PURPOSE To use 3D pseudocontinuous arterial spin labeling (3D PCASL) and dynamic susceptibility contrast-enhanced (DSC) perfusion MRI to differentiate progressive disease from pseudoprogression in patients with glioblastoma (GBM). METHODS Thirty-two patients with GBM who developed progressively enhancing lesions within the radiation field following resection and chemoradiation were included in this retrospective, single-institution study. The updated modified RANO criteria were used to establish progressive disease or pseudoprogression. Following 3D PCASL and DSC MR imaging, perfusion parameter estimates of cerebral blood flow (ASL-nCBF and DSC-nrCBF) and cerebral blood volume (DSC-nrCBV) were calculated. Additionally, contrast enhanced volumes were measured. Mann-Whitney U tests were used to compare groups. Linear discriminant analysis (LDA) and area under receiver operator characteristic curve (AUC) analyses were used to evaluate performance of each perfusion parameter and to determine optimal cut-off points. RESULTS All perfusion parameter measurements were higher in patients with progressive disease (mean, 95% CI ASL-nCBF 2.48, [2.03, 2.93]; DSC-nrCBF = 2.27, [1.85, 2.69]; DSC-nrCBV = 3.51, [2.37, 4.66]) compared to pseudoprogression (mean, 95% CI ASL-nCBF 0.99, [0.47, 1.52]; DSC-nrCBF = 1.05, [0.36, 1.74]; DSC-nCBV = 1.19, [0.34, 2.05]), and findings were significant at the p < 0.0125 level (p = 0.001, 0.003, 0.002; effect size: Cohen's d = 1.48, 1.27, and 0.92). Contrast enhanced volumes were not significantly different between groups (p > 0.447). All perfusion parameters demonstrated high AUC (0.954 for ASL-nCBF, 0.867 for DSC-nrCBF, and 0.891 for DSC-nrCBV), however, ASL-nCBF demonstrated the highest AUC and misclassified the fewest cases (N = 6). Lesions correctly classified by ASL but misclassified by DSC were located along the skull base or adjacent to large resection cavities with residual blood products, at areas of increased susceptibility. CONCLUSION Both 3D PCASL and DSC perfusion MRI techniques have nearly equivalent performance for the differentiation of progressive disease from pseudoprogression in patients with GBM. However, 3D PCASL is less sensitive to susceptibility artifact and may allow for improved classification in select cases.

中文翻译：

---

胶质母细胞瘤患者使用3D PCASL和DSC灌注MRI区分进行性疾病和假性疾病。

目的使用3D伪连续动脉自旋标记（3D PCASL）和动态磁化率对比增强（DSC）灌注MRI来区分胶质母细胞瘤（GBM）患者的疾病进展与假进展。方法该回顾性单机构研究纳入了32例GBM患者，这些患者在切除和化学放疗后在放射区域内逐渐增强了病变。更新后的经修改的RANO标准用于确定疾病进展或假进展。在3D PCASL和DSC MR成像之后，计算了脑血流量（ASL-nCBF和DSC-nrCBF）和脑血容量（DSC-nrCBV）的灌注参数估计值。另外，测量对比增强的体积。使用Mann-Whitney U检验比较各组。线性判别分析（LDA）和接收者操作员特征曲线下的面积（AUC）分析用于评估每个灌注参数的性能并确定最佳截止点。结果进行性疾病患者的所有灌注参数测量值均较高（平均值，95％CI ASL-nCBF 2.48，[2.03，2.93]; DSC-nrCBF = 2.27，[1.85，2.69]; DSC-nrCBV = 3.51，[2.37， 4.66]）与伪进行比较（平均值，95％CI ASL-nCBF 0.99，[0.47，1.52]； DSC-nrCBF = 1.05，[0.36，1.74]； DSC-nCBV = 1.19，[0.34，2.05]）和发现在p <0.0125的水平上具有显着性（p = 0.001、0.003、0.002；效应大小：Cohen d = 1.48、1.27和0.92）。两组之间的对比增强量没有显着差异（p> 0.447）。所有灌注参数均显示出较高的AUC（ASL-nCBF为0.954，DSC-nrCBF为0.867，DSC-nrCBV为0.891），但ASL-nCBF的AUC最高，而错误分类的病例最少（N = 6）。通过ASL正确分类但通过DSC错误分类的病变位于颅骨底或邻近具有大量血液残留的切除腔，位于易感性增加的区域。结论3D PCASL和DSC灌注MRI技术在区分GBM患者的进行性疾病和假性进展方面具有几乎相同的性能。但是，3D PCASL对磁化伪影的敏感性较低，并且可能允许在某些情况下改进分类。通过ASL正确分类但通过DSC错误分类的病变位于颅骨根部或邻近具有大量血液残留的切除腔，位于易感性增加的区域。结论3D PCASL和DSC灌注MRI技术在区分GBM患者的进行性疾病和假性进展方面具有几乎相同的性能。但是，3D PCASL对磁化伪影的敏感性较低，并且可能允许在某些情况下改进分类。通过ASL正确分类但通过DSC错误分类的病变位于颅骨根部或邻近具有大量血液残留的切除腔，位于易感性增加的区域。结论3D PCASL和DSC灌注MRI技术在区分GBM患者的进行性疾病和假性进展方面具有几乎相同的性能。但是，3D PCASL对磁化伪影的敏感性较低，并且可以在某些情况下改进分类。