Cortical lesion hotspots and association of subpial lesions with disability in multiple sclerosis.

BACKGROUND: Dramatic improvements in visualization of cortical (especially subpial) multiple sclerosis (MS) lesions allow assessment of impact on clinical course. OBJECTIVE: Characterize cortical lesions by 7 tesla (T) T2*-/T1-weighted magnetic resonance imaging (MRI); determine relationship with other MS pathology and contribution to disability. METHODS: Sixty-four adults with MS (45 relapsing-remitting/19 progressive) underwent 3 T brain/spine MRI, 7 T brain MRI, and clinical testing. RESULTS: Cortical lesions were found in 94% (progressive: median 56/range 2-203; relapsing-remitting: 15/0-168; p = 0.004). Lesion distribution across 50 cortical regions was nonuniform (p = 0.006), with highest lesion burden in supplementary motor cortex and highest prevalence in superior frontal gyrus. Leukocortical and white matter lesion volumes were strongly correlated (r = 0.58, p < 0.0001), while subpial and white matter lesion volumes were moderately correlated (r = 0.30, p = 0.002). Leukocortical (p = 0.02) but not subpial lesions (p = 0.40) were correlated with paramagnetic rim lesions; both were correlated with spinal cord lesions (p = 0.01). Cortical lesion volumes (total and subtypes) were correlated with expanded disability status scale, 25-foot timed walk, nine-hole peg test, and symbol digit modality test scores. CONCLUSION: Cortical lesions are highly prevalent and are associated with disability and progressive disease. Subpial lesion burden is not strongly correlated with white matter lesions, suggesting differences in inflammation and repair mechanisms.

On the generalizability of diffusion MRI signal representations across acquisition parameters, sequences and tissue types: Chronicles of the MEMENTO challenge.

Diffusion MRI (dMRI) has become an invaluable tool to assess the microstructural organization of brain tissue. Depending on the specific acquisition settings, the dMRI signal encodes specific properties of the underlying diffusion process. In the last two decades, several signal representations have been proposed to fit the dMRI signal and decode such properties. Most methods, however, are tested and developed on a limited amount of data, and their applicability to other acquisition schemes remains unknown. With this work, we aimed to shed light on the generalizability of existing dMRI signal representations to different diffusion encoding parameters and brain tissue types. To this end, we organized a community challenge - named MEMENTO, making available the same datasets for fair comparisons across algorithms and techniques. We considered two state-of-the-art diffusion datasets, including single-diffusion-encoding (SDE) spin-echo data from a human brain with over 3820 unique diffusion weightings (the MASSIVE dataset), and double (oscillating) diffusion encoding data (DDE/DODE) of a mouse brain including over 2520 unique data points. A subset of the data sampled in 5 different voxels was openly distributed, and the challenge participants were asked to predict the remaining part of the data. After one year, eight participant teams submitted a total of 80 signal fits. For each submission, we evaluated the mean squared error, the variance of the prediction error and the Bayesian information criteria. The received submissions predicted either multi-shell SDE data (37%) or DODE data (22%), followed by cartesian SDE data (19%) and DDE (18%). Most submissions predicted the signals measured with SDE remarkably well, with the exception of low and very strong diffusion weightings. The prediction of DDE and DODE data seemed more challenging, likely because none of the submissions explicitly accounted for diffusion time and frequency. Next to the choice of the model, decisions on fit procedure and hyperparameters play a major role in the prediction performance, highlighting the importance of optimizing and reporting such choices. This work is a community effort to highlight strength and limitations of the field at representing dMRI acquired with trending encoding schemes, gaining insights into how different models generalize to different tissue types and fiber configurations over a large range of diffusion encodings.

Tractography reproducibility challenge with empirical data (TraCED): The 2017 ISMRM diffusion study group challenge.

