Contrast-Free Detection of Focused Ultrasound-Induced Blood-Brain Barrier Opening Using Diffusion Tensor Imaging

Abstract

Focused ultrasound (FUS) has emerged as a non-invasive technique to locally and reversibly disrupt the blood-brain barrier (BBB). Here, we investigate the use of diffusion tensor imaging (DTI) as a means of detecting FUS-induced BBB opening at the absence of an MRI contrast agent. A non-human primate (NHP) was repeatedly treated with FUS and preformed circulating microbubbles to transiently disrupt the BBB (n = 4). T1- and diffusion-weighted MRI scans were acquired after the ultrasound treatment, with and without gadolinium-based contrast agent, respectively. Both scans were registered with a high-resolution T1-weighted scan of the NHP to investigate signal correlations. DTI detected an increase in fractional anisotropy from 0.21 ± 0.02 to 0.38 ± 0.03 (82.6 ± 5.2% change) within the targeted area one hour after BBB opening. Enhanced DTI contrast overlapped by 77.22 ± 9.2% with hyper-intense areas of gadolinium-enhanced T1-weighted scans, indicating diffusion anisotropy enhancement only within the BBB opening volume. Diffusion was highly anisotropic and unidirectional within the treated brain region, as indicated by the direction of the principal diffusion eigenvectors. Polar and azimuthal angle ranges decreased by 35.6% and 82.4%, respectively, following BBB opening. Evaluation of the detection methodology on a second NHP (n = 1) confirmed the across-animal feasibility of the technique. In conclusion, DTI may be used as a contrast-free MR imaging modality in lieu of contrast-enhanced T1 mapping for detecting BBB opening during focused-ultrasound treatment or evaluating BBB integrity in brain-related pathologies.

Fig. 1.

A. Timeline of the experimental procedure. The baseline scans were acquired once one month before the initiation of the sonications. SWI and DTI precede Gadolinium injection followed by T1-weighted imaging. The procedure was repeated once per month for six treatments in total, yet only four were successfully completed. B. Transducer orientation relative to the brain fixed on the stereotactic frame. C. Gradient direction coordinates are presented to confirm that diffusion sampling occurred in the entire sphere, a requirement imposed by the eddy current correction accuracy. D. Flowchart of the DTI processing pipeline. The raw data were transformed from DICOM to NIFTI format and all the directions were registered to the weight- and gradient-free image of the scan acquired before the sonications. The registered images were combined in a 4D format that was isolated from the surrounding brain tissue and corrected for eddy current artifacts. Then, the calculation of the tensor and the corresponding eigenvalues and eigenvectors resulted from the DTIFIT. The mean diffusivity and the fractional anisotropy maps were quantified by employing the appropriate equations. The difference in the FA values is reported herein, denoted as ΔFA, and resulted from the subtraction of the FA map obtained before the sonication from the FA map acquired following the sonication.

Fig. 2.

BBB opening detection with FA map, ΔFA and Gadolinium-Enhanced T1-weighted imaging for the four successfully completed experiments targeting a similar structure (Caudate nucleus region) in the same NHP. A. FA_POST maps reveal the BBB opening site despite the signal coming from other anisotropic structures. B. Normalization of the FA_POST maps (jet colormap) with the FAPRE map resulted in the pronounced BBB opening site shown in the ΔFA maps (jet colormap) overlayed on the anatomical scan (grayscale) for reference. C. The T1-weighted images’ ratio, RGd-T1, is presented herein overlayed on the same anatomical scan for comparison with the developing modality. D. Longitudinal assessment of the BBB opening volume shows comparability of the two modalities both qualitatively and quantitatively. E. Accordingly, the average BBB opening volume is similar between modalities. F,G. The FA value increased in all four cases following the sonication while increased on average by 82% from 0.21 ± 0.02 to 0.38 ± 0.03 (t[3]=27.73; P=0.0001).

Fig. 3.

A. DTI eigenvectors overlaid onto the ΔFA map for the entire axial brain plane and the magnified striatal region. BBB opening initiates an increase in the directionality of the water molecule diffusion compared to the intact barrier shown by the consistent direction of the arrows in the sonicated area compared to the contralateral side within the ROI (black square). B. Polar and azimuthal angle distributions in the ipsilateral and contralateral hemispheres within the ROI are presented showing the narrow range of polar and azimuthal angles of the principal eigenvector only at the site of sonication (_v1POST-IPSI). C. Cumulative results showed a decrease in the polar angle range on the order of 35.58% (t[3]=3.921; P=0.0295) compared to the ipsilateral side, v_1PRE-IPSI, and 53.86% (t[3]=4.887; P=0.0164) compared to the contralateral side, v_1POST-CONTRA. Similarly, the azimuthal angle range of the v_1POST-IPSI decreased by 82.44% (t[3]=3.699; P=0.0343) compared to v_1PRE-IPSI, and 84.55% (t[3]=7.462; P=0.005) compared v_1POST-CONTRA.

Fig. 4.

Evaluation of the developed methodology in a second NHP. A. The FA_POST map was normalized with the corresponding FA_PRE map to produce the ΔFA map that revealed a BBB opening volume about 75.14 mm³ similar to the 81.29 mm³ volume obtained from the RGd-T1w quantification. B. Polar and azimuthal angle distributions in the ipsilateral and contralateral hemispheres within the ROI are presented showing a decrease by 32.18% and 31.18% of the v_1POST-IPSI polar angle range compared to the v_1PRE-IPSI and v_1POST-CONTRA, respectively. Along the same lines, the azimuthal angle range of the v_1POST-IPSI decreased by 5% and 7.95% compared to v_1PRE-IPSI and to v_1POST-CONTRA respectively. C. The S₀ image along with the L₁ and MD maps are presented for image quality assurance.