乳腺超声中多普勒特征的提取及应用

Feature extraction and application of color Doppler in breast ultrasound image

为弥补以往单纯利用乳腺超声图像中灰度信息进行分类的过程中,易受噪声和其他因素影响而造成误分的缺点,提出了一种基于纹理特征和多普勒特征相结合的乳腺超声诊断系统.针对乳腺超声图像中的动态彩色多普勒信息建立数学模型进行特征提取,并结合图像中的纹理信息进行肿瘤的良恶性分类.实验结果表明,该系统的诊断准确性为92.30%,敏感性为91.42%,特异性为93.33%,假阳性率为94.11%,假阴性率为90.32%,接收特性AUC面积为0.9523,性能优于单纯利用纹理特征的分类结果,从而证明其在提高分类精度方面的有效性.

To reduce the impact of noises and other factors in traditional feature extraction of breast ultrasound based on gray level images, we propose a novel computer aided diagnosis system for breast cancer. Mathematical model was established using both texture features and color Doppler features to distinguish the benign and malignant tumors. The accuracy, sensitivity specificity, false positive rate and false negative rate of the proposed system are 92.30%, 91.42%, 93.33%, 94.11% and 90.32%, respectively. Experimental results show that the proposed system is valuable for radiologists to improve the accuracy of the diagnosis of breast cancer.