计算机辅助超声诊断提高卵巢癌诊断的准确性  

A Computer-Aided Diagnostic Algorithm can Improve Accuracy in Diagnosis of Ovarian Cancer

作　　者：杨娜[1] 韩燕燕[1] 张梅娜[1] 张英涛[2] 宫丽华[1] 曲延峻[1] 

机构地区：[1]哈尔滨医科大学附属第一医院妇产超声科,黑龙江哈尔滨150001 [2]哈尔滨工业大学计算机学院,黑龙江哈尔滨150001

出　　处：《现代生物医学进展》2017年第2期276-279,共4页Progress in Modern Biomedicine

基　　金：黑龙江省卫生厅科研课题(2013013)

摘　　要：目的:评估应用计算机辅助阴式超声诊断卵巢癌是否优于单独阴式超声诊断。方法:收集2013年9月-2015年3月在我院进行手术切除的卵巢肿块患者的术前阴式超声(TVS)图像资料,共124例(病理诊断良性80例,恶性44例),平均年龄54.2±9.7岁。建立的图像样本库,每名患者取4幅超声图像,共有496幅图像被分析。首先对图片库内的超声图像进行多模式序列超声识别系统的处理,应用灰度共生矩阵法,获得肿块的各种特征,将感兴趣区域分类为正常组织,良性肿块和恶性肿瘤。两名超声医生人工肉眼对卵巢肿块的原始阴式超声图片进行分析诊断,随后两名医师对计算机辅助诊断(CAD)处理后的图片再次分析诊断,以病理诊断为金标准,利用ROC曲线下肿块区域的面积参数评价单独TVS和TVS结合CAD诊断卵巢癌的准确度差异。比较应用CAD前后的特异性、敏感性、阳性预测值,阴性预测值等指标的差异。结果:TVS结合CAD诊断准确度高于单独TVS,单独TVS诊断ROC曲线下面积(Az)为0.791,95%可信区间[0.725-0.857],TVS结合CAD诊断Az为0.927,95%可信区间[0.883-0.970];TVS结合CAD组诊断卵巢癌的精确度(0.774 vs 0.919,P<0.05)、特异性(0.738 vs 0.933,P<0.05)及阳性预测值(0.650 vs 0.889,P<0.05)较单独TVS诊断明显提高。结论:计算机辅助诊断技术可以辅助超声医师提高卵巢癌诊断的准确度。Objective： We compared the efficacy of the transvaginal sonography（TVS） assisted with a computer-aided diagnostic（CAD） algorithm to it of TVS in discriminating the benign tumor from malignant adnexal masses. Methods： Women（n=124, with pathological diagnosis of benign 80 cases and malignant 44 cases） scheduled for surgical removal of at least one ovary and had TVS image data in First Affiliated Hospital of Harbin Medical University during from September, 2013 to March, 2015. The average ages of the patients are 54.2±9.7 years. We established the image database. In total, 496 images were acquired, and 4 images of each case were chosen for image processing. We used the multi-mode ultrasound sequence recognition system to reanalyze the images, applicated gray co-occurrence matrix method to obtain tumor texture features of the region of interest（ROI） and to classify the ROI as normal tissue,benign tumors and malignant tumors, then we compared the accuracy of diagnosis of ovarian cancer TVS group and TVS＋ CAD group which are executed by 2 ultrasound doctors. The mean area under the ROC curve（Az） was preformed to evaluate the diagnostic accuracy using pathological diagnosis as the gold standard. Furthermore, the comparing parameters we calculated are include the accuracy, the specificity, the sensitivity, the positive predictive value, and the negative predictive value within before and after application of CAD.Results： The diagnostic accuracy was improved after CAD application compared to TVS alone, ROC area under the curve（Az） was0.791 [95% CI 0.725-0.857] in TVS alone, and 0.927 [95% CI 0.883-0.970] in TVS＋ CAD; Comparing TVS and TVS＋ CAD,the accuracy（0.774 vs 0.919, P〈0.05）, the specificity（0.738 vs 0.933, P〈0.05） and the positive predictive value（0.650 vs 0.889,P〈0.05）were significantly improved. Conclusions： The CAD algorithm is significantly useful to improve the diagnostic accuracy in Ovarian cancer and decrease false negative rates.

关 键 词：卵巢癌的诊断 阴式超声 计算机辅助诊断 

分 类 号：R737.3[医药卫生—肿瘤] R445[医药卫生—临床医学]