Utilizing Machine Learning Techniques to Predict the Efficacy of Aerobic Exercise Intervention on Young Hypertensive Patients Based on Cardiopulmonary Exercise Testing

Figures

• 

  Figure 1. Illustration of cardiopulmonary exercise testing. To protect privacy, the patient's eyes were partially blurred.

• 

  Figure 2. Visualization of several indicators for a patient during the exercise phase of a CPET. (a) Heart rate. (b) Stroke volume. (c) Cardiac output. (d) Oxygen pulse. (e) Oxygen consumption/kilogram. (f) Tidal volume. (g) Ventilation volume/minute. (h) Respiratory exchange ratio. (i) Carbon dioxide ventilation equivalent.

• 

  Figure 3. The SRC schematic for a two-class problem, where y is the unlabeled sample. D1 and D2 represent subdictionaries composed of all normalized samples belonging to class 1 and class 2, respectively. The sparse representation α can be divided into two subvectors (α1 and α2) according to the number of columns in D1 and D2. The black-filled blocks in α represent nonzero elements.

• 

  Figure 4. The SRC-AL schematic for a two-class problem, where x1 and x2 represent the samples of class 1 and class 2, respectively, and y is the unlabeled sample. Due to the limited space, only two labeled samples are drawn in the figure. In fact, all labeled samples are used for dictionary learning. Dup' and Dlw' represent the upper and lower parts of the learned dictionary. αx1, αx2, and αy are the sparse representations of x1, x2, and y where the black-filled blocks represent nonzero elements.

• 

  Figure 5. Illustration of the confusion matrix.

• 

  Figure 6. Comparison between the predictive performance of each indicator obtained by machine learning models and that of BP change obtained by the baseline method. (a, b) The average and optimal performance of metabolic indicators, respectively.