Diagnostic Accuracy of Centrally Restricted Diffusion in the Differentiation of Treatment-Related Necrosis from Tumor Recurrence in High-Grade Gliomas

BACKGROUND AND PURPOSE: Centrally restricted diffusion has been demonstrated in recurrent high-grade gliomas treated with bevacizumab. Our purpose was to assess the accuracy of centrally restricted diffusion in the diagnosis of radiation necrosis in high-grade gliomas not treated with bevacizumab. MATERIALS AND METHODS: In this prospective study, we enrolled patients with high-grade gliomas who developed a new ring-enhancing necrotic lesion and who underwent re-resection. The presence of a centrally restricted diffusion within the ring-enhancing lesion was assessed visually on diffusion trace images and by ADC measurements on 3T preoperative diffusion tensor examination. The percentage of tumor recurrence and radiation necrosis in each surgical specimen was defined histopathologically. The association between centrally restricted diffusion and radiation necrosis was assessed using the Fisher exact test. Differences in ADC and the ADC ratio between the groups were assessed via the Mann-Whitney U test, and receiver operating characteristic curve analysis was performed. RESULTS: Seventeen patients had re-resected ring-enhancing lesions: 8 cases of radiation necrosis and 9 cases of tumor recurrence. There was significant association between centrally restricted diffusion by visual assessment and radiation necrosis (P = .015) with a sensitivity of 75% and a specificity of 88.9%, a positive predictive value 85.7%, and a negative predictive value of 80% for the diagnosis of radiation necrosis. There was a statistically significant difference in the ADC and ADC ratio between radiation necrosis and tumor recurrence (P = .027). CONCLUSIONS: The presence of centrally restricted diffusion in a new ring-enhancing lesion might indicate radiation necrosis rather than tumor recurrence in high-grade gliomas previously treated with standard chemoradiation without bevacizumab.

T he appearance of a new ring-enhancing lesion after surgery and chemoradiation for high-grade glioma (HGG) can be seen in tumor recurrence and radiation necrosis (RN). A radiation-induced ring-enhancing lesion contains coagulative necrosis secondary to vascular endothelial injury and fibrinoid necrosis of small blood vessels. 1 On the other hand, the nonenhancing necrotic com-ponent in recurrent glioblastoma contains liquefactive necrosis. 2 Differentiation between the 2 remains difficult on conventional MR imaging. [3][4][5][6] Diffusion restriction has been reported in HGGs treated with bevacizumab 2 and in radiation necrosis of metastatic brain lesions. 7 Our purpose was to assess the accuracy of centrally restricted diffusion in the diagnosis of RN in HGG not treated with bevacizumab.

Patient Population
We consecutively enrolled patients with HGG who developed a new enhancing lesion on follow-up MR imaging after resection, radiation (60 Gy in 30 fractions of 2 Gy during 6 weeks), and temozolomide therapy, from December 2012 to December 2016. The patients were part of a prospective diagnostic accuracy study on dynamic contrastenhanced and DSC MR imaging for glioma recurrence, approved by our institutional research ethics board. Written informed consent was obtained from all the patients. For the assessment of central diffusion restriction in newly developed necrotic enhancing lesions, the inclusion criteria were the presence of a central area of necrosis within the newly enhancing lesion and surgical re-resection of this lesion. The exclusion criteria were an entirely solid newly enhancing lesion, bevacizumab therapy, and an operation Ͼ3 months after the study MR imaging. A short interval between the operation and the MR imaging was chosen to ensure that the pathology results were representative of the lesions visualized on the imaging study.

MR Image Interpretation and Data Collection
The ADC maps were coregistered to the axial T1-weighted postcontrast images using a commercially available software (Olea Sphere 1.3; Olea Medical, La Ciotat, France). Qualitative and quantitative assessments were performed by 1 neuroradiologist (N.Z.) blinded to the clinical information and histopathologic results. The presence of centrally restricted diffusion within part or all of the necrotic component was visually assessed in each ring-enhancing lesion and recorded as present or absent. Central diffusion restriction was defined as higher signal on trace images with lower signal on ADC compared with the rest of the lesion. For quantitative analysis, ADC values were calculated from DTI. Separate ROIs (Fig 1C) were manually placed within the necrotic component of each lesion excluding the enhancing wall, as well as within the enhancing component of the lesions, and the minimum ADC was recorded. The ROI size was kept between 20 and 30 mm 2 . ROIs were placed in the contralateral normal-appearing white matter. We calculated the following ADC ratios: ADC Necrosis Ratio ϭ ADC Central Necrosis/ADC White Matter; ADC Enhancement Ratio ϭ ADC Enhancing Component/ADC White Matter; and ADC Necrosis/Enhancement Ratio ϭ ADC Central Necrosis/ADC Enhancing Component.

