MEDICAL IMAGE CLASSIFICATION BASED ON IMPROVED RESIDUAL NETWORK

Medical image classification plays an important role in computer-aided diagnosis and treatment. Traditional medical image classification methods usually have some problems, such as difficult manual marking features, complex calculation, low efficiency and poor classification effect. In this paper, a medical image aided diagnosis model based on AM-ResNet network is designed. The data was expanded by GAN and other methods to improve the generalization ability of the algorithm. The attention mechanism module was introduced into the AM-ResNet network to strengthen the image Feature extraction. The transfer learning method was adopted to speed up the convergence time of the model. In the experiment, the public breast cancer pathological image data set was used for diagnostic testing. The proposed model has an accuracy rate of about 97% for benign and malignant binary classification, and about 93% for multi-classification diagnosis of eight subtypes. The experimental results show that the model has high classification accuracy for classification tasks, and has good robustness and generalization ability.