Department of Radiology, Tongji Hospital, Tongji Medical College, Huazhong University of Science and Technology, 1095# Jiefang Avenue, Wuhan, 430030, China.
Department of Radiology, Weill Cornell Medicine, 407 E 61st Street Suite 107, New York, NY, 10065, USA.
Eur Radiol. 2019 Jun;29(6):2751-2759. doi: 10.1007/s00330-018-5921-1. Epub 2019 Jan 7.
Texture analysis performed on MRI images can provide additional quantitative information that is invisible to human assessment. This study aimed to evaluate the feasibility of texture analysis on preoperative conventional MRI images in predicting early malignant transformation from low- to high-grade glioma and compare its utility to histogram analysis alone.
A total of 68 patients with low-grade glioma (LGG) were included in this study, 15 of which showed malignant transformation. Patients were randomly divided into training (60%) and testing (40%) sets. Texture analyses were performed to obtain the most discriminant factor (MDF) values for both training and testing data. Receiver operating characteristic (ROC) curve analyses were performed on MDF values and 9 histogram parameters in the training data to obtain cutoff values for determining the correct rates of discrimination between two groups in the testing data.
The ROC analyses on MDF values resulted in an area under the curve (AUC) of 0.90 (sensitivity 85%, specificity 84%) for T2w FLAIR, 0.92 (86%, 94%) for ADC, 0.96 (97%, 84%) for T1w, and 0.82 (78%, 75%) for T1w + Gd and correctly discriminated between the two groups in 93%, 100%, 93%, and 92% of cases in testing data, respectively. In the astrocytoma subgroup, AUCs were 0.92 (88%, 83%) for T2w FLAIR and 0.90 (92%, 74%) for T1w + Gd and correctly discriminated two groups in 100% and 92% of cases. The MDF outperformed all 9 of the histogram parameters.
Texture analysis on conventional preoperative MRI images can accurately predict early malignant transformation of LGGs, which may guide therapeutic planning.
• Texture analysis performed on MRI images can provide additional quantitative information that is invisible to human assessment. • Texture analysis based on conventional preoperative MR images can accurately predict early malignant transformation from low- to high-grade glioma. • Texture analysis is a clinically feasible technique that may provide an alternative and effective way of determining the likelihood of early malignant transformation and help guide therapeutic decisions.
磁共振成像（MRI）图像的纹理分析可以提供人类评估无法察觉的额外定量信息。本研究旨在评估术前常规 MRI 图像纹理分析预测低级别胶质瘤向高级别胶质瘤恶性转化的可行性，并比较其与直方图分析的效用。
本研究共纳入 68 例低级别胶质瘤（LGG）患者，其中 15 例发生恶性转化。患者被随机分为训练（60%）和测试（40%）集。对纹理分析进行分析，以获得训练和测试数据的最具鉴别力因子（MDF）值。在训练数据中对 MDF 值和 9 个直方图参数进行接收者操作特征（ROC）曲线分析，以获得确定测试数据中两组之间正确鉴别率的截断值。
MDF 值的 ROC 分析在 T2w FLAIR 中得出曲线下面积（AUC）为 0.90（敏感性 85%，特异性 84%），ADC 为 0.92（86%，94%），T1w 为 0.96（97%，84%），T1w + Gd 为 0.82（78%，75%），并在测试数据中分别正确鉴别出两组 93%、100%、93%和 92%的病例。在星形细胞瘤亚组中，T2w FLAIR 的 AUC 为 0.92（88%，83%），T1w + Gd 为 0.90（92%，74%），并在 100%和 92%的病例中正确鉴别出两组。MDF 优于所有 9 个直方图参数。
术前常规 MRI 图像的纹理分析可以准确预测 LGG 的早期恶性转化，这可能有助于指导治疗计划。
MRI 图像上的纹理分析可以提供人类评估无法察觉的额外定量信息。
基于常规术前 MR 图像的纹理分析可以准确预测低级别胶质瘤向高级别胶质瘤的早期恶性转化。
纹理分析是一种可行的临床技术，它可能提供一种替代和有效的方法来确定早期恶性转化的可能性，并有助于指导治疗决策。
