Purpose

Though widely adopted for brain perfusion measurement, Dynamic Susceptibility Contrast (DSC) MRI with low molecular weight gadolinium based contrast agent (GBCA) is often confounded by GBCA’s leakage into interstitium space. An additional pre-DSC injection of GBCA (preload) is often used to mitigate the underestimation of relative cerebral blood volume (rCBV) due to GBCA leakage. Here, we use a linearization approach to uniquely quantify GBCA leakage rate constant and show that the preload practice is generally effective in mitigating the leakage effect in DSC with GBCA.

Method

Eight female athymic nude rats (weight, 200–240 g) were anesthetized with intraperitoneal injection of ketamine (60 mg/kg) and diazepam (7.5 mg/kg) for stereotactic inoculation of U87MG cells. MRIs were performed 13 days after tumor implantation. DSC images were acquired with a gradient-echo sequence using the following parameters: TR/TE/FA, 31 ms/5.0 ms/6°; 3.2 x 3.2 cm² field of view, 128 x 128 matrix, and three contiguous 1-mm-thick sections. After an initial 30-second baseline acquisition, 60 μL bolus of gadodiamide (Omniscan) was injected (3 mL/min) via a tail vein catheter with an infusion pump, followed by 240 μL saline flush. Three consecutive DSC MRI acquisitions were performed with ~ 5 minutes between two adjacent runs. Each previous GBCA injection thus served as a preload for the subsequent DSC scan. We term the first DSC-MRI without GBCA preload as “no-preload”, the second DSC as “single-dose preload”, and the third as “double-dose preload”. Based on the simple assumption that the extravasating and intravascular GBCA introduced R₂^* change can be linearly combined,^1,2 the pixel ΔR₂^* time-course is given in Eq. (1), $$\triangle R_2^*(t)=K_{1} \cdot\triangle\overline{R_2^*(t)} - K_{2}\int_{0}^{t}\triangle\overline{R_2^*(t')} dt'\ (1) $$ where ΔR₂^* represents the pixel time-course for R₂^* change, K₁ and K₂ are proportional constants for intravascular and extravasating contributions to ΔR₂^*, respectively. $$$ \triangle\overline{R_2^*} $$$ approximates relative blood R₂^* change by combining signal changes from all non-leaking brain tissue pixels.² Using a method^3,4 similar to Gjedde-Patlak linearization,^4,5 we divide Eq. (1) by $$$\triangle\overline{R_2^*(t)}$$$ then graph the transformed right-hand-side term vs. that associated to the K₂ term (see Fig. 2 labels for expressions). If a linear portion exists in the graph, we can identify a leakage rate constant, K_L, as the slope of that segment. Similar to that of Gjedde-Patlak plot, K_L is only calculated from the linear portion after the transformation.

Results

Fig. 1a shows the three normalized DSC signal time-courses for one animal. Only the “no-preload” time-course (red) shows the apparent leakage effect with a portion of post-GBCA signal increases above baseline. In 1b, color-matched ΔR₂^* time-courses were plotted for the 1a signals. Only the no-preload ΔR₂^* time-course falls below zero. 1b inset shows a post-GBCA DSC image. Fig. 2 plots results after the linearization (symbols) of a different animal, and their respective linear regression lines. The magnitude of the leakage rate constant (K_L) is much larger without preload, while differences for the ones with preloads are generally much smaller and closer to zero. Fig. 3. summarizes the lesion ROI K_L values from all animals. The leakage rate is much greater than zero without preload. On the other hand, neither of the mean K_L value for s. d. or d. d. preload is significantly different from zero with p-values of 0.41 and 0.075, respectively.

Discussion

Using a linearization approach similar to the Gjedde-Patlak linearization, a unique pseudo first-order leakage rate constant, K_L, is defined for DSC with GBCA. Therefore, the K_L approach can provide a systematic way in investigating the effectiveness of the preload method in mitigating the leakage effect in DSC rCBV quantification. The results here show that the mean K_L values from eight animals are not significantly differ from zero, underscoring the generally effective preload approach. Still, with a p-value from the double dose preload at only 0.075, it is likely that injection timing and preload dose still could play a non-negligible role in the overall rCBV quantification.

Acknowledgements

Grant Support: R01- NS33618, R01-NS34608, and the Walter S. and Lucienne Driskill Foundation.

References

1. Weisskoff et al. Proc. Int. Soc. Magn. Reson. Med. 2, 279 (1994). 2. Boxerman, Schmainda, Weisskoff, AJNR Am. J. Neuroradiol. 27, 859-867 (2006). 3. Li, Varallyay, Gahramonov, Rooney, and Neuwelt, Proc. Intl. Soc. Mag. Reson. Med. 22, 0724 (2014). 4. Li, Varallyay, Gahramanov, Fu, Rooney, Neuwelt, NMR Biomed. 30, 1-9 (2017). 5. Gjedde, J. Neurochem. 36,1463-1471 (1981). 6. Patlak, Blasberg, Fenstermacher, J. Cereb. Blood Flow Metab. 3, 1-7 (1983).