DCE-MRI is widely used in the clinic to determine leakage of a contrast medium in tumors. Recently, this method is increasingly applied to estimate the rather subtle leakage through the blood-brain barrier (BBB) in cerebrovascular and neurodegenerative pathology (e.g. cerebral small vessel disease (cSVD) and Alzheimer’s disease)^1,2. Several studies have shown that breakdown of the BBB might play a pivotal role in the pathophysiology of these diseases^3,4.

Measuring subtle leakage is challenging as the magnitude is several orders lower than that found in high-grade tumors. To have a clinical application, the measurement of subtle leakage using DCE-MRI has to be reproducible. The aim is to determine the reproducibility of the pharmacokinetic measures, the transfer constant (K_i) and fractional plasma volume (v_p), using dual temporal resolution DCE-MRI in patients with cerebrovascular disease. In addition, the effect of different vascular input functions (VIFs) on the reproducibility was investigated.

Patients: MR imaging was conducted twice in 14 patients with cSVD (n=10), cortical stroke (n=3) or intracerebral hemorrhage (n=1) (age 68.0±6.8y) to obtain a range of leakage values. Imaging was conducted on a 3.0 Tesla MR scanner (Philips Achieva TX) on two separate days with on average 2.1±2.3 days in between.

Data acquisition: Dual temporal resolution DCE-MRI was used for dynamic imaging to quantify the leakage rate through the BBB and the microvascular blood plasma space⁵. The acquisition compromises the combination of two dynamic sequences with different dynamic scan times (DST). The first sequence was applied during bolus injection (DST 3.2 s, TR/TE = 5.6/2.5 ms, FOV: 256x200x50 mm³, voxel size 2x2x5 mm³, 29 volumes). Subsequently, the second sequence was conducted (DST 30.5 s, TR/TE = 5.6/2.5 ms, FOV: 256x256x100 mm³, voxel size 1x1x2 mm³, 45 volumes). The contrast agent (Gadobutrol, dose 0.1 mmol/kg, 3 ml/s) was injected in the antecubital vein.

Data analysis: Voxel wise calculation of K_i and v_p was performed by using the Patlak graphical approach⁶. Histogram analysis was performed to calculate the mean v_p and the 75th percentile of K_i, to determine the strongest leakage region with respect to noise. The following regions of interest (ROIs) were analyzed: vascular lesion (VL), white matter (WM) and grey matter (GM) (both distant from the VL). The individual VIFs were taken from the sagittal sinus. To explore the influence of the VIF on the day-to-day variation, a session-averaged VIF per subject was also calculated over the two sessions.

Reproducibility was expressed by 2 measures: (i) the correlation coefficient (CV) to measure the relative within-subject variation and (ii) the intraclass correlation coefficient (ICC) that expresses the part of the total variance ascribed to biological variation rather than measurement error. Lower CV and higher ICC are indicative of good reproducibility. Bland-Altman plots and the 95% limits of agreement (LoA) were also depicted⁷.

No correlation between the magnitude and difference between sessions was present for the K_i and v_p in the WM and GM (Fig. 1,2).

Table 1 shows the reproducibility values. Fairly low CVs and high ICCs were found in the WM and GM for both measures. For the VLs, high ICC was observed for K_i, whereas for v_p it was low (Fig. 3).

Using session-averaged VIFs instead of individual VIFs, higher CVs and lower ICCs were observed for K_i. In contrast, lower CVs and higher ICCs were seen for the v_p .

The magnitude of K_i and v_p are comparable to the inter-session variability (LoA) (Fig. 1,2). Moreover, low CVs and moderate to high ICCs observed for K_i and v_p indicate a moderate reproducibility in the WM and GM. In the VLs, contradictory reproducibility values were observed for the K_i and v_p, for which more patients need to be investigated.

Furthermore, the reproducibility of K_i in the more vascularized GM (more leakage) was lower than in the WM. This might be caused by stronger partial volume effects in the GM by CSF and macrovessels, which are likely stronger in the (cortical) GM than WM.

Day-to-day variation might be induced by variation in the tissue signal and the VIF. Using session-averaged VIFs, it was observed that the reproducibility of v_p improved, whereas K_i did not show this effect.

These results demonstrate that despite the noisy nature of the measurement, determining subtle leakage in cerebrovascular patients using dual temporal DCE-MRI is moderately reproducible. Still, cautious interpretation of the K_i and v_p in individual patients is needed. Day-to-day variations may be partly compensated by using session-averaged (or standardized) VIFs.