Research On Segmentation And Classification Algorithm Of Breast Tumor Images

Breast cancer is one of the most malignant tumors that threaten women’s health.Early detection and diagnosis play a vital role in the prevention and treatment of breast cancer.With the development and application of medical image processing technology,the use of automatic analysis methods for quantitative analysis and segmentation of breast tumor images has become an important auxiliary method for modern medical diagnosis.Computer-aided diagnosis makes up for the shortcomings of large errors and low efficiency in manual detection,provides medical experts with a more intuitive and accurate judgment basis,and has important application value in the research of related diseases.This article mainly conducts in-depth research on the segmentation and classification of breast tumor images.Usually due to the influence of the detection system or human operation,the collected breast tumor images often have noise spots,low contrast between cells and background,etc.,which makes the accuracy of tumor images in the segmentation and classification process reduced.Therefore,in order to improve the diagnosis efficiency of breast tumors,this article will carry out work from the following two aspects,the main research contents are as follows:(1)A segmentation algorithm of breast tumor image based on improved watershed model is proposed.The algorithm uses the Canny operator to obtain the weighted gradient map,then combines the morphology to filter and reconstruct the gradient image,and uses the idea of gridding to determine the local minimum to construct the labeled image.Finally,the labeled image is used to modify the gradient image and find the segmentation result.The experimental results show that the improved watershed model proposed in this paper improves the merger ratio value and the segmentation accuracy by 31.86% and 21.88%,respectively,compared with the classic watershed algorithm;compared with the morphological watershed algorithm,its value is improved 18.21% and 11.36%.(2)A breast tumor classification algorithm based on PCA-SVM hybrid model is established.This model adopts PCA method to reduce the dimension of the data set in the input SVM to solve the problem of high cost of highdimensional data.Then,the improved grid search algorithm combined with cross-validation is used to optimize the parameters in the classifier to determine the optimal parameters.Finally,the breast tumor data set is input into a hybrid classifier for testing and analysis to verify the practicability of the PCA-SVM hybrid model.The experimental results show that the optimal classification accuracy of breast tumor cell test set based on the PCA-SVM hybrid model proposed in this paper reaches 97.86%,which is improved by 9.80%,18.32% and 10.76% respectively compared with LVQ neural network,BP neural network algorithm and decision tree.