BACKGROUND: Fiber tracking with diffusion-weighted MRI has become an essential tool for estimating in vivo brain white matter architecture. Fiber tracking results are sensitive to the choice of processing method and tracking criteria. PURPOSE: To assess the variability for an algorithm in group studies reproducibility is of critical context. However, reproducibility does not assess the validity of the brain connections. Phantom studies provide concrete quantitative comparisons of methods relative to absolute ground truths, yet do no capture variabilities because of in vivo physiological factors. The ISMRM 2017 TraCED challenge was created to fulfill the gap. STUDY TYPE: A systematic review of algorithms and tract reproducibility studies. SUBJECTS: Single healthy volunteers. FIELD STRENGTH/SEQUENCE: 3.0T, two different scanners by the same manufacturer. The multishell acquisition included b-values of 1000, 2000, and 3000 s/mm2 with 20, 45, and 64 diffusion gradient directions per shell, respectively. ASSESSMENT: Nine international groups submitted 46 tractography algorithm entries each consisting 16 tracts per scan. The algorithms were assessed using intraclass correlation (ICC) and the Dice similarity measure. STATISTICAL TESTS: Containment analysis was performed to assess if the submitted algorithms had containment within tracts of larger volume submissions. This also serves the purpose to detect if spurious submissions had been made. RESULTS: The top five submissions had high ICC and Dice >0.88. Reproducibility was high within the top five submissions when assessed across sessions or across scanners: 0.87-0.97. Containment analysis shows that the top five submissions are contained within larger volume submissions. From the total of 16 tracts as an outcome relatively the number of tracts with high, moderate, and low reproducibility were 8, 4, and 4. DATA CONCLUSION: The different methods clearly result in fundamentally different tract structures at the more conservative specificity choices. Data and challenge infrastructure remain available for continued analysis and provide a platform for comparison. LEVEL OF EVIDENCE: 5 Technical Efficacy Stage: 1 J. Magn. Reson. Imaging 2020;51:234-249.

3D whole brain segmentation using spatially localized atlas network tiles.

Detailed whole brain segmentation is an essential quantitative technique in medical image analysis, which provides a non-invasive way of measuring brain regions from a clinical acquired structural magnetic resonance imaging (MRI). Recently, deep convolution neural network (CNN) has been applied to whole brain segmentation. However, restricted by current GPU memory, 2D based methods, downsampling based 3D CNN methods, and patch-based high-resolution 3D CNN methods have been the de facto standard solutions. 3D patch-based high resolution methods typically yield superior performance among CNN approaches on detailed whole brain segmentation (>100 labels), however, whose performance are still commonly inferior compared with state-of-the-art multi-atlas segmentation methods (MAS) due to the following challenges: (1) a single network is typically used to learn both spatial and contextual information for the patches, (2) limited manually traced whole brain volumes are available (typically less than 50) for training a network. In this work, we propose the spatially localized atlas network tiles (SLANT) method to distribute multiple independent 3D fully convolutional networks (FCN) for high-resolution whole brain segmentation. To address the first challenge, multiple spatially distributed networks were used in the SLANT method, in which each network learned contextual information for a fixed spatial location. To address the second challenge, auxiliary labels on 5111 initially unlabeled scans were created by multi-atlas segmentation for training. Since the method integrated multiple traditional medical image processing methods with deep learning, we developed a containerized pipeline to deploy the end-to-end solution. From the results, the proposed method achieved superior performance compared with multi-atlas segmentation methods, while reducing the computational time from >30 h to 15 min. The method has been made available in open source (https://github.com/MASILab/SLANTbrainSeg).

Limits to anatomical accuracy of diffusion tractography using modern approaches.

Diffusion MRI fiber tractography is widely used to probe the structural connectivity of the brain, with a range of applications in both clinical and basic neuroscience. Despite widespread use, tractography has well-known pitfalls that limits the anatomical accuracy of this technique. Numerous modern methods have been developed to address these shortcomings through advances in acquisition, modeling, and computation. To test whether these advances improve tractography accuracy, we organized the 3-D Validation of Tractography with Experimental MRI (3D-VoTEM) challenge at the ISBI 2018 conference. We made available three unique independent tractography validation datasets - a physical phantom and two ex vivo brain specimens - resulting in 176 distinct submissions from 9 research groups. By comparing results over a wide range of fiber complexities and algorithmic strategies, this challenge provides a more comprehensive assessment of tractography's inherent limitations than has been reported previously. The central results were consistent across all sub-challenges in that, despite advances in tractography methods, the anatomical accuracy of tractography has not dramatically improved in recent years. Taken together, our results independently confirm findings from decades of tractography validation studies, demonstrate inherent limitations in reconstructing white matter pathways using diffusion MRI data alone, and highlight the need for alternative or combinatorial strategies to accurately map the fiber pathways of the brain.

