Ocrelizumab-treated patients with relapsing multiple sclerosis show volume loss rates similar to healthy aging.

BACKGROUND: Multiple sclerosis (MS) is a chronic autoimmune disease of the central nervous system characterized by two major and interconnected hallmarks: inflammation and progressive neurodegeneration. OBJECTIVE: The aim of this work was to compare neurodegenerative processes, in the form of global and regional brain volume loss rates, in healthy controls (HCs) and in patients with relapsing MS (RMS) treated with ocrelizumab, which suppresses acute inflammation. METHODS: Whole brain, white matter, cortical gray matter, thalamic, and cerebellar volume loss rates were assessed in 44 HCs that were part of a substudy in the OPERA II randomized controlled trial (NCT01412333) and 59 patients with RMS enrolled in the same substudy as well as age- and sex-matched patients in OPERA I (NCT01247324) and II. Volume loss rates were computed using random coefficients models over a period of 2 years. RESULTS: Ocrelizumab-treated patients showed global and regional brain volume loss rates that were approaching that of HCs. CONCLUSION: These findings are consistent with an important role of inflammation on overall tissue loss and the role of ocrelizumab in reducing this phenomenon.

Detection and clinical correlation of leukocortical lesions in pediatric-onset multiple sclerosis on multi-contrast MRI.

OBJECTIVE: To determine the frequency of cortical lesions (CLs) in patients with pediatric-onset multiple sclerosis (POMS) using multi-contrast magnetic resonance imaging (MRI), and the relationship between frontal CL load and upper limb dexterity assessed with the Nine-Hole Peg Test (9-HPT). METHODS: Participants completed the 9-HPT and were imaged on a 3T MRI scanner to collect T1-weighted three-dimensional (3D) magnetization prepared rapid gradient echo (MPRAGE), proton density-weighted, T2-weighted and fluid-attenuated inversion recovery (FLAIR) images. CLs were manually segmented using all MRI contrasts. RESULTS: We enrolled 24 participants with POMS (mean (standard deviation) age at first symptom: 13.3 (±2.7) years; mean age at scan: 18.8 (±3) years; mean disease duration of 5 (±3.2) years). A total of 391 CLs (mean, 16.3 ± 27.2; median, 7) were identified in 19 of 24 POMS patients (79%). The total number of CLs was positively associated with white matter lesion volume ( p = 0.04) but not with thalamic volume, age at the time of the scan, or disease duration. The number of frontal CLs was associated with slower performance on the 9-HPT ( p = 0.05). CONCLUSION: Multi-contrast 3T MRI led to a high rate of CL detection, demonstrating that cortical pathology occurs even in pediatric-onset disease. Frontal lobe CL count was associated with reduced manual dexterity, indicating that these CLs are clinically relevant.

Deep learning, data ramping, and uncertainty estimation for detecting artifacts in large, imbalanced databases of MRI images.

Magnetic resonance imaging (MRI) is increasingly being used to delineate morphological changes underlying neurological disorders. Successfully detecting these changes depends on the MRI data quality. Unfortunately, image artifacts frequently compromise the MRI utility, making it critical to screen the data. Currently, quality assessment requires visual inspection, a time-consuming process that suffers from inter-rater variability. Automated methods to detect MRI artifacts could improve the efficiency of the process. Such automated methods have achieved high accuracy using small datasets, with balanced proportions of MRI data with and without artifacts. With the current trend towards big data in neuroimaging, there is a need for automated methods that achieve accurate detection in large and imbalanced datasets. Deep learning (DL) is the ideal MRI artifact detection algorithm for large neuroimaging databases. However, the inference generated by DL does not commonly include a measure of uncertainty. Here, we present the first stochastic DL algorithm to generate automated, high-performing MRI artifact detection implemented on a large and imbalanced neuroimaging database. We implemented Monte Carlo dropout in a 3D AlexNet to generate probabilities and epistemic uncertainties. We then developed a method to handle class imbalance, namely data-ramping to transfer the learning by extending the dataset size and the proportion of the artifact-free data instances. We used a 34,800 scans (98% clean) dataset. At baseline, we obtained 89.3% testing accuracy (F1 = 0.230). Following the transfer learning (with data-ramping), we obtained 94.9% testing accuracy (F1 = 0.357) outperforming focal cross-entropy (92.9% testing accuracy, F1 = 0.304) incorporated for comparison at handling class imbalance. By implementing epistemic uncertainties, we improved the testing accuracy to 99.5% (F1 = 0.834), outperforming the results obtained in previous comparable studies. In addition, we estimated aleatoric uncertainties by incorporating random flips to the MRI volumes, and demonstrated that aleatoric uncertainty can be implemented as part of the pipeline. The methods we introduce enhance the efficiency of managing large databases and the exclusion of artifact images from big data analyses.

Using Deep Learning Algorithms to Automatically Identify the Brain MRI Contrast: Implications for Managing Large Databases.

