计算机辅助检测和诊断中K‐TIRADS、ACR‐TIRADS、ATA的诊断效能比较以及辅助超声医师诊断甲状腺结节的研究

Study on the comparison of diagnostic of K-TIRADS, ACR-TIRADS and ATA in CAD and diagnosis of thyroid nodules by computer-assisted ultrasonography

目的探讨计算机辅助检测和诊断(CAD)软件中 K‐TIRADS 、 ACR‐TIRADS 、 ATA风险分层的诊断效能以及辅助超声医师诊断甲状腺结节的应用价值.方法回顾性分析192个有术后病理的甲状腺结节资料,使用CAD软件对甲状腺结节分别进行K‐TIRADS 、 ACR‐TIRADS 、 ATA 分类,统计ROC曲线下面积、敏感性、特异性等得出最佳指南,以最佳指南为分类标准,采用双盲法对比同一超声医师在使用CAD前、结合CAD后对甲状腺结节的诊断能力.结果 CAD软件中 K‐TIRADS 、ACR‐T IRADS 、 A T A的AUC分别为 0 .88 、 0 .77 、 0 .62 ,两两比较差异有统计学意义( P < 0 .05 );K‐T IRADS 与AT A的特异性比较差异无统计学意义( P = 0 .176 ),但均高于 ACR‐T IRADS ( P <0 .05). CAD 、超声医师、超声医师结合CAD使用K‐TIRADS的AUC分别为0 .88 、 0 .80 、 0 .93 ,两两比较差异有统计学意义( P <0 .05);CAD和超声医师结合CAD敏感性比较差异无统计学意义( P =0 .163),但均高于超声医师人工判读( P <0 .05);CAD 、超声医师、超声医师结合CAD特异性两两比较差异有统计学意义( P <0 .05).结论 CAD软件中3种甲状腺超声风险分层系统均具有良好的诊断价值,其中K‐TI RADS的AUC最大,CAD软件辅助超声医师能提高对诊断甲状腺结节的诊断效能,具有较好的临床应用前景.

Objective To explore the diagnostic efficiency of K-TIRADS, ACR-TIRADS and ATA risk stratification in computer-aided detection and diagnosis(CAD) software and the application value of CAD-assisted ultrasound physicians in diagnosing thyroid nodules. Methods One hundred and ninety-two thyroid nodules with postoperative pathological results were retrospectively analyzed. All of them were graded by K-TIRADS, ACR-TIRADS and ATA with CAD software, and the best guide was recognized by calculating the area under the ROC curve, sensitivity and specificity. Then, based on the best guidelines for the classification criteria, the double-blind method was used to compare the ability of the same ultrasonologist to diagnose thyroid nodules before and after CAD. Results The AUC value of K-TIRADS, ACR-TIRADS, ATA was 0.88, 0.77, 0.62 respectively in the CAD software. The difference between the two groups was statistically significant (P<0.05). There was no significant difference in the specificity between K-TIRADS and ATA(P=0.176), which were both higher than ACR-TIRADS with statistically significant differences (P<0.05). The AUC value of the diagnosis among CAD itself, ultrasound physicians and physicians combined CAD was 0.88, 0.80, 0.93, respectively. The difference between the two groups was statistically significant (P<0.05). There was no significant difference in the sensitivity between CAD itself and physicians combined CAD(P=0.163), which were both higher than ultrasound physicians with statistical significant differences(P<0.05). Among ultrasound physicians, CAD itself and physicians combined CAD, the difference in specificity between the two groups was statistically significant(P<0.05). Conclusions All the three risk stratification systems of thyroid ultrasound in CAD software have good diagnostic values, among which K-TIRADS has the largest AUC. The CAD software can assist ultrasound physicians to improve the thyroid nodule diagnostic performance, and has a good clinical application prospect.