Figure 1.

Proposed convolutional neural network model for estimating stroke volume from arterial pressure waveform. (A) Overall model architecture. (B) Details of the inception module. (C) Details of the nonlocal module. The variable k indicates kernel size of the convolution layer, d means the dilation rate of the convolution layer, and p represents pooling rate. GlobalAvgPool indicates the global average pooling layer, which computes the mean value for each feature map and supplies abstracted feature maps to the flattening layer. Dotted arrows represent dropout of 0.5, while solid arrows mean full connections. BS: batch size; FC: fully connected.