CT纹理分析技术鉴别甲状腺良恶性结节可行性研究

Can computed tomography(CT) texture analysis help to differentiate benign and malignant nodules of thyroid

目的探讨CT纹理分析技术在鉴别甲状腺良恶性结节中的价值。方法回顾性分析经我院手术病理证实的甲状腺病变病人35例,共42个病灶,其中恶性结节26个,良性结节16个。所有病人治疗前均行颈部增强CT扫描。将DICOM格式的CT增强图像(层厚和层间距均为5 mm)导入CT Kinetics软件进行纹理及直方图分析得到未经滤过的原始细纹理图像。CT纹理分析主要参数包括熵值、偏度、峰态、平均像素值和像素分布的标准差。甲状腺良恶性结节间纹理参数比较采用独立样本t检验或Mann-Whitney U检验,并对有统计学意义的纹理参数进行受试者操作特征(ROC)曲线分析,确定诊断阈值。结果甲状腺恶性结节的熵值、偏度、峰态、像素值和标准差分别为6.65±0.92、0.63±1.37、0.69±1.23、84.08±23.36和18.14±3.31;良性结节分别为5.96±0.54、0.59±1.42、0.51±1.17、72.00±24.52和20.05±6.10。熵值在甲状腺良恶性结节间差异有统计学意义(P<0.05),偏度、峰态、像素值和标准差在甲状腺良恶性结节间差异均无统计学意义(均P>0.05)。ROC曲线分析显示,以熵值6.09为鉴别甲状腺结节良恶性的阈值,其ROC曲线下面积、敏感度和特异度分别为0.733、71.3%和70.0%。结论 CT纹理参数对鉴别甲状腺结节的良恶性有一定帮助。

Objective This study was aimed to determine the ability of texture analysis of contrast-enhanced computed tomography （CT） in differentiating benign and malignant thyroid nodules. Methods The clinical data of 35 patients with 42 thyroid nodules were retrospectively analyzed. Patients were classified as malignant nodules（n=26） or benign nodules （n=16） group based on their histological examination. All nodules were examined by contrast-enhance neck CT before surgery. The DICOM format enhancement CT images （thickness/gap=5 mm） were imported into CT Kinetics software and the parameters of CT texture were calculated automatically with texture and histogram algorithm to generate unfilter fine texture images. The parameters of CT texture, including entropy, skewness, kurtosis, mean pixels and standard deviation （SD） were derived from contrast-enhanced CT images using unfilter fine texture. The diagnostic value of CT texture parameters were examined with Student’s t-test or Mann-Whitney U test, and receiver operating characteristic （ROC） curve analysis was used to determine optimal CT texture parameters in differentiating malignant and benign lesions. Results For malignant thyroid nodules, the values of entropy, skewness, kurtosis, mean pixels and SD were 6.65±0.92, 0.63±1.37, 0.69±1.23, 84.08±23.36 and 18.14±3.31, respectively. For benign thyroid nodules, the values of entropy, skewness, kurtosis, mean pixels and SD were 5.96 ±0.54, 0.59 ±1.42, 0.51 ±1.17, 72.00 ±24.52 and 20.05 ±6.10, respectively. The entropy values of malignant thyroid nodules were significantly higher than that of the benign thyroid lesions （P〈0.05）;While the values of skewness, kurtosis, mean pixels and SD did not differ significantly （all P〉0.05） between the malignant and benign nodules of thyroid. When entropy value of 6.09 was used as a cut-off for differentiating benign from malignant thyroid lesions, the AUC, sensitivity, and specificity in differentiating benign from malignant thyroid lesions were 0.733, 71.3%, and 70.0%, respectively. Conclusions CT texture analysis is helpful in differentiating the benign from malignant thyroid nodules.