Enhancing MR vascular Fingerprinting with realistic microvascular geometries

Abstract

Magnetic resonance (MR) vascular Fingerprinting proposes to use the MR Fingerprinting framework to quantitatively and simultaneously map several characteristics that emerge from vascular structure much smaller than voxel size. The initial implementation assessed the local blood oxygenation saturation (SO₂), blood volume fraction (BVf), and vessel averaged radius (R) in humans and rodent brains using simple 2D representations of the vascular network during dictionary generation. In order to improve the results and possibly extend the approach to pathological environments and other biomarkers, we propose in this study to use 3D realistic vascular geometries in the numerical simulations. 28,000 different synthetic voxels containing vascular networks segmented from whole-brain healthy mice microscopy images were created. A Bayesian-based regression model was used for map reconstruction. We show in 8 healthy and 9 tumor-bearing rats that realistic vascular representations yield microvascular estimates in better agreement with the literature than 2D or 3D cylindrical models. Furthermore, tumoral blood oxygenation variations observed with the proposed approach are the only ones correlating with in vivo optic-fiber measurements performed in the same animals.

Keywords: Fingerprinting; MRI; brain; oxygenation; tumor; vascular.

Fig. 1.

(A) Top. Examples of the whole-brain vascular network and a single 744 µm thick slice from dataset 1. Both are eroded for visibility. Bottom. Three examples of MRI-sized voxels obtained. (B) Distributions of BVf and mean radius in the 28,000 voxels generated. (C) Overview of the MRF process: a sequence is used both for numerical simulations and for in-vivo acquisitions. Parametric maps are obtained either by comparing acquisitions to simulations (DBM), or by using Dictionary-based Learning (DBL).

Fig. 2.

Comparison of parametric maps obtained through MRvF using the 3D-micro dictionary for healthy and tumoral tissue. For each of the 4 parameters, the top and bottom lines correspond to the result from dictionary-based learning (DBL), and dictionary-based matching (DBM), respectively. Values outside the presented range are clipped to the corresponding extrema.

Fig. 3.

Comparison of parametric maps obtained through MRvF using the DBL reconstruction and the 3D-micro dictionary. Healthy and tumoral brains are presented for each parameter column. The top three rows correspond to MRvF results with the 3D-micro, the 2D-synth, and the 3D-synth dictionaries, respectively. The bottom row corresponds to the vascular parameters obtained with the analytical approach.

Fig. 4.

Quantitative estimates of the 4 parameters for the different methods. All results from MRvF were obtained with the DBL method. “Healthy” values are averaged in the striatum for each animal, “Tumor” values from the lesion, and “Contra.” from a contralateral ROI matching the tumor’s location and size. Stars indicate the significance with a p-value$\le$0.05.

Fig. 5.

pO₂measurements obtained through Oxylite measurements. Six animals were implanted with optic fiber measuring the pO₂at two different depths, both in the lesion and the contralateral hemisphere. Between 5 and 10 measurements were performed at each depth. (left) Boxplots of the values obtained in both zones on all animals. The mean value in each hemisphere for each animal is shown. (right) Measurements obtained in one animal. Boxplots and actual values, at two different depths, are shown. Stars indicate the significance with a p-value ≤ 0.05, for a 2-sample t-test.