Empirical single sample quantification of bias and variance in Q-ball imaging.

PURPOSE: The bias and variance of high angular resolution diffusion imaging methods have not been thoroughly explored in the literature and may benefit from the simulation extrapolation (SIMEX) and bootstrap techniques to estimate bias and variance of high angular resolution diffusion imaging metrics. METHODS: The SIMEX approach is well established in the statistics literature and uses simulation of increasingly noisy data to extrapolate back to a hypothetical case with no noise. The bias of calculated metrics can then be computed by subtracting the SIMEX estimate from the original pointwise measurement. The SIMEX technique has been studied in the context of diffusion imaging to accurately capture the bias in fractional anisotropy measurements in DTI. Herein, we extend the application of SIMEX and bootstrap approaches to characterize bias and variance in metrics obtained from a Q-ball imaging reconstruction of high angular resolution diffusion imaging data. RESULTS: The results demonstrate that SIMEX and bootstrap approaches provide consistent estimates of the bias and variance of generalized fractional anisotropy, respectively. The RMSE for the generalized fractional anisotropy estimates shows a 7% decrease in white matter and an 8% decrease in gray matter when compared with the observed generalized fractional anisotropy estimates. On average, the bootstrap technique results in SD estimates that are approximately 97% of the true variation in white matter, and 86% in gray matter. CONCLUSION: Both SIMEX and bootstrap methods are flexible, estimate population characteristics based on single scans, and may be extended for bias and variance estimation on a variety of high angular resolution diffusion imaging metrics.

Towards Portable Large-Scale Image Processing with High-Performance Computing.

High-throughput, large-scale medical image computing demands tight integration of high-performance computing (HPC) infrastructure for data storage, job distribution, and image processing. The Vanderbilt University Institute for Imaging Science (VUIIS) Center for Computational Imaging (CCI) has constructed a large-scale image storage and processing infrastructure that is composed of (1) a large-scale image database using the eXtensible Neuroimaging Archive Toolkit (XNAT), (2) a content-aware job scheduling platform using the Distributed Automation for XNAT pipeline automation tool (DAX), and (3) a wide variety of encapsulated image processing pipelines called "spiders." The VUIIS CCI medical image data storage and processing infrastructure have housed and processed nearly half-million medical image volumes with Vanderbilt Advanced Computing Center for Research and Education (ACCRE), which is the HPC facility at the Vanderbilt University. The initial deployment was natively deployed (i.e., direct installations on a bare-metal server) within the ACCRE hardware and software environments, which lead to issues of portability and sustainability. First, it could be laborious to deploy the entire VUIIS CCI medical image data storage and processing infrastructure to another HPC center with varying hardware infrastructure, library availability, and software permission policies. Second, the spiders were not developed in an isolated manner, which has led to software dependency issues during system upgrades or remote software installation. To address such issues, herein, we describe recent innovations using containerization techniques with XNAT/DAX which are used to isolate the VUIIS CCI medical image data storage and processing infrastructure from the underlying hardware and software environments. The newly presented XNAT/DAX solution has the following new features: (1) multi-level portability from system level to the application level, (2) flexible and dynamic software development and expansion, and (3) scalable spider deployment compatible with HPC clusters and local workstations.

Contribution of new and chronic cortical lesions to disability accrual in multiple sclerosis.

