Using Machine Learning to Identify Organ System Specific Limitations to Exercise via Cardiopulmonary Exercise Testing

Cardiopulmonary Exer cise Testing (CPET) is a unique physiologic medical test used to evaluate human response to progressive maximal exercise stress. Depending on the degree and type of deviation from the normal physiologic response, CPET can help identify a patient's specific limitations to exercise to guide clinical care without the need for other expensive and invasive diagnostic tests. However, given the amount and complexity of data obtained from CPET, interpretation and visualization of test results is challenging. CPET data currently require dedicated training and significant experience for proper clinician interpretation. To make CPET more accessible to clinicians, we investigated a simplified data interpretation and visualization tool using machine learning algorithms. The visualization shows three types of limitations (cardiac, pulmonary and others); values are defined based on the results of three independent random forest classifiers. To display the models' scores and make them interpretable to the clinicians, an interactive dashboard with the scores and interpretability plots was developed. This machine learning platform has the potential to augment existing diagnostic procedures and provide a tool to make CPET more accessible to clinicians.

The lowest VE/VCO2 ratio during exercise as a predictor of outcomes in patients with heart failure

BACKGROUND

The lowest minute ventilation (VE) and carbon dioxide production (VCO(2)) ratio during exercise has been suggested to be the most stable and reproducible marker of ventilatory efficiency in patients with heart failure (HF). However, the prognostic power of this index is unknown.

METHODS AND RESULTS

A total of 847 HF patients underwent cardiopulmonary exercise testing (CPX) and were followed for 3 years. The associations between the lowest VE/VCO(2) ratio, maximal oxygen uptake (peak VO(2)), the VE/VCO(2) slope, and major events (death or transplantation) were evaluated using proportional hazards analysis; adequacy of the predictive models was assessed using Akaike information criterion (AIC) weights. There were 147 major adverse events. In multivariate analysis, the lowest VE/VCO(2) ratio (higher ratio associated with greater risk) was similar to the VE/VCO(2) slope in predicting risk (hazard ratios [HR] per unit increment 2.0, 95% CI 1.1-3.4, and 2.2, 95% CI 1.3-3.7, respectively; P < .01), followed by peak VO(2) (HR 1.6, 95% CI 1.1-2.4, P=.01). Patients exhibiting abnormalities for all 3 responses had an 11.6-fold higher risk. The AIC weight for the 3 variables combined (0.94) was higher than any single response or any combination of 2. The model including all 3 responses remained the most powerful after adjustment for beta-blocker use, type of HF, and after applying different cut points for high risk.

CONCLUSIONS

The lowest VE/VCO(2) ratio adds to the prognostic power of conventional CPX responses in HF.

Ventilatory efficiency and resting hemodynamics in hypertrophic cardiomyopathy

PURPOSE

In patients with systolic heart failure, the ability of cardiopulmonary exercise testing (CPX) variables to reflect pathophysiology is well established. The relationship between CPX and pathophysiology has, however, not been thoroughly investigated in patients with nonobstructive hypertrophic cardiomyopathy (NHCM). The objective of this study was to assess the ability of CPX variables to reflect resting hemodynamics in patients with nonobstructive hypertrophic cardiomyopathy NHCM.

METHODS

We performed CPX and right heart catheterization on 83 subjects with NHCM (51 male/32 female, mean age = 38 +/- 10 yr, NYHA I-III mean = 1.7). Peak oxygen consumption ( O2) and minute ventilation/carbon dioxide ratio (V E/VCO2) at peak exercise were compared to resting hemodynamics including pulmonary artery systolic, diastolic and mean pressures (PASP, PADP and MPAP), and pulmonary capillary wedge pressure (PCWP).

RESULTS

Elevations in PCWP (> or = 15 mm Hg), PASP (> or =30 and > or = 40 mm Hg), PADP (> 15 mm Hg) and MPAP (> or = 20 mm Hg) were detected in 22, 33, 10, and 23% of subjects, respectively. Peak V E/VCO2 (positive correlation) and peak VO2 (negative correlation) correlated modestly with all pressure measurements (r = 0.33-0.51, P < 0.01 for all measurements). By receiver operating curve analysis, a V E/VCO2 >35.5 exhibited the best diagnostic accuracy with a curve areas of 0.81 for PAP > or = 30 mm Hg (sensitivity/specificity = 86%/67%), 0.87 for PAP > or = 40 mm Hg (77%/100%), 0.86 for MPAP > 20 mm Hg (83%/79%), and 0.84 for PCWP > or = 15 mm Hg (80%/76%).

CONCLUSIONS

CPX can accurately identify abnormal resting hemodynamics in patients with NHCM. Further testing of this modality in other forms of diastolic dysfunction may be warranted.