灰阶超声影像组学对甲状腺结节的鉴别诊断

Application value of gray-scale ultrasonography in the differential diagnosis of benign and malignant thyroid nodules

目的分析灰阶超声影像组学在甲状腺结节良、恶性鉴别诊断中的应用价值。方法选择本院2019年5月至2022年5月接诊的286例甲状腺结节患者,均给予灰阶超声影像检查,将病理活检穿刺诊断结果作为本次研究金标准,统计灰阶超声诊断准确率、灵敏度、特异度,Kappa检验灰阶超声与金标准的一致性。比较良性结节、恶性结节声像图特征,比较良性结节、恶性结节时间-强度曲线(TIC曲线)参数。以Pyradiorllics程序提取甲状腺癌影像组学理特征。结果灰阶超声诊断准确率是97.6%,灵敏度是90.0%,特异度是99.6%。Kappa检验灰阶超声与金标准的一致性较好(Kappa值=0.786,P<0.05)。恶性结节沙粒样(36.7%)、粗大钙化率(63.3%)、后方无衰减率(80.0%)、内部回声不均匀率(88.3%)、低回声率(50.0%)、边界不清晰率(71.7%)、形态不规则率(91.7%)均高于良性结节(13.3%、2.6%、20.4%、11.5%、4.4%、13.3%、9.3%),差异有统计学意义(P<0.05)。恶性结节始-峰时间差低于良性结节,恶性结节上升支斜率高于良性结节,差异有统计学意义(P<0.05)。恶性结节峰值强度、达峰时间、始增时间与良性结节比较,差异无统计学意义(P>0.05)。从Pyradiorllics程序提取甲状腺癌影像组学特征152个,以lasso降维筛选出影像组学特征3种,分别是游程矩阵特征(RLM)8个、灰度共生矩阵特征(GLCM)2个、灰度直方图特征1个。结论灰阶超声影像组学在鉴别甲状腺结节良、恶性方面具有较高的准确率、灵敏度、特异度,同时可提供良、恶性病灶声像图特征、TIC曲线参数等信息,具有一定的临床价值。

Objective To analyze the application value of gray-scale ultrasonography in the differential diagnosis of benign and malignant thyroid nodules.Methods A total of 286 patients with thyroid nodules who were admitted to our hospital from May 2019 to May 2022 were selected and given gray-scale ultrasound imaging examination.The results of pathological biopsy puncture diagnosis were taken as the gold standard of this study,and gray-scale ultrasound was counted.Diagnostic accuracy,sensitivity,specificity,Kappa test gray-scale ultrasound and the consistency of the gold standard.The sonographic features of benign nodules and malignant nodules were compared,and the time-intensity curve(TIC curve)parameters of benign nodules and malignant nodules were compared.The radiomic features of thyroid cancer were extracted by Pyradiorllics program.Results The diagnostic accuracy of gray-scale ultrasound was 97.6%,the sensitivity was 90.0%,and the specificity was 99.6%.The consistency between gray-scale ultrasound and the gold standard of Kappa test was good,P<0.05(Kappa value=0.786).Malignant nodule sand-like(36.7%),coarse calcification rate(63.3%),posterior non-attenuation rate(80.0%),internal echo inhomogeneous rate(88.3%),hypoechoic rate(50.0%),unclear boundary rate(71.7%)and irregularity rate(91.7%)were higher than those of benign nodules(13.3%,2.6%,20.4%,11.5%,4.4%,13.3%,9.3%)(P<0.05).The time difference between the beginning and the peak of malignant nodules was lower than that of benign nodules,and the slope of ascending branch of malignant nodules was higher than that of benign nodules(P<0.05).The peak intensity,time to peak,and time to increase of malignant nodules were compared with those of benign nodules,and there had no significant difference(P>0.05).One hundred and fifty-two thyroid cancer radiomic features were extracted from the Pyradiorllics program,and 3 radiomic features were screened by lasso dimensionality reduction,of which there were 8 RLM(run-length matrix features),2 GLCM(gray-level co-occurrence matrix features),and one gray-level histogram feature.Conclusion Gray-scale ultrasound radiomics has high accuracy,sensitivity and specificity in distinguishing benign and malignant thyroid nodules,and can provide information such as sonographic features and TIC curve parameters of benign and malignant lesions.