Neuroimaging science has seen a recent explosion in dataset size driving the need to develop database management with efficient processing pipelines. Multi-center neuroimaging databases consistently receive magnetic resonance imaging (MRI) data with unlabeled or incorrectly labeled contrast. There is a need to automatically identify the contrast of MRI scans to save database-managing facilities valuable resources spent by trained technicians required for visual inspection. We developed a deep learning (DL) algorithm with convolution neural network architecture to automatically infer the contrast of MRI scans based on the image intensity of multiple slices. For comparison, we developed a random forest (RF) algorithm to automatically infer the contrast of MRI scans based on acquisition parameters. The DL algorithm was able to automatically identify the MRI contrast of an unseen dataset with <0.2% error rate. The RF algorithm was able to identify the MRI contrast of the same dataset with 1.74% error rate. Our analysis showed that reduced dataset sizes caused the DL algorithm to lose generalizability. Finally, we developed a confidence measure, which made it possible to detect, with 100% specificity, all MRI volumes that were misclassified by the DL algorithm. This confidence measure can be used to alert the user on the need to inspect the small fraction of MRI volumes that are prone to misclassification. Our study introduces a practical solution for automatically identifying the MRI contrast. Furthermore, it demonstrates the powerful combination of convolution neural networks and DL for analyzing large MRI datasets.

Recent advances in diffusion MRI modeling: Angular and radial reconstruction.

Recent advances in diffusion magnetic resonance image (dMRI) modeling have led to the development of several state of the art methods for reconstructing the diffusion signal. These methods allow for distinct features to be computed, which in turn reflect properties of fibrous tissue in the brain and in other organs. A practical consideration is that to choose among these approaches requires very specialized knowledge. In order to bridge the gap between theory and practice in dMRI reconstruction and analysis we present a detailed review of the dMRI modeling literature. We place an emphasis on the mathematical and algorithmic underpinnings of the subject, categorizing existing methods according to how they treat the angular and radial sampling of the diffusion signal. We describe the features that can be computed with each method and discuss its advantages and limitations. We also provide a detailed bibliography to guide the reader.

Apparent intravoxel fibre population dispersion (FPD) using spherical harmonics.

The vast majority of High Angular Resolution Diffusion Imaging (HARDI) modeling methods recover networks of neuronal fibres, using a heuristic extraction of their local orientation. In this paper, we present a method for computing the apparent intravoxel Fibre Population Dispersion (FPD), which conveys the manner in which distinct fibre populations are partitioned within the same voxel. We provide a statistical analysis, without any prior assumptions on the number or size of these fibre populations, using an analytical formulation of the diffusion signal autocorrelation function in the spherical harmonics basis. We also propose to extract features of the FPD obtained in the group of rotations, using several metrics based on unit quaternions. We show results on simulated data and on physical phantoms, that demonstrate the effectiveness of the FPD to reveal regions with crossing tracts, in contrast to the standard anisotropy measures.

Efficient and robust computation of PDF features from diffusion MR signal.

We present a method for the estimation of various features of the tissue micro-architecture using the diffusion magnetic resonance imaging. The considered features are designed from the displacement probability density function (PDF). The estimation is based on two steps: first the approximation of the signal by a series expansion made of Gaussian-Laguerre and Spherical Harmonics functions; followed by a projection on a finite dimensional space. Besides, we propose to tackle the problem of the robustness to Rician noise corrupting in-vivo acquisitions. Our feature estimation is expressed as a variational minimization process leading to a variational framework which is robust to noise. This approach is very flexible regarding the number of samples and enables the computation of a large set of various features of the local tissues structure. We demonstrate the effectiveness of the method with results on both synthetic phantom and real MR datasets acquired in a clinical time-frame.

Evaluation of q-space sampling strategies for the diffusion magnetic resonance imaging.

We address the problem of efficient sampling of the diffusion space for the Diffusion Magnetic Resonance Imaging (dMRI) modality. While recent scanner improvements enable the acquisition of more and more detailed images, it is still unclear which q-space sampling strategy gives the best performance. We evaluate several q-space sampling distributions by an approach based on the approximation of the MR signal by a series expansion of Spherical Harmonics and Laguerre-Gaussian functions. With the help of synthetic experiments, we identify a subset of sampling distributions which leads to the best reconstructed data.

Efficient computation of PDF-based characteristics from diffusion MR signal.

We present a general method for the computation of PDF-based characteristics of the tissue micro-architecture in MR imaging. The approach relies on the approximation of the MR signal by a series expansion based on Spherical Harmonics and Laguerre-Gaussian functions, followed by a simple projection step that is efficiently done in a finite dimensional space. The resulting algorithm is generic, flexible and is able to compute a large set of useful characteristics of the local tissues structure. We illustrate the effectiveness of this approach by showing results on synthetic and real MR datasets acquired in a clinical time-frame.