Reliability of dynamic susceptibility contrast perfusion metrics in pre- and post-treatment glioma

In neuro-oncology, dynamic susceptibility contrast magnetic resonance imaging (DSC-MRI) can play an important role in non-invasive tumor grading, prediction of treatment response [1]. In the diagnostic work-up, DSC-MRI can be used for characterisation of glioma genotype, as it is known that genetic differences in glioma subtypes correlate with the glioma vasculature. This results in specific perfusion profiles of different subtypes of glioma, as described in a recent review and meta-analysis of our group [2]. However, DSC-MRI has a major role to play in the post-treatment setting, especially with regard to the differentiation between true early tumor progression and pseudoprogression. Due to the infiltrative nature of gliomas, total resection is almost impossible and residual glioma cells can further infiltrate the brain tissue surrounding the resection cavity. This results in an enhancing lesion on pre- and post-contrast T1-weighted images. Although these characteristics are generally reliable to assess residual or recurring neoplasm, an important imaging dilemma exist. Different treatment methods (i.e., chemotherapy and radiation) increase the permeability of the vascular walls which leads to contrast-leakage. This leakage can also be observed as a new enhancing lesion, though is called pseudoprogression. Therefore, differentiation between early tumor progression and pseudoprogression is poor with conventional MR images alone [3]. It has been reported that implementation of either DSC-MRI in routine radiological follow-up of glioblastoma can aid the detection of tumour recurrence [4], although varying practices complicate further clinical implementation [5‐7].

The technique of DSC-MRI relies on the susceptibility induced signal loss on T2*-weighted sequences, resulting from the passage of a bolus of gadolinium-based contrast agent. The most commonly used DSC parameter is Cerebral Blood Volume (CBV) which can be estimated [8, 9] and computed [10] based on the negative enhancement integral. Other parameters include Cerebral Blood Flow (CBF), Mean Transit Time (MTT) and Time To Peak (TTP) (Fig. 1). The estimated value of the area under the attenuation curve is proportional to the CBV but does not yield an absolute measurement. Therefore, the measurement is expressed relative to a standard reference, usually the contralateral white matter (relative CBV ratio: rCBV ratio) [11]. Overall, the rCBV ratio is an indicator of hypervascular regions and serves as the most robust parameter in DSC MRI [12].

DSC-MRI data is most commonly analyzed using region-of-interest (ROI) measurements, which remains an operator-dependent method [13‐15]. ROI measurement of tumor cerebral blood volume (CBV) has shown good reproducibility [16], but nevertheless remains operator-dependent and fairly subjective, with an unavoidable component of interobserver and intra-observer variability [17, 18]. For example, repeatability and reproducibility were found to be below 50% and 10% respectively in one multicentre study [16]. However, it has been shown that a well-established image-review process can help to assess perfusion metrics more objectively. It has also been recommended to assess DSC-MRI data in the context of clinical trials by two experienced readers. In case of disagreement, an adjudicator could be consulted [16]. Albeit, this is clearly time consuming and thus impractical for clinical practice. However, much remains unknown with regard to measuring DSC-MRI data, including whether the shape of the ROI correlates with the variability in perfusion metrics. Also, as the aforementioned multicenter trial was performed on post-operative gliomas, the repeatability and reproducibility of DSC-MRI data in non-operated gliomas remains unknown. Finally, it remains unknown whether less experienced readers can provide robust results by using a well-established image-review process upfront to the assessment. Next to these unclarities, no broad consensus has been reached on the acquisition technique, post-processing algorithms, analysis methods or interpretation guidelines. Thereby, the use of DSC-MRI as an imaging biomarker is severely hampered, despite its extensive use in clinical practice [19].

This study aimed to investigate whether ROI shape impacts the repeatability and reproducibility of measured DSC-MRI metrics. The study was carried out after defining a well-established image-review protocol by three non-experienced readers. Finally, we assessed the repeatability and reproducibility of the DSC-MRI metrics in glioma patients in the pre- and post-surgical setting.

Table 1 provides information with regard to the demographics of the included patients. Table 2 provides an overview of the mean perfusion metrics per group. Below, we provide more specific information with regard to perfusion metrics and the impact of the differently shaped ROIs in the different subregions.

