Synopsis

Keywords: Relaxometry, Ex-Vivo Applications, Quantitative Imaging

The arterial distribution territories of the brain remain poorly characterized and are of major importance for treating ischemic strokes. We propose a new quantitative MRI approach for ex vivo mapping of vascular territories using paramagnetic gelatin injection into cerebral arteries to enhance MRI contrast at the gelatin level. This method was applied on one healthy human brain injected into the middle cerebral artery and accurately quantified the volume fraction of blood using a dedicated protocol and a modeling of the injected tissue signal. We projected the resulting mapping along the deep white matter fiber bundles.

Introduction

Ischemic stroke is a leading cause of disability worldwide. Although invaluable imaging developments have been made to better detect and address this pathology, the arterial distribution territories of the brain remain poorly characterized, and deserves to be studied in more detail. While gray matter (GM) studies already identify corresponding main arteries vascular territories, white matter (WM) arterial supply remains lacunar in literature¹. Current studies are mainly based on dissection and visual inspection. Since vascularization of the WM from one brain to another is more variable than GM vascularization, a systematic way of mapping seems necessary to investigate interindividual variability in a large cohort.
Following an old idea from Duvernoy² who injected cerebral vessels of ex vivo brains with black-colored gelatin before slicing to optically reveal their microvasculature, we propose a new ex vivo approach based on the post mortem simultaneous injection of paramagnetic gelatines into the arteries to map the cerebral vascular territories in the WM using quantitative MRI (qMRI).

Materials and methods

Tissue preparation: Following Smirnov et al. injection method³, a 90-year-old post mortem brain obtained from the Body Donation Program (Université de Tours) has been simultaneously injected with 1) 5%-gelatin solution colored with green dry pigment (SP vert 8, halogenated copper-phthalocyanine pigment), showing a specific paramagnetic signature after being formalin-fixed, in the middle cerebral artery (MCA) and 2) amagnetic pure 5%-gelatin solution in the anterior and posterior cerebral arteries in order to keep competitive pressure among distribution territories, mimicking the physiological blood flow.
Acquisition Protocol: A dedicated qMRI protocol was designed on a 3T Siemens Prisma MRI to highlight gelatin contrast diffused in the brain including: 1) a T₂-weighted Multi-Slice Multi-Echo (MSME) sequence (32 echo times T_E from 12.9ms to 412.8ms, 1.2mm isotropic, read bandwidth RBW=200Hz, repetition time T_R=10s) and 2) 45 T₁-weighted VFA-SPGR sequences (T_R=10ms, 45 flip angles FA ranging from 1° to 45°, 1.2mm isotropic). MSME-T_Es and VFA-SPGR-FAs were sampled to obtain T₂-weighted and T₁-weighted NMR signals revealing both parenchyma and gelatin compartments.
Multicompartment relaxometry model: Within a voxel, we modeled the measured signal stemming from various compartments (parenchyma and green gelatin) using a multi-compartment model^4-6. It is a common assumption to consider a "slow exchange" model of T₂ relaxometry signal: T₂ relaxation times are shorter compared to the time of water exchange between compartments. The measured signal results from the exponential contribution of the 2 compartments weighted by their volume fractions yielding such that:$$s_{MSME}(T_E)=\rho_{gelatin}f_{gelatin}e^{-\frac{T_E}{T_{2\;gelatin}}}+\rho_{parenchyma}(1–f_{gelatin})e^{-\frac{T_E}{T_{2\;parenchyma}}}$$On the contrary, a "fast exchange" model was used to model the T₁ relaxation effects between compartments, since water mixing is supposed to be fast compared to the range of T₁ values. The 2 compartments can therefore be compared to a single compartment characterized by an apparent relaxation time T_1eq:$$s_{VFA\;SPGR}(\theta)=\left| s_\rho\rho\frac{\left(1-e^{-\frac{T_R}{T_{1eq}}}\right)sin(\theta)}{1-e^{-\frac{T_R}{T_1}}cos(\theta)}\right|$$where$$\frac{1}{T_{1eq}}=\frac{(1–f_{gelatin})}{T_{1\;parenchyma}}+\frac{f_{gelatin}}{T_{1\;gelatin}}$$Post-processing: A robust WM mask was computed using the FSL fast method⁷. Quantitative gelatin (corresponding to the arterial compartment) and parenchyma volume fractions were estimated with previous relaxometric equations using a non-linear programming optimizer (based on the Nelder-Mead simplex⁸). The magnetic signature of doped gelatin was fixed to T₁/T₂=1580ms/225ms stemming from a preliminary estimation within a gelatin sample and the WM magnetic signature was fixed to T₁/T₂=690ms/110ms established from a non-MCA irrigated area. The post mortem brain was further matched to the MNI ICBM 152 Non-Linear Asymmetric 2009c template⁹ using the ANTs diffeomorphic registration toolbox^10,11, allowing the Chauvel’s human deep WM atlas¹² to be projected onto the quantitative MCA volume fraction map.

Results

Fig.1 illustrates the preparation and post-processing pipeline from the gelatin injection step till the projection of MCA volume contribution for each deep WM bundle. Fig.2 depicts axial and coronal slices of the MCA volume fraction map overlaid onto the corresponding proton density map computed from the T2-weighted qMRI data. Fig.3 shows the statistical analysis highlighting the MCA contribution in the vascularization of deep WM bundles. Fig.4 illustrates the vascularization of the most MCA-irrigated deep WM bundles (arcuate, frontal aslants, inferior fronto-occipital, inferior/middle longitudinal, optic radiations, uncinate and ventral visual stream).

Discussion

After injection, green gelatin was observed in a presumed PCA territory, probably arriving there retrogradely through the leptomeningeal or deep anastomoses. Interhemispheric variability of the anastomoses and in peripheral vascular resistance, could explain the left-right asymmetric presence of green gelatin in the occipital lobes. A validation procedure on serial sections after paraffin embedding is currently under treatment.
In summary, we accurately quantified the arterial blood supply in all deep WM bundles and identified main MCA perfused areas. It typically ranges from 0% to 12% of the volume, which is in good agreement with the literature. Regarding the left-right asymmetry observed in the occipital lobe, we could also notice a net increase of the MCA supply within deep WM bundles originating from the occipital regions.

Conclusion

The developed approach not only allowed to map the MCA vascular territory but also provided accurate quantification of the local volume fraction of supplied blood. We believe that our work paves the way for the elaboration of a precise atlas of the cerebral vascular territories, particularly in the WM. In the future, we aim to simultaneously map the 3 main vascular territories (of the anterior, middle and posterior cerebral arteries) using doped gelatins characterized by different magnetic signatures.