Calcification Classification Model Based On Prior Knowledge And Attention Mechanism

In recent years,due to the increase in the incidence of thyroid nodules,more and more people have undergone thyroid ultrasound examination.It is hard to avoid misjudgment when doctors check thyroid ultrasound images for a long time.The degree of nodule calcification is an important basis for the diagnosis of thyroid cancer.If we can provide a reference calcification information for the judgment of doctors through the thyroid nodule calcification classification model,it will be of great significance.Traditional CNN(Convolutional Neural Networks)has a general performance on thyroid nodule data sets.Aiming at the limitation of neural network,it is improved in this paper.On the basis of feature fusion,this paper puts forward a calcification classification model based on prior knowledge and attention mechanism.By introducing prior knowledge,the neural network can pay more attention to calcification.The introduction of attention mechanism can enable the neural network to better learn the internal relationship between calcification features and focus on the important features.In order to verify the superiority of the experimental model,this paper compares it with other classification models from the perspective of the four indexes such as accuracy,macroP,macroR and macroF1.The experimental results show that the model in this paper performs best in four indexes and has the best classification effect.