CT平扫图像纹理分析鉴别腮腺多形性腺瘤与恶性肿瘤的初步研究

Preliminary Study on Differentiating Pleomorphic Adenoma and Malignant Tumors of the Parotid Gland by Texture Analysis of Non-Enhanced CT Images

目的:探讨CT平扫图像纹理分析方法鉴别腮腺多形性腺瘤及腮腺恶性肿瘤的可行性。方法:对经过CT平扫检查且经病理证实的45例腮腺多形性腺瘤和12例腮腺原发恶性肿瘤进行回顾性分析。应用软件工具在这些CT平扫图像上对病灶进行感兴趣区(ROI)的勾画,获得18个纹理参数。使用非参数检验对腮腺多形性腺瘤与恶性肿瘤的纹理参数进行统计学分析,并作ROC曲线,以评估有统计学差异的纹理参数的诊断效果。结果:由CT平扫的图像生成的18个不同的纹理参数中,共有10种纹理参数有统计学差异(P <0.05),其中中位数强度、平均值、体素值和、均方根误差、离均差诊断效能较高,AUC均大于0.8。结论:用CT平扫图像纹理分析方法可以鉴别腮腺多形性腺瘤和腮腺恶性肿瘤。

Objective: To evaluate the possibility of texture analysis on non-enhanced CT images for differentiating pleomorphic adenoma and malignant tumors of the parotid gland. Materials and Methods: 45 cases of parotid pleomorphic adenoma and 12 cases of primary malignant tumor of the parotid gland confirmed by pathology were included in this retrospective study. Texture analysis software was used to delineate region of interest(ROI) on the lesions of these non-enhanced CT images, and obtained 18 texture parameters. Nonparametric test was used to statistically analyze the texture parameters of parotid pleomorphic adenomas and malignant tumors, and the ROC curve was used to evaluate the diagnostic effect of parameters with statistical difference. Results: Among the 18 different CT texture parameters generated on these non-enhanced CT images, 10 texture parameters were statistically significant between the PA group and the MT group(P < 0.05). Among them, Median Intensity, Mean Value Voxel Value Sum, RMS, Mean Deviation have higher diagnostic efficacies, the AUCs are all greater than 0.8. Conclusion: Texture analysis of CT plain scan images can be used to identify parotid pleomorphic adenomas and parotid malignancies.