Segmentation Of Breast Cancer And Prediction Of Immune Complete Response To Neoadjuvnat Chemotherapy

Dynamic contrast-enhanced magnetic resonance imaging is the most effective imaging method for breast cancer detection in clinical practice.The segmentation of breast cancer tumor regions can provide clinicians with reference lesion regions and input regions for prognosis prediction tasks.With the continuous development of deep learning technology,its advantages of end-to-end learning and adaptive learning of feature extraction methods can effectively improve segmentation accuracy.Image segmentation models based on convolutional neural networks have developed rapidly in the field of medical image processing.This paper conducts segmentation of breast cancer lesions and predicts whether subsequent patients can achieve pathological complete remission after receiving neoadjuvant chemotherapy.The data is used as training data,and the data from Henan Provincial Hospital is used as validation data.Deep neural network is used for step-by-step segmentation.First,the breast region is segmented,and then the breast cancer lesion region is segmented on the basis of the breast region.For singlecenter data,when it is applied to multi-center,due to the parameter differences between imaging devices,there are often inter-domain differences between data.In this paper,for the domain generalization problem,a method of histogram matching is proposed to improve the multicenter generalization ability of the model.The experimental results show that by introducing the histogram information of other centers,the generalization ability of the model can be improved without transmitting the patient’s private information.Based on the image segmentation results,the lesion area was used as the input,and the three-dimensional convolutional neural network was used to extract image features to predict the pathological complete remission results.Finally,through image-text fusion,the prediction accuracy of pathological complete remission is improved.Then,based on the image segmentation results,the lesion area was used as the input of the prediction model,and the three-dimensional convolutional neural network was used to extract image features to predict the pathological complete remission results.Finally,through image-molecular typing characteristics fusion,the prediction accuracy of pathological complete remission is improved.