Skip to main content
. 2022 Oct 4;3(1):e36660. doi: 10.2196/36660

Figure 1.

Figure 1

Schematic diagram of the multiple-input convolutional neural network (MI-CNN) architecture. There are two parts in the MI-CNN: the convolution part and classifier part. The convolution part consists of up to 16 MI-CNN inputs (depending on the type of MI-CNN), and each MI-CNN input has three convolutional sections and one max-pooling layer. The classifier part is composed of three fully connected (FC) layers. CXR: chest X-ray radiograph; ReLu: rectified linear unit.