The aim of this retrospective study was to qualitatively and quantitatively compare the perfusion values obtained from ASL with DSC MRI in patients treated for malignant brain tumor and to evaluate the response to treatment. It was observed that ASL nCBF values correlated very well with DSC nCBV and perfectly with DSC nCBF. It was determined that the ASL nCBF, DSC nCBV and DSC nCBF values of RRT were statistically and significantly higher than the ASL nCBF, DSC nCBV and DSC nCBF values of TRC.
Histopathological confirmation with biopsy is often required for definitive diagnosis of brain tumors. However, biopsy is invasive and challenging for some areas [21]. The treatment response criteria of the RANO study group guide the clinical examination and can help in the evaluation of perfusion maps [22]. Recently, with developing MRI techniques, new non-invasive imaging methods have been used to guide treatment plans by using physiological features of different tumors such as cellularity, oxygenation, vascularity and microstructure [23,24,25]. In DSC, which is based on T2-weighted imaging, a parallel imaging scan is required to cover the entire brain [26]. DSC is routinely used to differentiate low- and high-grade tumors and tumor recurrence and treatment-related changes. DSC provides perfusion parameters that correlate with histological structure such as rCBV and rCBF. However, its disadvantages include that it requires contrast injection and may cause a decrease in CBV values due to contrast material extravasation. ASL provides absolute CBF values by using an endogenous tracker [27, 28]. ASL perfusion is a method that is developing day by day and can obtain absolute CBF value noninvasively and without the use of contrast [24].
In a study conducted on 30 patients with a history of treatment for GBM (1,5 T, pASL), ASL nCBF values were determined to be higher than DSC nCBV values. No difference was observed between patients with and without tumor recurrence. Researchers have suggested that since the single-PLD ASL method was used, it may cause bias in the evaluation [16]. Ye et al. [29] documented that in patients treated for glioma, ASL nCBF values were higher than DSC nCBV and that there was a significant difference in both parameters between the recurrent gliomas and radiation necrosis groups. On the contrary, in the study conducted by Manning et al. [30], it was found that ASL nCBF values were smaller than DSC nCBV, but similar to DSC nCBF, and the values in the progression group were statistically significantly higher for all three parameters. In ASL examination of 26 patients with a history of treatment for GBM (1,5 T, pASL), Seeger et al. [31] did not observe any difference between the group with tumor recurrence and stable disease (2.41 ± 1.3 vs 1.66 ± 0.5, respectively). In the same study, it was reported that DSC nCBF and DSC nCBV values were higher than ASL nCBF, and that there was a significant difference in DSC-related parameters between tumor recurrence and radiation necrosis groups. Jovanovic et al. [32] reported that DSC nCBV values were similar to other studies and ASL nCBF values were lower compared to other studies.
In our study, uncorrected DSC nCBV values were found to be higher than both ASL nCBF and uncorrected DSC nCBF and previous study results. In the study conducted by Arisawa et al. [33] using histogram analysis in benign and malignant glial tumors, it was stated that the maximum values were found to be higher in DSC nCBV, and they suggested that this may be due to the inability to completely remove pixels belonging to vascular structures in the histogram analysis. Moreover, in a study conducted by Hashido et al. [28], using radiomics for comparison, similar results were obtained. Therefore, it was thought that the use of manual measurement in our study did not affect the results. Previous studies have suggested that taking the reference value from the white matter and low spatial resolution may cause the ASL nCBF measurement to be smaller than normal [34, 35]. In our study, manual measurements, not using corrected values, reference values being taken from white matter, and heterogeneous malignant tumor histologies may be responsible for these results. Also, in our study, it was observed that the DSC nCBV, DSC nCBF and ASL nCBF values of RRT patients were significantly higher than those of TRC patients. These results of our study confirm the results of previous studies.
In previous studies evaluating brain tumor perfusion, the correlation of diagnostic comparison parameters was also evaluated. Ata et al. (1,5 T) reported an excellent correlation (r = 0.91, P < 0.001) between DSC nCBF and nCBV values [36]. In the linear correlation analysis of Jovanovic et al., a very good correlation was found between ASL nCBF and DSC nCBV (r: 0.733) [26]. Lavrova et al. [37] determined that there was a moderate correlation (r: 0.60–0.67) between ASL nCBF and uncorrected DSC nCBV, and a good correlation (r: 0.72–0.78) between ASL CBF and corrected nCBV. Researchers also observed a strong correlation in contrasted gliomas (r: 0.65–0.80) and low correlation in non-contrasted gliomas (r: 0.58–0.73) and brain metastases (r: 0.14–0.40) when they looked at disease specificities [37]. Xu et al. [38] found an excellent correlation between DSC nCBF-nCBV, ASL nCBF-DSC nCBF and ASL nCBF-DSC nCBV in the quantitative evaluation of patients receiving treatment for glial tumors. Rau et al. [9] determined a moderate correlation between DSC nCBF and nCBV in high-grade gliomas. However, they did not observe a statistically significant correlation between ASL nCBF and DSC parameters. In another study, it was determined that there was a close linear correlation between normalized perfusion values obtained from ASL and DSC [39]. In our study, it was observed that uncorrected DSC nCBF showed excellent correlation with both uncorrected DSC nCBV and ASL nCBF. Additionally, uncorrected DSC nCBV showed a very good correlation with ASL nCBF. The different correlation rates determined in the studies may be related to the method used, the number of patients and the histological structure of the tumor.