Histopathologic Analysis
Following surgical resection, an experienced neuropathologist (J.W. or G.H.J.), blinded to the imaging findings, determined the presence or absence of viable tumor and/or radiation necrosis using the 2007 World Health Organization classification and estimated the percentage of viable tumor and radiation necrosis within the specimen. Visual assessment was performed under ϫ4 high-power-field microscopy to determine the approximate percentage area of radiation necrosis or tumor recurrence relative to the whole tissue area. The following criteria were used  for the diagnosis of radiation necrosis: the presence of coagulative necrosis, hyalinized vessels, and microcalcifications. For the diagnosis of recurrent tumor, we used the following criteria: the presence of increased cellularity and nuclear pleomorphism. Given that most lesions consisted of a mixture of tumor-and treatment-related changes in different proportions, a lesion was categorized as predominantly tumor recurrence or predominantly RN based on the component with the higher proportion in the specimen (Fig 2).

Statistical Analysis
The association between the presence of centrally restricted diffusion and pathologically proved radiation necrosis was assessed using the Fisher exact test, and the diagnostic accuracy of this finding was calculated using receiver operating characteristic curve analysis. Differences in the minimum ADC and ADC ratio of the necrotic component as well as the enhancing rim between radiation necrosis and tumor recurrence were assessed via the Mann-Whitney U test. A P value of .05 indicated statistical significance. All data were analyzed with MedCalc for Windows (Version 12; MedCalc Software, Mariakerke, Belgium).

RESULTS
Thirty-seven patients with HGG with new necrotic peripherally enhancing lesions were identified. Twenty-three patients had reresection of the identified lesions, and 14 had nonsurgical management. Three patients had an operation Ͼ3 months after the study MR imaging. Three patients were subsequently excluded on the basis of the pathology results: Two had equal proportions of necrosis and tumor, and 1 had hemorrhage and pus with no evidence of tumor recurrence or RN (Fig 3). Of note, the lesion excluded for presence of pus and blood product in the center of the lesion at the operation had central diffusion restriction on imaging.
There were 13 men and 4 women in the included patients. Average age was 55.9 Ϯ 10.3 years (range, 33.3-77.6 years). The average duration between the operation and the MR imaging examination was 25.8 Ϯ 23.99 days (range, 1-78 days).

Qualitative Assessment
Nine patients had predominant tumor recurrence, and 8 had predominant RN (Fig 3). The presence of centrally restricted diffusion in the necrotic lesions was significantly associated with the predominance of RN on histopathologic analysis (P ϭ .015) (Figs 1 and 4 and Table 1). Central diffusion restriction was seen in 6 patients with RN. From the 9 patients with predominant tumor recurrence, only 1 patient had central diffusion restriction. On histopathologic analysis, radiation necrosis comprised approximately 30% of this patient's surgical specimen. The presence of central diffusion restriction had a sensitivity of 75% and a specificity of 88.9%, with a positive predictive value of 85.7% and a negative predictive value of 80% for the diagnosis of radiation necrosis in a newly developed necrotic lesion after treatment of HGG. From the 20 patients who were excluded, 9 patients had central diffusion restriction on visual inspection.

Quantitative Assessment
There was a significant difference in the minimum ADC and ADC ratio in the necrotic component of the tumor recurrence and RN lesions (P ϭ .027) with a lower median ADC and ADC ratio in the RN group compared with the tumor recurrence group (Table  2). On the basis of the receiver operating characteristic curve analysis (area under the curve ϭ 0.82; 95% CI, 0.5-0.96; P ϭ .002), an ADC cutoff value of Յ1316 ϫ 10 Ϫ6 mm 2 /s provides 100% specificity for RN but only 55.6% sensitivity. Using a cutoff of Յ1197.03 ϫ 10 Ϫ6 mm 2 /s results in 80% specificity and 66.7% sensitivity for the identification of RN. Similarly, an ADC necrosis ratio cutoff value  of Յ1.78 provides 100% specificity and 55.6% sensitivity for the identification of RN (Fig 5). No significant difference was found in the ADC and ADC ratio of the enhancing components between tumor recurrence and RN (P ϭ .63 and P ϭ .85) ( Table 2). However, there was a significant difference in the necrotic-to-enhancing component ADC ratio (P ϭ .043), with a lower ratio in the RN group (Table 2). On the basis of the receiver operating characteristic curve analysis (area under the curve ϭ 0.79; 95% CI, 0.53-0.95; P ϭ .001), a cutoff value of Յ0.95 provides 88.9% sensitivity and 62.5% specificity for RN (Fig 5).

