosted classification trees, support vector machines) using a 10 times 3-fold cross-validation (CV) scheme. RESULTS: ML algorithms achieved high prediction accuracies: support vector machines (AUC 0.83; 95% CI 0.73-0.93), boosted classification trees (AUC 0.80; 95% CI 0.68-0.92), lasso penalized logistic regression (AUC 0.78; 95% CI 0.67-0.89), random forest of classification trees (AUC 0.85; 95% CI 0.75-0.95), random forest of regression trees (AUC 0.87; 95% CI 0.78-0.96). In contrast to the best of several conventional formulae (by Aduen et al.), this ML algorithm is based on several echocardiographic signs and feature selection, with estimated right atrial pressure (RAP) being of minor importance. CONCLUSIONS: Using ML, we were able to predict pulmonary hypertension based on a broader set of echocardiographic data with little reliance on estimated RAP compared to an existing formula with non-inferior performance. With the conceptual advantages of a broader and unbiased selection and weighting of data our ML approach is suited for high level assistance in PH prediction.

Echocardiographic Estimation of Mean Pulmonary Artery Pressure: A Comparison of Different Approaches to Assign the Likelihood of Pulmonary Hypertension.

BACKGROUND: Current guidelines advise using echocardiography for noninvasive estimation of the likelihood that a patient has pulmonary hypertension (PH). To estimate the echocardiographic probability of PH, the maximal tricuspid regurgitation velocity (TR Vmax) is recommended as the main parameter to use over more complex algorithms that provide an estimation of pulmonary artery pressure. This preference is based on concerns about inaccuracies and amplification of measurement errors that can occur from using derived variables. However, this has not been examined systematically. METHODS: A retrospective database analysis was performed of invasively determined measurements of right heart pressure in 90 patients, corresponding echocardiographic estimations of pulmonary artery pressure, and additional parameters obtained within 24 hours. Several algorithms were compared for their correlations and accuracy parameters. RESULTS: Although a Bland-Altman analysis demonstrated that all examined algorithms exhibited inaccuracies that could be clinically relevant in individuals, algorithms estimating mean pulmonary artery pressure (PAPm) on the basis of tricuspid regurgitation generally exhibited stronger correlations with invasively determined PAPm and more accurate identification of PH than did TR Vmax. Echocardiographic estimation of right atrial pressure >15 mm Hg exhibited the highest odds ratio for invasively confirmed PH, suggesting that this parameter is of additional diagnostic value. Indeed, algorithms that also considered right atrial pressure performed best, whereas empirical algorithms, TR Vmax, and methods relying on pulmonary acceleration time exhibited weaker performance. CONCLUSIONS: Although all methods are associated with inaccuracies, echocardiographically determined PAPm was superior to the current guideline recommendation of using TR Vmax with regard to its correlation with invasively determined PAPm and the presence of PH. PAPm may be considered as an alternative to TR Vmax for evaluating the echocardiographic probability of PH.

Neural respiratory drive and cardiac function in patients with obesity hypoventilation syndrome following initiation of non-invasive ventilation.

BACKGROUND: Chronic hypercapnic respiratory failure (HRF) in obesity hypoventilation syndrome (OHS) is commonly treated using non-invasive ventilation (NIV). We hypothesised that treatment of OHS would improve neural respiratory drive index (NRDI) and cardiac function. METHODS: Fourteen patients (8 females) with OHS, who were admitted for initiation of domiciliary NIV, were prospectively studied. Patients had (mean ± SD): age (53±10 years), body mass index (BMI) (50.1±10.8 kg/m2), and pCO2 (7.3±0.9 kPa). NRDI was assessed by surface electromyogram of the parasternal intercostals. Cardiac function was assessed by transthoracic echocardiography (TTE). All measurements were performed at baseline, 6 weeks, and 3 months. RESULTS: NRDI improved on day one following NIV set-up comparing to baseline (484.2±214.8 vs. 316.5±106.5 AU) and this improvement was maintained at 6 weeks (369.1±173.2 AU) and at 3 months (351.2±167.1 AU) (P=0.004). No significant differences were identified in terms of cardiac function between baseline and 3 months [tricuspid annular plane systolic excursion (TAPSE) (24.6±5.8 vs. 23.0±4.0 mm, P=0.317); systolic pulmonary artery (PA) pressures (36.7±15.2 vs. 44.5±23.9 mmHg, P=0.163]. CONCLUSIONS: NIV improves NRDI in patients with OHS, while the cardiac function over a three-month period remains unchanged.