ROC curve analysis was performed to evaluate the diagnostic performance of the parameters in the applied methods. Jovanovic et al. [32] determined that when the ROC curve was used, DSC nCBV had 100% sensitivity and specificity (AUC = 1.000; p < 0.001) when the cut-off value was 2.89. They also observed that ASL had 100% sensitivity and 73.7% specificity when the cut-off value was 0.995, and 92.3% sensitivity and 92.9% specificity when the cut-off value was 1.02 (AUC = 0.967; p < 0.001). In another study, the cut-off value in ROC analysis was determined as 2.18 for ASL nCBF (84.6% specificity, 53.9% sensitivity, AUC: 69%), and the cut-off value for DSC nCBV was determined as 2.24 (84.6% specificity, 77.3% sensitivity, AUC: 80%) [31]. In their ROC curve analysis, Xu et al. [38] found the cut-off values to be 1.11 (AUC 0.88, sensitivity 100%, specificity 75%) for ASL nCBF, 2.36 (AUC 0.86, sensitivity 70%, specificity 91%) for DSC nCBF and 3.64 (AUC 0.82, sensitivity 58%, specificity 100%) for DSC nCBV. Lavrova et al. [37] measured AUC values of 0.73–0.80 for ASL nCBF and DSC nCBV. While researchers determined AUC values of 0.78 and 0.77–0.80 for ASL nCBF and DSC nCBV in enhancing gliomas, they found them to be 0.72 for ASL nCBF and 0.87–0.93 for DSC nCBV in brain metastases. Özsunar et al. [16] determined that the ASL technique exhibited higher sensitivity and specificity compared to DSC in detecting recurrent tumors (88% sensitivity-89% specificity vs. 86% sensitivity-70% specificity, respectively). Rau et al. [9] found DSC nCBV to be superior to DSC nCBF and ASL nCBF in predicting recurrence (AUC: 0.71, AUC: 0.59 and AUC: 0.58, respectively). In another study, AUC values for ASL nCBF, DSC nCBF and DSC nCBV were determined as 0.95, 0.86 and 0.89, respectively [30]. In the evaluation of patients receiving treatment for GBM, Choi et al. [40] determined that the diagnostic accuracy of DSC when used alone was 75.8%, and with the addition of ASL perfusion, the accuracy rate increased to 88.7%.
In our study, according to the ROC curve analysis results, the cut-off values were determined as 1.211 (93% sensitivity, 82% specificity) for uncorrected DSC nCBV, 0.896 for (93% sensitivity, 82% specificity) uncorrected DSC nCBF and 0.829 (78% sensitivity, 75% specificity) for ASL nCBF. ASL nCBF, DSC nCBF and DSC nCBV cut-off values in our study were observed to be lower than previous studies [28, 30,31,32,33,34,35,36,37,38]. Additionally, the detection of lower cut-off values in all parameters (ASL nCBF, DSC nCBV, DSC nCBF) compared to the literature has been attributed to differences in the number of patients, the larger size of the group with TBD, the lower minimum values compared to the literature, the lower measurements taken from the operation cavity without recurrence, or the use of low-area ROI in these regions, resulting in lower minimum values compared to the literature.
In our study, the sensitivity and specificity values for uncorrected DSC nCBV were generally found to be higher when compared to the corrected DSC nCBV values reported in the literature [28, 30,31,32,33,34,35,36,37,38]. As for uncorrected DSC nCBF, the sensitivity in our study was higher compared to the literature, while the specificity was similar or lower.and corrected DSC nCBF [30, 38]. It has been considered that the higher sensitivity detected in DSC nCBV and DSC nCBF compared to the literature may also be attributed to the use of uncorrected values. In comparison to the literature, the sensitivity of ASL nCBF is lower, but its specificity is higher than some studies [16, 31, 32, 38]. Our ASL imaging techniques are similar to only one study among these [38]. In a study comparing perfusions in primary and secondary brain tumors [37], no significant correlation was found between ASL and DSC perfusions in metastases, and a decrease in ASL nCBF AUC values compared to corrected DSC nCBV was observed in cases with brain metastases. The discrepancy in sensitivity and specificity between our study and Xu et al. [38] was thought to be possibly due to the high number of metastases in our study.