Cortical lesions are common in multiple sclerosis and are associated with disability and progressive disease. We asked whether cortical lesions continue to form in people with stable white matter lesions and whether the association of cortical lesions with worsening disability relates to pre-existing or new cortical lesions. Fifty adults with multiple sclerosis and no new white matter lesions in the year prior to enrolment (33 relapsing-remitting and 17 progressive) and a comparison group of nine adults who had formed at least one new white matter lesion in the year prior to enrolment (active relapsing-remitting) were evaluated annually with 7 tesla (T) brain MRI and 3T brain and spine MRI for 2 years, with clinical assessments for 3 years. Cortical lesions and paramagnetic rim lesions were identified on 7T images. Seven total cortical lesions formed in 3/30 individuals in the stable relapsing-remitting group (median 0, range 0-5), four total cortical lesions formed in 4/17 individuals in the progressive group (median 0, range 0-1), and 16 cortical lesions formed in 5/9 individuals in the active relapsing-remitting group (median 1, range 0-10, stable relapsing-remitting versus progressive versus active relapsing-remitting P = 0.006). New cortical lesions were not associated with greater change in any individual disability measure or in a composite measure of disability worsening (worsening Expanded Disability Status Scale or 9-hole peg test or 25-foot timed walk). Individuals with at least three paramagnetic rim lesions had a greater increase in cortical lesion volume over time (median 16 µl, range -61 to 215 versus median 1 µl, range -24 to 184, P = 0.007), but change in lesion volume was not associated with disability change. Baseline cortical lesion volume was higher in people with worsening disability (median 1010 µl, range 13-9888 versus median 267 µl, range 0-3539, P = 0.001, adjusted for age and sex) and in individuals with relapsing-remitting multiple sclerosis who subsequently transitioned to secondary progressive multiple sclerosis (median 2183 µl, range 270-9888 versus median 321 µl, range 0-6392 in those who remained relapsing-remitting, P = 0.01, adjusted for age and sex). Baseline white matter lesion volume was not associated with worsening disability or transition from relapsing-remitting to secondary progressive multiple sclerosis. Cortical lesion formation is rare in people with stable white matter lesions, even in those with worsening disability. Cortical but not white matter lesion burden predicts disability worsening, suggesting that disability progression is related to long-term effects of cortical lesions that form early in the disease, rather than to ongoing cortical lesion formation.

T1 /T2 ratio from 3T MRI improves multiple sclerosis cortical lesion contrast.

BACKGROUND AND PURPOSE: Cortical demyelinated lesions are prevalent in multiple sclerosis (MS), associated with disability, and have recently been incorporated into MS diagnostic criteria. Presently, advanced and ultrahigh-field MRIs-not routinely available in clinical practice-are the most sensitive methods for detection of cortical lesions. Approaches utilizing MRI sequences obtainable in routine clinical practice remain an unmet need. We plan to assess the sensitivity of the ratio of T1 -weighted and T2 -weighted (T1 /T2 ) signal intensity for focal cortical lesions in comparison to other high-field imaging methods. METHODS: 3-Tesla and 7-Tesla MRI collected from 10 adults with MS were included in the study. T1 /T2 images were calculated by dividing 3T T1 -weighted (T1 w) images by 3T T2 -weighted (T2 w) fluid-attenuated inversion recovery images for each participant. A total of 614 cortical lesions were identified using 7T T2 *w and T1 w images and corresponding voxels were assessed on registered 3T images. Signal intensities were compared across 3T imaging sequences, including T1 /T2 , T1 w, T2 w, and inversion recovery susceptibility-weighted imaging with enhanced T2 weighting (IR-SWIET) images. RESULTS: T1 /T2 images demonstrated a larger contrast between median lesional and nonlesional cortical signal intensity (median ratio = 1.29, range: 1.19-1.38) when compared to T1 w (1.01, 0.97-1.10, p < .002), T2 w (1.17, 1.07-1.26, p < .002), and IR-SWIET (1.21, 1.01-1.29, p < .03). CONCLUSION: T1 /T2 images are sensitive to cortical lesions. Approaches incorporating T1 /T2 could improve the accessibility of cortical lesion detection in research settings and clinical practice.