Physiological Reduction in Left Ventricular Contractile Function in Healthy Postpartum Women: Potential Overlap with Peripartum Cardiomyopathy.

AIMS: Peripartum cardiomyopathy is a potentially life-threatening cause of heart failure, commoner in Afro-Caribbean than Caucasian women. Its diagnosis can be challenging due to physiological changes in cardiac function that also occur in healthy women during the early postpartum period. This study aimed to (i) establish the overlap between normal cardiac physiology in the immediate postpartum period and pathological changes in peripartum cardiomyopathy ii) identify any ethnicity-specific changes in cardiac function and cardiac biomarkers in healthy postpartum women. METHODS AND RESULTS: We conducted a cross-sectional study of 58 healthy postpartum women within 48 hours of delivery and 18 matched non-pregnant controls. Participants underwent cardiac assessment by echocardiography and strain analysis, including 3D echocardiography in 40 postpartum women. Results were compared with 12 retrospectively studied peripartum cardiomyopathy patients. Healthy postpartum women had significantly higher left ventricular volumes and mass, and lower ejection fraction and global longitudinal strain than non-pregnant controls. These parameters were significantly more impaired in peripartum cardiomyopathy patients but with overlapping ranges of values. Healthy postpartum women had higher levels of adrenomedullin, placental growth factor (PlGF) and soluble fms-like tyrosine kinase-1 (sFlt1) compared to controls. The postpartum state, adrenomedullin, sFlt1 and the sFlt1:PlGF ratio were independent predictors of LV remodelling and function in healthy postpartum women. CONCLUSION: Healthy postpartum women demonstrate several echocardiographic indicators of left ventricular remodelling and reduced function, which are associated with altered levels of angiogenic and cardiac biomarkers.

Pulmonary Haemodynamics in Sickle Cell Disease Are Driven Predominantly by a High-Output State Rather Than Elevated Pulmonary Vascular Resistance: A Prospective 3-Dimensional Echocardiography/Doppler Study.

AIMS: Patients with sickle cell disease have significant morbidity and mortality. Pulmonary hypertension is suggested to be an important contributor but its nature and severity in these patients and how best to non-invasively assess it are controversial. We hypothesised that a high-output state rather than primary pulmonary vascular pathology may be the major abnormality in sickle cell disease. This study aimed to evaluate the characteristics and severity of pulmonary hypertension in patients with sickle cell disease using detailed echocardiography. METHODS AND RESULTS: We undertook a prospective study in 122 consecutive stable outpatients with sickle cell disease and 30 age, gender and ethnicity-matched healthy controls. Echocardiographic evaluation included 3D ventricular volumes, sphericity, tissue Doppler, and non-invasive estimation of pulmonary vascular resistance. 36% of patients had a tricuspid regurgitant velocity ≥2.5 m.s(-1) but only 2% had elevated pulmonary vascular resistance and the prevalence of right ventricular dysfunction was very low. Patients with raised tricuspid regurgitant velocity had significantly elevated biventricular volumes and globular left ventricular remodelling, related primarily to anaemia. In a subgroup of patients who underwent cardiac catheterization, invasive pulmonary haemodynamics confirmed the echocardiographic findings. CONCLUSIONS: Elevated cardiac output and left ventricular volume overload secondary to chronic anaemia may be the dominant factor responsible for abnormal cardiopulmonary haemodynamics in patients with sickle cell disease. 3D echocardiography with non-invasive estimation of pulmonary vascular resistance represents a valuable approach for initial evaluation of cardiopulmonary haemodynamics in sickle cell disease.