探讨MRI与CT的Haralick纹理参数诊断良恶性肺结节价值
Exploring the Haralick Texture Parameters of MRI and CT for the Diagnosis of Value of Benign and Malignant Pulmonary Nodules
楼俭茹 1孟思 1覃群 1邓镇生 2刘玉银 1刘海洲 1刘明1
作者信息
- 1. 三亚中心医院(海南省第三人民医院)放射科(海南三亚 572000)
- 2. 三亚中心医院(海南省第三人民医院)胸外科(海南三亚 572000)
- 折叠
摘要
目的 通过提取肺结节MRI与CT图像的Haralick纹理参数,探讨其对于肺结节良恶性的诊断效能.方法 选取40名就诊于三亚中心医院肺亚实性结节的患者为研究对象,根据病理结果将患者分为良性组(n=20),恶性组(n=20).对所有入组患者分别进行CT以及MRI扫描,选取病灶较清晰的图像层面中提取角二阶矩、对比度、自相关、逆差矩、熵5个纹理特征参数,采用单因素方差分析比较良性组与恶性组之间的纹理特征差异.结果 CT组的熵无统计学意义(P＞0.05),其余纹理参数均有统计学意义;T1WI与T2WI组中的逆差矩和熵均有统计学意义(P＜0.05),角二阶矩、对比度、自相关均无统计学意义(P＞0.05);DWI组中仅对比度有统计学意义(P＜0.05).比较各组AUC值发现,CT组的AUC值普遍较高,MRI组的T1WI中熵的AUC值最高(AUC=0.970).结论 CT纹理分析仍具有一定优势,加入MRI多序列检查能够提高肺良恶性结节的诊断准确性.
Abstract
Objective This study leverages Haralick texture parameters to analyze image characteristics of benign and malignant pulmonary nodules,assessing the diagnostic efficacy of MRI and CT imaging in differentiating these conditions.Methods A total of 40 patients with pulmonary subsolid nodules from Sanya Central Hospital were selected for the study.Based on pathological results,the patients were divided into a benign group(n=20)and a malignant group(n=20).Each patient underwent both CT and MRI scans.From the clearest lesion images,five texture features were extracted:angular second moment,contrast,autocorrelation,inverse difference moment,and entropy.One-way ANOVA was used to compare the differences in these texture features between the benign and malignant groups.Results In the CT group,entropy was not statistically significant(P＞0.05),while the other texture parameters were statistically significant.In the T1WI and T2WI groups,both inverse difference moment and entropy were statistically significant(P＜0.05),whereas angular second moment,contrast,and autocorrelation were not statistically significant(P＞0.05).In the DWI group,only contrast was statistically significant(P＜0.05).Comparing the AUC values across groups,the CT group generally had higher AUC values,with the highest AUC value observed for entropy in the T1WI group(AUC=0.970).Conclusion CT texture analysis still has certain advantages,but incorporating multiple MRI sequences can improve the diagnostic accuracy for distinguishing between benign and malignant pulmonary nodules.
关键词
肺结节/纹理分析/磁共振成像(MR)/计算机体层摄影(CT)Key words
Pulmonary Nodule/Texture Analysis/Magnetic Resonance imaging/Computed Tomography引用本文复制引用
基金项目
海南省卫生健康行业科研项目(21A200132)
出版年
2024
NETL
NSTL
万方数据