Cortical lesions uniquely predict motor disability accrual and form rarely in the absence of new white matter lesions in multiple sclerosis.

Background and objectives: Cortical lesions (CL) are common in multiple sclerosis (MS) and associate with disability and progressive disease. We asked whether CL continue to form in people with stable white matter lesions (WML) and whether the association of CL with worsening disability relates to pre-existing or new CL. Methods: A cohort of adults with MS were evaluated annually with 7 tesla (T) brain magnetic resonance imaging (MRI) and 3T brain and spine MRI for 2 years, and clinical assessments for 3 years. CL were identified on 7T images at each timepoint. WML and brain tissue segmentation were performed using 3T images at baseline and year 2. Results: 59 adults with MS had ≥1 7T follow-up visit (mean follow-up time 2±0.5 years). 9 had "active" relapsing-remitting MS (RRMS), defined as new WML in the year prior to enrollment. Of the remaining 50, 33 had "stable" RRMS, 14 secondary progressive MS (SPMS), and 3 primary progressive MS. 16 total new CL formed in the active RRMS group (median 1, range 0-10), 7 in the stable RRMS group (median 0, range 0-5), and 4 in the progressive MS group (median 0, range 0-1) (p=0.006, stable RR vs PMS p=0.88). New CL were not associated with greater change in any individual disability measure or in a composite measure of disability worsening (worsening Expanded Disability Status Scale or 9-hole peg test or 25-foot timed walk). Baseline CL volume was higher in people with worsening disability (median 1010µl, range 13-9888 vs median 267µl, range 0-3539, p=0.001, adjusted for age and sex) and in individuals with RRMS who subsequently transitioned to SPMS (median 2183µl, range 270-9888 vs median 321µl, range 0-6392 in those who remained RRMS, p=0.01, adjusted for age and sex). Baseline WML volume was not associated with worsening disability or transition from RRMS to SPMS. Discussion: CL formation is rare in people with stable WML, even in those with worsening disability. CL but not WML burden predicts future worsening of disability, suggesting that the relationship between CL and disability progression is related to long-term effects of lesions that form in the earlier stages of disease, rather than to ongoing lesion formation.

Logistic Regression-Based Model Is More Efficient Than U-Net Model for Reliable Whole Brain Magnetic Resonance Imaging Segmentation.

OBJECTIVES: Automated whole brain segmentation from magnetic resonance images is of great interest for the development of clinically relevant volumetric markers for various neurological diseases. Although deep learning methods have demonstrated remarkable potential in this area, they may perform poorly in nonoptimal conditions, such as limited training data availability. Manual whole brain segmentation is an incredibly tedious process, so minimizing the data set size required for training segmentation algorithms may be of wide interest. The purpose of this study was to compare the performance of the prototypical deep learning segmentation architecture (U-Net) with a previously published atlas-free traditional machine learning method, Classification using Derivative-based Features (C-DEF) for whole brain segmentation, in the setting of limited training data. MATERIALS AND METHODS: C-DEF and U-Net models were evaluated after training on manually curated data from 5, 10, and 15 participants in 2 research cohorts: (1) people living with clinically diagnosed HIV infection and (2) relapsing-remitting multiple sclerosis, each acquired at separate institutions, and between 5 and 295 participants' data using a large, publicly available, and annotated data set of glioblastoma and lower grade glioma (brain tumor segmentation). Statistics was performed on the Dice similarity coefficient using repeated-measures analysis of variance and Dunnett-Hsu pairwise comparison. RESULTS: C-DEF produced better segmentation than U-Net in lesion (29.2%-38.9%) and cerebrospinal fluid (5.3%-11.9%) classes when trained with data from 15 or fewer participants. Unlike C-DEF, U-Net showed significant improvement when increasing the size of the training data (24%-30% higher than baseline). In the brain tumor segmentation data set, C-DEF produced equivalent or better segmentations than U-Net for enhancing tumor and peritumoral edema regions across all training data sizes explored. However, U-Net was more effective than C-DEF for segmentation of necrotic/non-enhancing tumor when trained on 10 or more participants, probably because of the inconsistent signal intensity of the tissue class. CONCLUSIONS: These results demonstrate that classical machine learning methods can produce more accurate brain segmentation than the far more complex deep learning methods when only small or moderate amounts of training data are available (n ≤ 15). The magnitude of this advantage varies by tissue and cohort, while U-Net may be preferable for deep gray matter and necrotic/non-enhancing tumor segmentation, particularly with larger training data sets (n ≥ 20). Given that segmentation models often need to be retrained for application to novel imaging protocols or pathology, the bottleneck associated with large-scale manual annotation could be avoided with classical machine learning algorithms, such as C-DEF.

