FIGURE 5.

Model of the cardiac aligned responses. (a) First, two principal components (PCs) for all five subjects plotted as a function of the heartbeat cycle. Time zero is the photoplethysmography (PPG) peak. The first principal component is plotted in the solid line; the second component is plotted as a dashed line. (b) The cross‐validated variance explained by an increasing number of principal components plotted for each subject. (c) Canonical principal components across subjects as a function of heartbeat cycle. (d) Histogram showing the distribution of the relative root mean square errors across voxels. A R _mse < 1 indicates that the model trained on four out of five subjects explained half of the data in the fifth subject better compared to the other half of the data from the same subject. Note that the expected value of a perfect model and Gaussian measurement noise lies around $\frac{1}{\sqrt{2}}$. (e) The model with these two principal components can predict various response shapes, the waveforms show time ranging from −0.50 to 0.66 of the heartbeat cycle (indicated in green in (c))