Suppr超能文献

磁共振图像纹理分析鉴别良恶性黏液样软组织肿瘤:一项回顾性对比研究。

Texture analysis of magnetic resonance image to differentiate benign from malignant myxoid soft tissue tumors: A retrospective comparative study.

机构信息

Department of Radiology, Seoul National University Bundang Hospital, Bundang-gu, Seongnam-si, Gyeonggi-do, Korea.

出版信息

PLoS One. 2022 May 19;17(5):e0267569. doi: 10.1371/journal.pone.0267569. eCollection 2022.

Abstract

It is important to differentiate between benign and malignant myxoid tumors to establish the treatment plan, determine the optimal surgical extent, and plan postoperative surveillance, but differentiation may be complicated by imaging-feature overlap. Texture analysis is used for quantitative assessment of imaging characteristics based on mathematically calculated pixel heterogeneity and has been applied to the discrimination of benign from malignant soft tissue tumors (STTs). In this study, we aimed to assess the diagnostic value of the texture features of conventional magnetic resonance images for the differentiation of benign from malignant myxoid STTs. Magnetic resonance images of 39 patients with histologically confirmed myxoid STTs of the extremities were analyzed. Qualitative features were assessed and compared between the benign and malignant groups. Texture analysis was performed, and texture features were selected based on univariate analysis and Fisher's coefficient. The diagnostic value of the texture features was assessed using receiver operating curve analysis. T1 heterogeneity showed a statistically significant difference between benign and malignant myxoid STTs, with substantial inter-reader reliability. The sensitivity, specificity, positive predictive value, negative predictive value, and accuracy of T1 heterogeneity were 55.6%, 83.3%, 88.2%, 45.5%, and 64.1%, respectively. Among the texture features, T2w-WavEnLL_s-3 showed good diagnostic performance, and T2w-WavEnLL_s-4 and GeoW4 showed fair diagnostic performance. The logistic regression model including T1 heterogeneity and T2_WavEnLL_s-4 showed good diagnostic performance. However, there was no statistically significant difference between the overall qualitative assessment by a radiologist and the predictor model. Geometry-based and wavelet-derived texture features from T2-weighted images were significantly different between benign and malignant myxoid STTs. However, the texture features had a limited additive value in differentiating benign from malignant myxoid STTs.

摘要

区分良性和恶性黏液样肿瘤对于制定治疗计划、确定最佳手术范围和规划术后监测非常重要,但影像学特征的重叠可能会使鉴别变得复杂。纹理分析用于基于数学计算的像素异质性对影像学特征进行定量评估,已应用于鉴别良性和恶性软组织肿瘤(STT)。本研究旨在评估常规磁共振成像纹理特征对区分四肢黏液样 STT 的良恶性的诊断价值。分析了 39 例经组织学证实的四肢黏液样 STT 患者的磁共振图像。评估并比较了良性和恶性组之间的定性特征。进行了纹理分析,并基于单变量分析和费希尔系数选择纹理特征。使用受试者工作特征曲线分析评估纹理特征的诊断价值。T1 异质性在良性和恶性黏液样 STT 之间存在统计学显著差异,具有较大的读者间可靠性。T1 异质性的敏感性、特异性、阳性预测值、阴性预测值和准确性分别为 55.6%、83.3%、88.2%、45.5%和 64.1%。在纹理特征中,T2w-WavEnLL_s-3 具有良好的诊断性能,T2w-WavEnLL_s-4 和 GeoW4 具有较好的诊断性能。包括 T1 异质性和 T2_WavEnLL_s-4 的逻辑回归模型显示出良好的诊断性能。然而,放射科医生的整体定性评估与预测模型之间没有统计学上的显著差异。T2 加权图像的基于几何和基于小波的纹理特征在良性和恶性黏液样 STT 之间存在显著差异。然而,纹理特征在区分良性和恶性黏液样 STT 方面的附加价值有限。

https://cdn.ncbi.nlm.nih.gov/pmc/blobs/6481/9119440/4d8889ba84c4/pone.0267569.g001.jpg

文献AI研究员

20分钟写一篇综述,助力文献阅读效率提升50倍。

立即体验

用中文搜PubMed

大模型驱动的PubMed中文搜索引擎

马上搜索

文档翻译

学术文献翻译模型,支持多种主流文档格式。

立即体验