Progressive multifocal leukoencephalopathy lesion and brain parenchymal segmentation from MRI using serial deep convolutional neural networks.

Progressive multifocal leukoencephalopathy (PML) is a rare opportunistic brain infection caused by the JC virus and associated with substantial morbidity and mortality. Accurate MRI assessment of PML lesion burden and brain parenchymal atrophy is of decisive value in monitoring the disease course and response to therapy. However, there are currently no validated automatic methods for quantification of PML lesion burden or associated parenchymal volume loss. Furthermore, manual brain or lesion delineations can be tedious, require the use of valuable time resources by radiologists or trained experts, and are often subjective. In this work, we introduce JCnet (named after the causative viral agent), an end-to-end, fully automated method for brain parenchymal and lesion segmentation in PML using consecutive 3D patch-based convolutional neural networks. The network architecture consists of multi-view feature pyramid networks with hierarchical residual learning blocks containing embedded batch normalization and nonlinear activation functions. The feature maps across the bottom-up and top-down pathways of the feature pyramids are merged, and an output probability membership generated through convolutional pathways, thus rendering the method fully convolutional. Our results show that this approach outperforms and improves longitudinal consistency compared to conventional, state-of-the-art methods of healthy brain and multiple sclerosis lesion segmentation, utilized here as comparators given the lack of available methods validated for use in PML. The ability to produce robust and accurate automated measures of brain atrophy and lesion segmentation in PML is not only valuable clinically but holds promise toward including standardized quantitative MRI measures in clinical trials of targeted therapies. Code is available at: https://github.com/omarallouz/JCnet.

AUTOMATIC LABELING OF CORTICAL SULCI USING SPHERICAL CONVOLUTIONAL NEURAL NETWORKS IN A DEVELOPMENTAL COHORT.

In this paper, we present the automatic labeling framework for sulci in the human lateral prefrontal cortex (PFC). We adapt an existing spherical U-Net architecture with our recent surface data augmentation technique to improve the sulcal labeling accuracy in a developmental cohort. Specifically, our framework consists of the following key components: (1) augmented geometrical features being generated during cortical surface registration, (2) spherical U-Net architecture to efficiently fit the augmented features, and (3) postrefinement of sulcal labeling by optimizing spatial coherence via a graph cut technique. We validate our method on 30 healthy subjects with manual labeling of sulcal regions within PFC. In the experiments, we demonstrate significantly improved labeling performance (0.7749) in mean Dice overlap compared to that of multi-atlas (0.6410) and standard spherical U-Net (0.7011) approaches, respectively (p < 0.05). Additionally, the proposed method achieves a full set of sulcal labels in 20 seconds in this developmental cohort.

Improving human cortical sulcal curve labeling in large scale cross-sectional MRI using deep neural networks.

BACKGROUND: Human cortical primary sulci are relatively stable landmarks and commonly observed across the population. Despite their stability, the primary sulci exhibit phenotypic variability. NEW METHOD: We propose a fully automated pipeline that integrates both sulcal curve extraction and labeling. In this study, we use a large normal control population (n = 1424) to train neural networks for accurately labeling the primary sulci. Briefly, we use sulcal curve distance map, surface parcellation, mean curvature and spectral features to delineate their sulcal labels. We evaluate the proposed method with 8 primary sulcal curves in the left and right hemispheres compared to an established multi-atlas curve labeling method. RESULTS: Sulcal labels by the proposed method reasonably well agree with manual labeling. The proposed method outperforms the existing multi-atlas curve labeling method. COMPARISON WITH EXISTING METHOD: Significantly improved sulcal labeling results are achieved with over 12.5 and 20.6 percent improvement on labeling accuracy in the left and right hemispheres, respectively compared to that of a multi-atlas curve labeling method in eight curves (p≪0.001, two-sample t-test). CONCLUSION: The proposed method offers a computationally efficient and robust labeling of major sulci.