DISCUSSION
The results of the current study suggest that evaluation of diffusion restriction in the nonenhancing necrotic center rather than the peripheral enhancing component of ring-enhancing lesions developing after treatment of HGG may be a useful tool in differentiating RN from tumor recurrence. Central diffusion restriction is associated with a predominance of radiation necrosis on pathology, with a lower minimum ADC in the lesion center in RN compared with tumor recurrence.
Most of the previously published studies assessing the diagnostic accuracy of DWI in differentiating tumor recurrence and RN focused on assessment of diffusion restriction and ADC values in the enhancing component of the lesion with the expectation that the increased cellularity in tumor recurrence and increased extracellular space associated with cell death in RN would result in diffusion restriction and lower ADC values with tumor recurrence. [8][9][10] However, the results were not consistent. [11][12][13][14][15] A meta-analysis of the diagnostic accuracy of DWI showed moderate diagnostic performance and recommended against the use of DWI alone in differentiating RN and tumor recurrence. 16 Like the results of the current study, evaluation of the central zone of radiation necrosis and malignant glioma in rat models identified a pattern of central diffusion restriction in radiation necrosis with significantly lower ADC values compared with malignant glioma. The diffusion restriction in the central zone of radiation necrosis corresponded to coagulative necrosis on histopathology. 12 Asao et al 17 reported that most of the radiation necrosis lesions in their patient population, consisting predominantly of astrocytic neoplasms, had heterogeneous signal intensity on DWI with hyperintensities within the lesions. There are also a few reported cases in the literature with diffusion restriction in the center of radiation necrosis, simulating central diffusion restriction in the abscess. 18,19 In fact, one of the cases that showed central diffusion restriction in our series was found to have infection and hemorrhage within the lesion center at the operation.
Nonenhancing diffusion restriction has also been reported in bevacizumab-treated HGG 20,21 as well as bevacizumab-treated brain metastasis. 22 Nguyen et al 2 demonstrated that progressively expanding diffusion restriction in HGG treated with bevacizumab pathologically represents coagulative necrosis surrounded by viable tumor cells. They also showed lower ADC in the diffusion-restricted necrosis compared with the hypercellular nonenhancing tumor.
In their series of treated metastatic lesions, Cha et al 7 described a 3-layer pattern of radiation necrosis with a central core of lique-   fied necrosis showing increased ADC, a nonenhancing middle layer of coagulative necrosis with low ADC, and an outer layer of elevated CBV representing a zone of active inflammation. Our findings of diffusion restriction centrally in the nonenhancing component along the inner aspect of the enhancing ring likely correspond to the middle layer in their 3-layer pattern. They also showed lower minimum ADC values in RN compared with tumor recurrence, comparable with the current study. 7 In this study, we included only surgically proved cases, leading to the small number of patients in this series, which is the main limitation to the study. Although only surgically proved cases were included, the pathologic samples were not limited to the areas of diffusion restriction, another limitation to this study. While we classified the lesions into predominant tumor recurrence or RN on the basis of the component with the higher proportion in the surgical specimen, the proportion of tumor in the cases classified as predominant RN in this series ranged from 5% to 20%. However, it can still be argued from a pathologic point of view that the presence of any amount of tumor cells in the specimen should be considered evidence of tumor recurrence. Future studies with image-guided stereotactic biopsy of the regions of centrally restricted diffusion would be of great value in confirming the current findings. The results of this study cannot be generalized to patients treated with bevacizumab or lesions with no necrosis.

CONCLUSIONS
This study shows that the presence of central nonenhancing restricted diffusion in a newly necrotic ring-enhancing lesion might indicate RN rather than tumor recurrence in HGG previously treated with standard chemotherapy without bevacizumab.