In our study, AUC values for ASL nCBF, uncorrected DSC nCBF and uncorrected DSC nCBV were determined as 0.84, 0.95 and 0.95, respectively. Our study found that diagnostically, all three parameters can be used in the differentiation between TBD (treatment-related changes) and NRT (non-residual tumor), but uncorrected DSC nCBV and DSC nCBF are superior. Similarly to literature findings [9, 32, 37], studies have correlated with our results, showing higher AUC values for DSC nCBV compared to ASL nCBF. Some studies have reported higher AUC values for ASL nCBF [30, 38]. However, Xu et al. did not find the difference in AUC values to be statistically significant [38]. Manning et al. conducted a study similar to ours (3 T, pCASL), with higher AUC values for ASL nCBF [30]. The differences could be attributed to Manning et al. having only GBM operative cases, a lower number of pseudoprogression cases (n:7), and a smaller overall patient population (n:32), as well as the use of different ASL reading techniques [30].
In our study, in qualitative assessment, it was observed that two readers could evaluate DSC CBV more consistently with each other and with the results compared to the other two parameters. The sensitivity was found to be higher compared to other techniques. In ASL CBF and DSC CBF maps, higher specificity was observed. In contrast to previous studies in the literature [16, 36], our study found that the specificity of ASL CBF may be higher than its sensitivity. Jarnum et al., compared the 3D pCASL method with the DSC method in terms of susceptibility artifacts and noted that the artifacts were statistically significantly less in ASL. In this case, they suggested that using FSE instead of GRE EPI could be effective [41]. In the study conducted by Manning et al., the 3D pCASL technique was employed, and signal acquisition was performed using the spiral Fast Spin Echo (FSE) method. This study also indicates that sensitivity artifacts were less in ASL compared to DSC [30]. The difference in sensitivity in our qualitative assessment compared to the literature may be attributed to the unclear efficacy of ASL in metastatic lesions, the continued inadequacy of SNR despite the preference for patients with tumor could be evaluated GRASE pCASL, and the potential errors in labeling due to susceptibility artifacts in treated cases. The higher specificity compared to sensitivity, contrary to studies in the literature, may be linked to the sensitivity to artifacts in the GRASE reading technique, resulting in the elimination of false positives, as suggested by Maral et al. [42]. Similarly, the more pronounced susceptibility artifacts in the GRASE method compared to FSE may explain the differences between our study and the quantitative evaluations and specificity-sensitivity levels in the literature [30]. DSC CBF also showed similar sensitivity and specificity in visual assessment compared to ASL CBF.
ROI measurements, an operator-dependent method, are used in DSC-MRI data analysis [36]. Tumoral CBV measurement remains operator dependent and highly subjective [43]. In a multicenter study, reproducibility and repeatability were reported to be less than 10 and 50%, respectively [44]. A good review process can help evaluate perfusion measurements more objectively. For this reason, it has also been recommended that DSC-MRI data be evaluated by two experienced observers [45]. In our study, DSC and ASL results were evaluated separately by two observers. Interobserver agreement was good for ASL CBF (0.714), good for DSC CBF (0.790), and excellent for DSC CBV (0.822). Additionally, intra-observer agreement was generally good.
The strengths of our study were that the patient group is larger compared to previous studies, there were different groups with numerically similar distribution in the lesion groups, patients with tumor could be evaluated with 3D GRASE pCASL, and there was no bias in favor of ASL in visual evaluation due to sufficient time between readings. There are some limitations in our study. First, corrected values were not used in DSC measurements and patient results were obtained through follow-up and control evaluations rather than histopathologically. Second, the treatment-related changes group was not divided into radiation necrosis/pseudoprogression and stable disease during follow-up. Third, the values taken from the operation cavity were found to be low in cases that underwent surgery and did not have significant radiation necrosis. Lastly, the limitation is that the ROI sizes used are not standard due to different lesion sizes and the study was conducted retrospectively.
In conclusion, we observed that the DSC nCBF value was perfectly correlated with DSC nCBV and ASL nCBF, and DSC nCBV was very well correlated with ASL nCBF. DSC nCBV, DSC nCBF and ASL nCBF values of RRT were significantly higher than those of TRC. It was determined that all three parameters DSC nCBV, DSC nCBF and ASL nCBF were usable from a diagnostic perspective, but the highest rate belonged to DSC nCBV. It was also observed that DSC CBV had higher sensitivity and ASL CBF had higher specificity. In visual evaluation, DSC CBV is the test that can be evaluated best among all three parameters. In the qualitative evaluation, ASL CBF was found to have low sensitivity and high specificity. Prospective studies with larger patient participation will make important contributions.
Add Comment