Attentive Octave Convolutional Capsule Network for Medical Image Classification

Medical image classification plays an essential role in disease diagnosis and clinical treatment. More more research efforts have been dedicated to the design of effective methods for medical classification. As framework, capsule network (CapsNet) can realize translation equivariance. Lots current applies networks analysis. In this paper, we propose attentive octave convolutional (AOC-Caps) AOC-Caps, AOC module is used replace traditional convolution operation. The purpose process fuse high- low-frequency information input simultaneously, weigh important parts automatically. Following module, a matrix expectation maximization (EM) algorithm applied update routing weights. proposed AOC-Caps comparative are tested on seven datasets, including PathMNIST, DermaMNIST, OCTMNIST, PneumoniaMNIST, OrganMNIST_Axial, OrganMNIST_Coronal, OrganMNIST_Sagittal, which from MedMNIST. experiments, baselines include CNN models, automated machine learning (AutoML) methods, related methods. experimental results demonstrate that achieves better performance most datasets.