Fast learning of fiber orientation distribution function for MR tractography using convolutional neural network

Abstract

Purpose: In diffusion-weighted magnetic resonance imaging (DW-MRI), the fiber orientation distribution function (fODF) is of great importance for solving complex fiber configurations to achieve reliable tractography throughout the brain, which ultimately facilitates the understanding of brain connectivity and exploration of neurological dysfunction. Recently, multi-shell multi-tissue constrained spherical deconvolution (MSMT-CSD) method has been explored for reconstructing full fODFs. To achieve a reliable fitting, similar to other model-based approaches, a large number of diffusion measurements is typically required for MSMT-CSD method. The prolonged acquisition is, however, not feasible in practical clinical routine and is prone to motion artifacts. To accelerate the acquisition, we proposed a method to reconstruct the fODF from downsampled diffusion-weighted images (DWIs) by leveraging the strong inference ability of the deep convolutional neural network (CNN). Methods: The method treats spherical harmonics (SH)-represented DWI signals and fODF coefficients as inputs and outputs, respectively. To compensate for the reduced gradient directions with reduced number of DWIs in acquisition in each voxel, its surrounding voxels are incorporated by the network for exploiting their spatial continuity. The resulting fODF coefficients are fitted with applying the CNN in a multi-target regression model. The network is composed of two convolutional layers and three fully connected layers. To obtain an initial evaluation of the method, we quantitatively measured its performance on a simulated dataset. Then, for in vivo tests, we employed data from 24 subjects from the Human Connectome Project (HCP) as training set and six subjects as test set. The performance of the proposed method was primarily compared to the super-resolved MSMT-CSD with the decreasing number of DWIs. The fODFs reconstructed by MSMT-CSD from all available 288 DWIs were used as training labels and the reference standard. The performance was quantitatively measured by the angular correlation coefficient (ACC) and the mean angular error (MAE). Results: For the simulated dataset, the proposed method exhibited the potential advantage over the model reconstruction. For the in vivo dataset, it achieved superior results over the MSMT-CSD in all the investigated cases, with its advantage more obvious when a limited number of DWIs were used. As the number of DWIs was reduced from 95 to 25, the median ACC ranged from 0.96 to 0.91 for the CNN, but 0.93 to 0.77 for the MSMT-CSD (with perfect score of 1). The angular error in the typical regions of interest (ROIs) was also much lower, especially in multi-fiber regions. The average MAE for the CNN method in regions containing one, two, three fibers was, respectively, 1.09°, 2.75°, and 8.35°smaller than the MSMT-CSD method. The visual inception of the fODF further 3101 confirmed this superiority. Moreover, the tractography results validated the effectiveness of the learned fODF, in preserving known major branching fiber...

Deep learning‐based method for reducing residual motion effects in diffusion parameter estimation

“…In this study, we used a deep learning‐ (DL‐) based parameter estimation method to address residual motion effects. This was motivated by the fact that DL‐based methods have recently been shown to have excellent results in estimating diffusion‐derived measures from highly undersampled data 27‐31 . Here, we demonstrate this new approach by combining a DL‐based method using a hierarchical convolutional neural network (H‐CNN) 29 with a motion assessment and data rejection procedure (Figure 1C).…”

“…This was motivated by the fact that DL-based methods have recently been shown to have excellent results in estimating diffusion-derived measures from highly undersampled data. [27][28][29][30][31] Here, we demonstrate this new approach by combining a DL-based method using a hierarchical convolutional neural network (H-CNN) 29 with a motion assessment and data rejection procedure ( Figure 1C). The proposed technique is evaluated for recovering DKI-and DTI-derived measures from motion-contaminated data.…”

FOD-Net: A Deep Learning Method for Fiber Orientation Distribution Angular Super Resolution

“…Existing deep learning models Lin et al (2019); Tian et al (2020) for dMRI angular super-resolution tasks are based on fixed-size dMRI images, which is not in principle suitable for our fiber orientation map angular super-resolution task because the size of dMRI data can vary considerably between acquisition protocols. To address this, the proposed FOD-Net takes (sequentially) as input an FOD patch cropped from a singleshell LARDI FOD image and outputs the angular super-resolved version of the central voxel of this patch.…”

“…A growing interest in healthy and diseased brain connectomics has highlighted a critical need to bridge the gap between high-angular resolution, multi-shell dMRI acquisitions and the restricted practical capacity of clinical MRI investigations. Pioneering work Scherrer et al (2011, 2012); Lin et al (2019); Tian et al (2020) has attempted to overcome the limitations of clinical protocols by striving for dMRI reconstruction, which in turn reconstruct more reliable fiber tracking connectivity information from dMRI itself. However, these methods require specific dMRI acquisition protocols (for example, a certain number of gradient directions), which is inflexible and impractical in general clinical environments.…”

Diffusion Imaging in the Post HCP Era

“…9) is one reason that DL models will be further advanced and used to extract even more information than what can easily be described with conventional techniques. Recent studies report accurate diffusion tensor imaging (DTI) reconstruction 93 and a fiber orientation distribution function 94 reconstruction, using neural networks, which promise to improve tractography even in complex tissue geometry areas and demonstrate the extraction of increased information from existing data. DL‐based reconstruction methods have also shown the ability to generate images for diffusion kurtosis imaging (DKI) and the NODDI model with less data, 95 and the transition from these retrospective analysis to prospective analyses will elucidate the applications that will require larger amounts of data or will perform well with less data, and from which more information can be obtained.…”