数字钼靶计算机辅助诊断系统检测微钙化团簇的诊断能力评价

Diagnostic abilities of CAD system for assessing microcalcification clusters in digitized mammograms

目的 :自主开发的计算机辅助诊断系统 MCAD ,检测数字化乳腺片上的微钙化团簇 ,提高早期乳腺癌的诊断准确率。方法 :从互联网上的权威性数字化乳腺影像数据库下载 6 9个病例 ,用于有效性研究 ;对国内 2 9例含微钙化团簇的乳腺癌x线片进行回顾性研究 ;对国内 82 6例乳腺筛查x线片进行前瞻性研究。将MCAD系统对所有数字化乳腺片上的微钙化团簇检测结果与放射科医师诊断结果比较。结果 :在有效性研究中 ,MCAD系统诊断微钙化团簇的敏感性显著高于两位放射科医师 (0 .92 5vs0 .716 ,0 .75 ,P <0 .0 1) ,特异性则低于放射科医师 (0 .719vs 0 .938,0 .938,P <0 .0 0 1) ,平均每幅图像上有假团簇 0 .6个。在回顾性研究中 ,MCAD系统的诊断敏感性高于放射科医师 (0 .914vs 0 .716 ,0 .8,P =0 .0 98) ,假阳性团簇的检出数显著高于放射科医师 (0 .93vs 0 .0 3,0 .0 7,P <0 .0 0 1)。在前瞻性研究中 ,使用MCAD系统亦获得了更高的诊断准确率。结论 :MCAD系统对微钙化团簇的检测具有高敏感度 ,可帮助放射科医师发现隐匿于乳腺片复杂背景中的微钙化团簇 。

Objective:To test a mammography computer-aided diagnosis system(MCAD),which was developed by us.This system was applied to detect the microcalcification clusters(MCCs) in digitized mammograms,and to help the radiologists improve the diagnostic accuracy of the early breast cancer.Methods:To begin with,69 cases of mammograms from authoritative Internet digital mammography database were employed for validation study.Furthermore,29 native cases of breast cancer with MCCs were retrospectively studied.In the end,826 native cases of screening mammography were prospective analyzed.The MCCs were detected by MCAD system in all digitized mammograms.The results were compared with that of the experienced radiologists.Results:In validation study,the MCAD system had significant higher sensitivity and lower specificity than the radiologists in the detection of MCCs(sensitivity was 0.925 vs 0.716 and 0.75,P<0.01;specificity was 0.719 vs 0.938 and 0.938,P<0.001,respectively).The average false positive detection of MCAD system was 0.6 MCCs per image.In the retrospective study,the sensitivity of MCAD system was higher than the radiologists,but no significant difference(0.914 vs 0.716 and 0.8,P=0.098),while the false positive detection was significantly higher than the radiologists(0.93 vs 0.03 and 0.07,P<0.001).In prospective study,the application of MCAD system achieved a higher diagnostic accuracy.Conclusion:The MCAD system could provide high sensitivity in the detection of MCCs.It could help the radiologists detect MCCs hidden among complex background of mammograms,and is valuable in the early diagnosis of breast cancer.