Histogram analysis based on 3D-amide proton transfer weighted and apparent diffusion coefficient imaging in predicting ATRX mutation in IDH-mutant WHO grading 2/3 gliomas
10.3760/cma.j.cn115354-20240530-00317
- VernacularTitle:3D-APTw及ADC直方图分析预测 IDH突变型WHO 2/3级胶质瘤中 ATRX基因突变的价值研究
- Author:
Xia ZOU
1
;
Xinran YAN
;
Yuxin LI
;
Yaoming QU
;
Haitao WEN
;
Andong MA
;
Shizhong ZHANG
;
Zhibo WEN
Author Information
1. 南方医科大学珠江医院影像诊断科，广州 510282
- Keywords:
Glioma;
Amide proton transfer weighted imaging;
Apparent diffusion coefficient;
Histogram analysis;
Alpha-thalassemia/mental retardation syndrome X-linked
- From:
Chinese Journal of Neuromedicine
2024;23(7):659-668
- CountryChina
- Language:Chinese
-
Abstract:
Objective:To evaluate the role of histogram analysis based on amide proton transfer weighted (APTw) and apparent diffusion coefficient (ADC) imaging in predicting alpha-thalassemia/mental retardation syndrome X-linked ( ATRX) mutation in isocitrate dehydrogenase ( IDH)-mutant WHO grading 2/3 gliomas. Methods:Seventy-eight patients with IDH-mutant WHO grading 2/3 gliomas, admitted to and confirmed by surgical pathology in Department of Functional Neurosurgery, Neurosurgery Center, Zhujiang Hospital, Southern Medical University from June 2017 to October 2023, including 52 with ATRX wild and 26 with ATRX mutant-type, were selected. Preoperative 3D-APTw and ADC imaging data were collected; after post-processing, the lesions were segmented using lesion outlining method based on inclusion of peri-tumor edema and lesion outlining method based on tumor entity, respectively; after that, the histogram features (the 10 th percentile, 90 th percentile, maximum, mean, median, minimum, skewness, kurtosis, entropy, range, uniformity, and variance) were extracted from 3D-APTw and ADC imaging, respectively. Univariate Logistic regression was used to compare the differences in histogram features between patients in the ATRX mutant group and ATRX wild-type group, and multivariate Logistic regression was used to screen the independent predictors for ATRX mutation (a Logistic regression prediction model was constructed). Predictive values of independent predictors and Logistic regression prediction models in ATRX mutation were evaluated by receiver operating characteristic (ROC) curve. Results:(1) With lesion outlining method based on inclusion of peri-tumor edema, univariate analysis indicated significant difference between ATRX mutant group and ATRX wild-type group in 9 histogram features: relative 3D-APTw minimum, 3D-APTw skewness, relative ADC 90 th percentile, relative ADC mean, relative ADC median, ADC kurtosis, ADC skewness, ADC uniformity, and ADC entropy ( P<0.05). With lesion outlining method based on tumor entity, univariate analysis indicated significant difference between ATRX mutant group and ATRX wild-type group in 9 histogram features: relative 3D-APTw 90 th percentile, 3D-APTw skewness, relative ADC 90 th percentile, relative ADC mean, relative ADC median, ADC kurtosis, ADC skewness, ADC uniformity and ADC entropy ( P<0.05). (2) With lesion outlining method based on inclusion of peri-tumor edema, multivariate Logistic regression showed that 3D-APTw skewness and ADC kurtosis were the independent predictor for ATRX mutation in IDH mutant WHO grading 2/3 glioma patients ( OR=0.168, 95% CI: 0.034-0.800, P=0.025; OR=0.508, 95% CI: 0.319-0.807, P=0.004). The constructed Logistic regression prediction model was P(Y=1|X)=1/1+e -(1.827-1.785×3D-APTw skewness-0.678×ADC kurtosis). With lesion outlining method based on tumor entity, multivariate Logistic regression showed that 3D-APTw skewness and ADC kurtosis were independent predictors for ATRX mutation in IDH mutant WHO grading 2/3 glioma patients ( OR=0.164, 95% CI: 0.034-0.791, P=0.024; OR=0.496, 95% CI: 0.312-0.788, P=0.003); the constructed Logistic regression prediction model was P(Y=1|X)=1/1+e -(1.585-1.810×3D-APTw skewness-0.702×ADC kurtosis). (3) ROC curve analysis showed that, with lesion outlining method based on inclusion of peri-tumor edema, area under ROC curve (AUC) of 3D-APTw skewness and ADC kurtosis was 0.725 (95% CI: 0.608-0.842, P=0.001) and 0.794 (95% CI: 0.685-0.904), respectively ( P<0.001); AUC of Logistic regression prediction model was 0.836 (95% CI: 0.729-0.942, P<0.001), and its sensitivity and specificity were 73.10% and 90.40% when the best threshold was 0.505. ROC curve showed that, with lesion outlining method based on tumor entity, AUC of 3D-APTw skewness and ADC kurtosis was 0.705 (95% CI: 0.587-0.823, P=0.003) and 0.808 (95% CI: 0.704-0.913), respectively ( P<0.001); AUC of Logistic regression prediction model was 0.844 (95% CI: 0.739-0.949, P<0.001), and its sensitivity and specificity were 84.60% and 80.80% when the best threshold was 0.399. Conclusion:Histogram analysis based on 3D-APTw and ADC imaging can predict ATRX mutation in IDH mutant WHO grading 2/3 gliomas to a certain extent.
