Artikel
Using BOLD cerebrovascular reactivity to discriminate glioblastoma versus radionecrosis
Anwendung von BOLD zerebrovaskulärer Reaktivität zur Unterscheidung zwischen Radionekrose und Rezidiv vom Gliobastoma
Suche in Medline nach
Autoren
Veröffentlicht: | 8. Mai 2019 |
---|
Gliederung
Text
Objective: We aim to improve the differential diagnosis of post-treatment contrast-enhancing lesions. To recognize recurrent glioblastoma after treatment, we investigate if relevant CVR differences exist between newly diagnosed brain glioblastomas and radionecrosis. For this purpose, we utilized blood oxygenation-level-dependent functional MRI (BOLD-fMRI) to study CVR.
Methods: We identified eight patients with diagnosed radionecrosis and eight age/gender matched patients with de novo diagnosed glioblastomas. BOLD-CVR was acquired on a 3-Tesla MRI scanner. We performed an analysis of intralesional (as identified as T1-weighted imaging contrast-enhancing lesion) and perilesional BOLD-CVR in the two groups for comparison. The CVR data were fitted with a sigmoidal curve. We performed a logistic regression analysis to obtain the predicted probabilities of group classification. The goodness of fit of the logistic regression was evaluated with a receiver operating characteristic (ROC) curve analysis.
Results: Intralesional CVR values were lower in radionecrosis than in glioblastoma patients (0.0009±0.06 vs 0.06±0.05; p=0.04). The ROC curve showed a good ability of intralesional BOLD-CVR do discriminate the two groups (sensitivity: 87,5%, specificity 75%, AUC (95%-confidence-interval): 0.81 (0.592,1.0)). Perilesionally, a faster recovery – from the lesion towards the outside – of the CVR values was observed in the radionecrosis group compared to glioblastomas, which showed also perilesional CVR impairment 3 cm away from the lesion. In radionecrosis cases, the CVR normalize – in average – at 2 cm away from the lesion. Using the parameters of the sigmoidal fit, the classification of the pathology could be performed with a ROC area under the curve of 0.95 with a 95%-confidence-interval of 0.85–1.0).
Conclusion: Recognizing different hemodynamic features both inside and around the lesion in “pure” glioblastomas versus radionecrosis delivers a scheme that, if confirmed by further investigations, can offer another tool to help in the differential diagnosis of recurrent glioblastoma after radiotherapy. However, due to the presence of mixed radionecrosis-recurrences observed frequently in the clinical practice, the small cohorts included in our study, and the possible bias in identifying radionecrosis patients, further confirmation – in form of blinded prospective studies – of this diagnostic approach is required before solid conclusion on its clinical applicability are provided.
Figure 1 [Fig. 1]
Figure 2 [Fig. 2]