D-BRAIN: Anatomically Accurate Simulated Diffusion MRI Brain Data

Abstract

Diffusion Weighted (DW) MRI allows for the non-invasive study of water diffusion inside living tissues. As such, it is useful for the investigation of human brain white matter (WM) connectivity in vivo through fiber tractography (FT) algorithms. Many DW-MRI tailored restoration techniques and FT algorithms have been developed. However, it is not clear how accurately these methods reproduce the WM bundle characteristics in real-world conditions, such as in the presence of noise, partial volume effect, and a limited spatial and angular resolution. The difficulty lies in the lack of a realistic brain phantom on the one hand, and a sufficiently accurate way of modeling the acquisition-related degradation on the other. This paper proposes a software phantom that approximates a human brain to a high degree of realism and that can incorporate complex brain-like structural features. We refer to it as a Diffusion BRAIN (D-BRAIN) phantom. Also, we propose an accurate model of a (DW) MRI acquisition protocol to allow for validation of methods in realistic conditions with data imperfections. The phantom model simulates anatomical and diffusion properties for multiple brain tissue components, and can serve as a ground-truth to evaluate FT algorithms, among others. The simulation of the acquisition process allows one to include noise, partial volume effects, and limited spatial and angular resolution in the images. In this way, the effect of image artifacts on, for instance, fiber tractography can be investigated with great detail. The proposed framework enables reliable and quantitative evaluation of DW-MR image processing and FT algorithms at the level of large-scale WM structures. The effect of noise levels and other data characteristics on cortico-cortical connectivity and tractography-based grey matter parcellation can be investigated as well.

Fig 1. Brain like phantom creation.

Proposed pipeline: from realistic inputs and a simulated DW MRI scanner to brain-like D-BRAIN data. The tissue volume fractions have a resolution of 0.7 × 0.7 × 0.7 mm³, streamlines step-size is 0.7 mm.

Fig 2. D-BRAIN.

Anatomical MR image (a) and corresponding diffusion-weighted images for D-BRAIN data. B-values of 1000 s mm⁻² (b), 2500 s mm⁻² (c) and 10000 s mm⁻² (d). For each picture, the intensity values have been optimized for visual purposes.

Fig 3. Phantom WM bundles tractography.

A portion of the corticospinal tract (a), pathways of the forceps minor (b), tracts following the cingulum (c), streamlines belonging to the fornix (d), and part of the uncinate fasciculus (e).

Fig 4. Phantom WM bundles tractography.

The upper portion of the corticospinal tract (a) has a lower FA in the region highlighted because of the contribution of other fibers coming from the body of the corpus callosum. In (b), the crossing of the corticospinal tracts (blue) and the lateral projections of the corpus callosum (red) are clearly visible.

Fig 5. Estimated phantom fODF for different SNR.

Yellow boxes highlight the effect of noise level. Top left: ROI. ROI magnification for: SNR = 30 (a), SNR = 20 (b), and SNR = 15 (c). B-value = 1000 s mm².

Fig 6. Estimated phantom fODF for different resolutions.

Top left: ROI. ROI magnification for: voxel size = 1.4 × 1.4 × 1.4 mm³ (a), voxel size = 2.1 × 2.1 × 2.1 mm³ (b). SNR = inf.

Fig 7. D-BRAIN FA maps before and after k-space downsampling.

Voxel size is 0.7 × 0.7 × 0.7 mm³ (a), 1.4 × 1.4 × 1.4 mm³ (b) and 2.1 × 2.1 × 2.1 mm³ (c). Yellow arrows highlight the Gibbs ringing artifact.

Fig 8. Estimated phantom fODF for different b-values.

Yellow boxes highlight the effect of the b-value. Top left: ROI. ROI magnification for: b-value = 1000 s mm² (a), b-value = 2500 s mm² (b), b-value = 10000 s mm² (c). SNR = inf.

Fig 9. Connectivity analysis.

The 82 GM parcels used as network nodes (a) superimposed to the FA map. In (b), the corresponding connectivity matrix built counting the number of streamlines connecting each couple of parcels (intensities displayed in logarithmic scale).

Fig 10. Estimated network connectivity measures.

Variability across different SNR and resolutions of D-BRAIN data. Ground truth: from noiseless data whose voxel size is 1.4 × 1.4 × 1.4 mm³. High Resolution: from noisy data whose voxel size is 1.4 × 1.4 × 1.4 mm³. Low Resolution: from noisy data whose voxel size is 2.1 × 2.1 × 2.1 mm³.