文章摘要

文章信息/Info

Title:: Classification of benign and malignant lesions of breast lesions based on ultrasound radio frequency time series analysis

Author(s):: LIN Chunyi1; ZOU Bo1; ZHOU Jianhua2; 1.School of Electronic and Information Engineering, South China University of Technology, Guangzhou 510640, China;2.Cancer Center, Sun Yat-sen University, Guangzhou 510080

Keywords:: Ultrasonic RF time series; Tissue characterization; Breast lesions benign and malignant classification; Image assisted diagnosis; Time series analysis

摘要:: 尝试为乳腺病灶良恶性分类提供高精度且无创的影像辅助诊断手段，提出了基于超声射频（radio frequency，RF）时间序列分析的乳腺病灶良恶性分类方法。以275例女性乳腺病灶为样本，使用常规超声探头采集多帧超声回波RF信号，提取RF时间序列的分形维数（fractal dimension，FD）、频域特征和时域特征，以支持向量机（support vector machine，SVM）和随机森林作为分类器对乳腺病灶进行良恶性分类。分类结果如下：SVM受试者工作特征曲线下的面积（area under receiver operating characteristics curve，AUC）为0.914，分类准确率为85.37%，随机森林AUC为0.937，分类准确率为91.46%。与以往研究相比，提高了乳腺病灶良恶性的分类精度，并开发了乳腺病灶良恶性分类系统，可以给医生提供诊断参考。

Abstract:: Try to provide high-accuracy and noninvasive imaging-assisted diagnostic methods for benign and malignant breast lesions,a method for classification of benign and malignant breast lesions based on ultrasound radio frequency time series analysis was proposed. For 275 samples of breast lesions, a conventional ultrasonic probe was used to acquire multi-frame ultrasound echo radio frequency (RF) signal. Fractal dimension, frequency domain and time domain characteristics from the ultrasonic RF time series were extracted, and then these breast samples were classified by using machine learning algorithm support vector machine (SVM) and random forests. The classification results were as follows: area under receiver operating characteristics curve(AUC) of SVM was 0.914, the classification accuracy rate was 85.37%, the AUC of the random forest was 0.937, and the classification accuracy rate was 91.46%. Compared with the previous studies, the classification accuracy of benign and malignant breast lesions is improved, and the benign and malignant breast lesions classification system is developed, which can provide diagnostic reference for doctors.