计算机辅助检测在乳腺癌X线诊断中的应用

Clinical application of computer-aided detection in breast cancers on mammograms

目的 评价计算机辅助检测 (CAD)系统在乳腺癌X线诊断中的临床应用价值。方法使用CAD系统对 136例乳腺癌钼靶X线片进行检测 ,以手术病理结果为标准对CAD标记结果进行统计分析 ,并与放射科医生的检测结果进行比较。结果 CAD系统对微小钙化的敏感性为 93 2 %(4 1/ 4 4) ,特异性为 2 2 8% (2 1/ 92 ) ,准确性为 4 5 6 % (6 2 / 136 ) ;CAD系统对肿块及结构紊乱病灶敏感性为 82 9% (10 7/ 12 9) ,特异性为 71 4 % (5 / 7) ,准确性为 82 4 % (112 / 136 )。经Kappa检验 ,CAD系统与放射科医生的检测结果一致性具有显著性意义 (Kappa肿块 =0 4 2 4 ,P肿块 =0 0 0 0 ;Kappa钙化 =0 36 5 ,P钙化 =0 0 13)。结论 CAD系统具有敏感性高、特异性较低的特点。

Objective To evaluate the clinical value of computer-aided detection (CAD) system in the diagnosis of breast cancer on mammograms. Methods The mammograms of 136 pathology-proved breast cancer cases were analyzed by CAD system. The results of CAD were compared with that made by radiologists. Results The sensitivity, specificity, and accuracy of CAD system to microcalcification lesions were 93.2% (41/44), 22.8%(21/92), and 45.6%(62/136), respectively. The sensitivity, specificity, and accuracy of CAD system to mass lesions were 82.9%(107/129), 71.4%(5/7), and 82.4%(112/136), respectively. There was middle level agreement between the detection of CAD system and that of the radiologists (Kappa_(mass)=0.424, P_(mass)=0.000; Kappa_(cal)=0.365, P_(cal)=0.013). Conclusion CAD system has the characteristics of high sensitivity and low specificity. It could be the aid of radiologists in breast cancer detection on mammograms.