Coronary Calcium Detection using 3D Attention Identical Dual Deep Network Based on Weakly Supervised Learning.

Coronary artery calcium (CAC) is biomarker of advanced subclinical coronary artery disease and predicts myocardial infarction and death prior to age 60 years. The slice-wise manual delineation has been regarded as the gold standard of coronary calcium detection. However, manual efforts are time and resource consuming and even impracticable to be applied on large-scale cohorts. In this paper, we propose the attention identical dual network (AID-Net) to perform CAC detection using scan-rescan longitudinal non-contrast CT scans with weakly supervised attention by only using per scan level labels. To leverage the performance, 3D attention mechanisms were integrated into the AID-Net to provide complementary information for classification tasks. Moreover, the 3D Gradient-weighted Class Activation Mapping (Grad-CAM) was also proposed at the testing stage to interpret the behaviors of the deep neural network. 5075 non-contrast chest CT scans were used as training, validation and testing datasets. Baseline performance was assessed on the same cohort. From the results, the proposed AID-Net achieved the superior performance on classification accuracy (0.9272) and AUC (0.9627).

Enabling Multi-Shell b-Value Generalizability of Data-Driven Diffusion Models with Deep SHORE.

Intra-voxel models of the diffusion signal are essential for interpreting organization of the tissue environment at micrometer level with data at millimeter resolution. Recent advances in data driven methods have enabled direct comparison and optimization of methods for in-vivo data with externally validated histological sections with both 2-D and 3-D histology. Yet, all existing methods make limiting assumptions of either (1) model-based linkages between b-values or (2) limited associations with single shell data. We generalize prior deep learning models that used single shell spherical harmonic transforms to integrate the recently developed simple harmonic oscillator reconstruction (SHORE) basis. To enable learning on the SHORE manifold, we present an alternative formulation of the fiber orientation distribution (FOD) object using the SHORE basis while representing the observed diffusion weighted data in the SHORE basis. To ensure consistency of hyper-parameter optimization for SHORE, we present our Deep SHORE approach to learn on a data-optimized manifold. Deep SHORE is evaluated with eight-fold cross-validation of a preclinical MRI-histology data with four b-values. Generalizability of in-vivo human data is evaluated on two separate 3T MRI scanners. Specificity in terms of angular correlation (ACC) with the preclinical data improved on single shell: 0.78 relative to 0.73 and 0.73, multi-shell: 0.80 relative to 0.74 (p < 0.001). In the in-vivo human data, Deep SHORE was more consistent across scanners with 0.63 relative to other multi-shell methods 0.39, 0.52 and 0.57 in terms of ACC. In conclusion, Deep SHORE is a promising method to enable data driven learning with DW-MRI under conditions with varying b-values, number of diffusion shells, and gradient directions per shell.

A deep learning approach to estimation of subject-level bias and variance in high angular resolution diffusion imaging.

The ability to evaluate empirical diffusion MRI acquisitions for quality and to correct the resulting imaging metrics allows for improved inference and increased replicability. Previous work has shown promise for estimation of bias and variance of generalized fractional anisotropy (GFA) but comes at the price of computational complexity. This paper aims to provide methods for estimating GFA, bias of GFA and standard deviation of GFA quickly and accurately. In order to provide a method for bias and variance estimation that can return results faster than the previously studied statistical techniques, three deep, fully-connected neural networks are developed for GFA, bias of GFA, and standard deviation of GFA. The results of these networks are compared to the observed values of the metrics as well as those fit from the statistical techniques (i.e. Simulation Extrapolation (SIMEX) for bias estimation and wild bootstrap for variance estimation). Our GFA network provides predictions that are closer to the true GFA values than a Q-ball fit of the observed data (root-mean-square error (RMSE) 0.0077 vs 0.0082, p < .001). The bias network also shows statistically significant improvement in comparison to the SIMEX-estimated error of GFA (RMSE 0.0071 vs. 0.01, p < .001).

