Classification Of Breast Tumor Ultrasound Images Based On Machine Learning And Clinical Features

Breast cancer is one of the most common types of cancer in women,a fatal threat to women’s lives and health.With the development of modern medical science,if breast cancer is diagnosed early,patients’ survival rate will be greatly improved.Breast tumors are usually examined by CT,MRI,molybdenum target X-ray,far infrared,ultrasound(US)and other methods.Among them,breast ultrasound is widely used in clinical diagnosis because of its routine,fast,safe and painless.However,because of the high sensitivity of ultrasonic instruments,it is susceptible to the influence of different tissues of the human body and environment,resulting in a lot of speckle noise,which will cause interference to doctors’ diagnosis.In view of the challenge brought by speckle noise interference in breast ultrasound image classification,the automatic breast ultrasound image diagnosis system is studied from two aspects: traditional machine learning and deep learning.On the one hand,a classification method based on multi-features and support vector machine is proposed.Multi-features are composed of characteristic features and deep learning features.Initially,an improved level set algorithm is used to segment the lesion in breast ultrasound images,which provided an accurate calculation of characteristic features,such as orientation,edge indistinctness,characteristics of posterior shadowing region and shape complexity;Simultaneously,we use transfer learning to construct a pre-trained model as feature extractor to extract the deep learning features of breast ultrasound images;Finally,the multi-features are fused and fed to support vector machine for the further classification of breast ultrasound images.The proposed model when tested on unknown samples provided a classification accuracy of 92.5% for cancerous and non-cancerous tumors.On the other hand,the fuzzy enhancement and bilateral filtering algorithm are used to process the original breast ultrasound image firstly.Then,various decomposition images representing breast tumors’ clinical characteristics are obtained using the original image and the mask image.Considering the diversity of the benign and malignant characteristic information represented by each decomposition image,the decomposition images are fused through the RGB channel,and three kinds of fusion images are generated.Then,from a series of candidate deep learning models,transfer learning is used to select the best model as its base model to extract deep learning features.Finally,while training the classification network,the adaptive spatial feature fusion technology is used to train the weight network to complete the deep learning feature fusion and classification.The experimental results show that the accuracy,precision,specificity,sensitivity/recall,F1 score,and AUC of the proposed method are 0.9548,0.9811,0.9833,0.9392,0.9571,and 0.9883,respectively.Our research is able to automate breathe cancer detection and has strong clinical utility.