Xiaoxin Li1, Yanwei Miao2, Liang Han1, Junyi Dong1, Yan Guo3, Yuqing Shang4, Lizhi Xie5, Qingwei Song1, Ailian Liu1. 1. Department of Radiology, the First Affiliated Hospital of Dalian Medical University, Dalian, Liaoning, China. 2. Department of Radiology, the First Affiliated Hospital of Dalian Medical University, Dalian, Liaoning, China. Electronic address: ywmiao716@163.com. 3. Life Science, GE Healthcare, Shenyang, Liaoning, China. 4. Department of Chronic Disease Epidemiology, Yale School of Public Health, Yale University, New Haven, Connecticut, USA. 5. GE Healthcare, MR Research China, Beijing, China.
Abstract
PURPOSE: To evaluate the application of conventional MRI histogram analysis based on the whole tumor measurement on assessing meningioma grading. MATERIALS AND METHODS: This retrospective study was approved by the institutional review board. A total amount of 90 patients with meningioma were enrolled and the preoperative MRI of them were analyzed. To be specific, the patient group were consisted of 45 patients with grade I, 38 with grade II, and 7 with grade III meningioma. Grade I meningioma is classified as low grade meningioma (LGM), whereas Grade II and III meningioma were combined and classified as high grade meningioma (HGM). ROIs were drawn along the edge of the tumor on each section of T1WI, T2WI, and contrasted T1WI. 3D ROI signal intensity histogram and all its parameters were obtained. Independent t-test and Kruskal-Wallis test were used for comparison between two groups. Univariate logistic regression analysis and Spearman's correlation analysis were used to screen for the parameters with high predictive efficiency, while multivariate logistic regression analysis was used to determine the optimal model for the classification of meningioma. RESULTS: There were significant differences observed between HGM and LGM groups regarding to histogram volume count, uniformity of three sequences, range of T1WI and T2WI, kurtosis, standard deviation, variance, max intensity of T2WI, skewness, mean deviation, minimum intensity, mean value, the 5th percentile, the 10th percentile, the 25th percentile, the 50th percentile, the 75th percentile, and the 90th percentile of contrasted T1WI. Volume count and uniformity were high predictive parameters in distinguishing HGM from LGM. Logistic regression model included contrasted T1WI histogram parameters (i.e. minimum intensity, volume count, skewness, uniformity, and the 75th percentile) showed the best diagnostic efficiency for meningioma grade, with a sensitivity and specificity of 83.9% and 77.4% (AUC = 0.834, cutoff value = 0.413), respectively. The optimal model was achieved with a sensitivity of 71.4% and a specificity of 78.6% in the test set (AUC = 0.791, cutoff value = 0.413). CONCLUSIONS: Histogram analysis of conventional MRI based on 3D tumor measurement can be applied in the assessment of meningioma grading in clinical.
PURPOSE: To evaluate the application of conventional MRI histogram analysis based on the whole tumor measurement on assessing meningioma grading. MATERIALS AND METHODS: This retrospective study was approved by the institutional review board. A total amount of 90 patients with meningioma were enrolled and the preoperative MRI of them were analyzed. To be specific, the patient group were consisted of 45 patients with grade I, 38 with grade II, and 7 with grade III meningioma. Grade I meningioma is classified as low grade meningioma (LGM), whereas Grade II and III meningioma were combined and classified as high grade meningioma (HGM). ROIs were drawn along the edge of the tumor on each section of T1WI, T2WI, and contrasted T1WI. 3D ROI signal intensity histogram and all its parameters were obtained. Independent t-test and Kruskal-Wallis test were used for comparison between two groups. Univariate logistic regression analysis and Spearman's correlation analysis were used to screen for the parameters with high predictive efficiency, while multivariate logistic regression analysis was used to determine the optimal model for the classification of meningioma. RESULTS: There were significant differences observed between HGM and LGM groups regarding to histogram volume count, uniformity of three sequences, range of T1WI and T2WI, kurtosis, standard deviation, variance, max intensity of T2WI, skewness, mean deviation, minimum intensity, mean value, the 5th percentile, the 10th percentile, the 25th percentile, the 50th percentile, the 75th percentile, and the 90th percentile of contrasted T1WI. Volume count and uniformity were high predictive parameters in distinguishing HGM from LGM. Logistic regression model included contrasted T1WI histogram parameters (i.e. minimum intensity, volume count, skewness, uniformity, and the 75th percentile) showed the best diagnostic efficiency for meningioma grade, with a sensitivity and specificity of 83.9% and 77.4% (AUC = 0.834, cutoff value = 0.413), respectively. The optimal model was achieved with a sensitivity of 71.4% and a specificity of 78.6% in the test set (AUC = 0.791, cutoff value = 0.413). CONCLUSIONS: Histogram analysis of conventional MRI based on 3D tumor measurement can be applied in the assessment of meningioma grading in clinical.