This study showed that readers with varying experience can deliver largely consistent DSC perfusion metrics in untreated glioma patient when working with a well-established image-review protocol. Consistency and DSC perfusion metrics were not impacted by shape of the placed ROI. Based on our findings, the most robust method to assess DSC-MR images in untreated patients comprised the use of circular ROIs and the usage of the CBV metric. With regard to the stability of tumor subregions, the whole tumor region and peritumoral region were considered to be consistently identified, whereas the results derived from the tumor hotspot varied more.

ROI-placement in the DSC-MR perfusion images of post-treatment patients, on the other hand, delivered highly variable results. In this group, shape of the ROIs did not further deteriorate ICC-values. In this group, the ICC-values of the circular ROI covering the whole tumor region provided the most optimal results with significant overlap between the three readers.

It was found, however, that specific tumor subregions yielded more consistent results as compared to other regions. These findings partially contradict the available literature, as it has been stated that polygonal ROIs improve accuracy of determining perfusion metrics [25]. Other groups also found that polygonal ROIs yielded the greatest inter-observer agreement as compared to circular ROIs [26‐28].

Regarding the deterioration of ICC values in the post-treatment group, this could be explained by the post-operative changes with regard to neuroanatomy, complicating interpretation of DSC MR perfusion images. A similar trend was observed by Barboriak et al. in dynamic contrast enhanced (DCE) MR perfusion imaging. They found that ICC-values for specific DCE metrics ranged from 0.90 to 1.00 when assessing baseline gliomas. However, the ICC-values of the posttreatment examinations deteriorated significantly and ranged from 0.48 to 0.76 [29]. This induces a dilemma in daily radiological practices as, following the European Association for Neuro-Oncology (EANO) guidelines, the role of DSC-MR perfusion imaging is to monitor efficacy of neurosurgical resection, radiotherapy and/or pharmacotherapy, or as surveillance imaging method after completion of the treatment [30]. Different readers could therefore interpret the same DSC-MR perfusion imaging scan differently, undermining the diagnostic effectiveness of, for example, distinguishing progression from pseudo-progression. This is especially important as developing more objective methods to obtain reproducible data is increasingly important in neuro-oncology imaging, especially when performing multicenter clinical trials. This could be of pivotal importance when using DSC perfusion imaging as a marker of angiogenesis to determine the effects of antiangiogenic treatment [31].

In the reading room, on the other hand, the present findings could also have a relevant impact. Research on MR perfusion in neuro-oncology focuses on the use of brain tumor vascularization and perfusion to predict WHO subtype and molecular diagnostics. In most studies, ROIs are drawn by different readers, independently [32‐34]. Due to the high ICC-values in the non-treated gliomas, one could argue that independent placement of ROIs by different observers, which is considered common practice in perfusion weighted imaging studies, is unnecessary. Another method to increase ICC-values in glioma perfusion research, could include the (semi-)automatic segmentation of the ROI. A recent publication on this topic showed that use of semiautomatic segmentation of contrast-enhancing tumor led to increased ICC and a lower coefficient of variation between readers [35].

The current study is partially limited by the inclusion of gliomas with different WHO grades. However, the majority of the included cases concerned glioblastoma (WHO grade IV), which limits the inclusion heterogeneity. Another limitation of the current study concerns its relative small sample size for each group and retrospective nature. One of the strengths of this paper concerns the inclusion of a matched posttreatment group which enables comparison of ICC-values of the pretreatment and posttreatment groups. These insights underline that measuring perfusion metrics in the clinical, often post-operative setting can be difficult and remain operator-dependent. A well-validated assessment protocol should therefore be established. Another strength concerns the use of differently shaped ROIs covering various areas of the tumor region, aiding in taking the first steps of establishing a well-validated assessment protocol for DSC-MR perfusion imaging in the clinical setting. Based on these findings, the authors would recommend different readers to assess post-operative DSC-MR perfusion imaging. However, pre-operative images could be assessed by one reader only.

Even inexperienced readers can provide consistent DSC perfusion metrics in pre-treatment glioma patients when using a well-established image-review protocol. However, in post-treatment glioma patients, double reading is warranted due to the high inconsistency of assessed DSC perfusion metrics. This study further showed that the ROI shape of preference can be chosen without impacting the quantified DSC perfusion metric assessment.