Registration-based image enhancement improves multi-atlas segmentation of the thalamic nuclei and hippocampal subfields.

Magnetic resonance imaging (MRI) is an important tool for analysis of deep brain grey matter structures. However, analysis of these structures is limited due to low intensity contrast typically found in whole brain imaging protocols. Herein, we propose a big data registration-enhancement (BDRE) technique to augment the contrast of deep brain structures using an efficient large-scale non-rigid registration strategy. Direct validation is problematic given a lack of ground truth data. Rather, we validate the usefulness and impact of BDRE for multi-atlas (MA) segmentation on two sets of structures of clinical interest: the thalamic nuclei and hippocampal subfields. The experimental design compares algorithms using T1-weighted 3 T MRI for both structures (and additional 7 T MRI for the thalamic nuclei) with an algorithm using BDRE. As baseline comparisons, a recent denoising (DN) technique and a super-resolution (SR) method are used to preprocess the original 3 T MRI. The performance of each MA segmentation is evaluated by the Dice similarity coefficient (DSC). BDRE significantly improves mean segmentation accuracy over all methods tested for both thalamic nuclei (3 T imaging: 9.1%; 7 T imaging: 15.6%; DN: 6.9%; SR: 16.2%) and hippocampal subfields (3 T T1 only: 8.7%; DN: 8.4%; SR: 8.6%). We also present DSC performance for each thalamic nucleus and hippocampal subfield and show that BDRE can help MA segmentation for individual thalamic nuclei and hippocampal subfields. This work will enable large-scale analysis of clinically relevant deep brain structures from commonly acquired T1 images.

Cortical Surface Parcellation using Spherical Convolutional Neural Networks.

We present cortical surface parcellation using spherical deep convolutional neural networks. Traditional multi-atlas cortical surface parcellation requires inter-subject surface registration using geometric features with slow processing speed on a single subject (2-3 hours). Moreover, even optimal surface registration does not necessarily produce optimal cortical parcellation as parcel boundaries are not fully matched to the geometric features. In this context, a choice of training features is important for accurate cortical parcellation. To utilize the networks efficiently, we propose cortical parcellation-specific input data from an irregular and complicated structure of cortical surfaces. To this end, we align ground-truth cortical parcel boundaries and use their resulting deformation fields to generate new pairs of deformed geometric features and parcellation maps. To extend the capability of the networks, we then smoothly morph cortical geometric features and parcellation maps using the intermediate deformation fields. We validate our method on 427 adult brains for 49 labels. The experimental results show that our method outperforms traditional multi-atlas and naive spherical U-Net approaches, while achieving full cortical parcellation in less than a minute.

Splenomegaly Segmentation on Multi-Modal MRI Using Deep Convolutional Networks.

The findings of splenomegaly, abnormal enlargement of the spleen, is a non-invasive clinical biomarker for liver and spleen diseases. Automated segmentation methods are essential to efficiently quantify splenomegaly from clinically acquired abdominal magnetic resonance imaging (MRI) scans. However, the task is challenging due to: 1) large anatomical and spatial variations of splenomegaly; 2) large inter- and intra-scan intensity variations on multi-modal MRI; and 3) limited numbers of labeled splenomegaly scans. In this paper, we propose the Splenomegaly Segmentation Network (SS-Net) to introduce the deep convolutional neural network (DCNN) approaches in multi-modal MRI splenomegaly segmentation. Large convolutional kernel layers were used to address the spatial and anatomical variations, while the conditional generative adversarial networks were employed to leverage the segmentation performance of SS-Net in an end-to-end manner. A clinically acquired cohort containing both T1-weighted (T1w) and T2-weighted (T2w) MRI splenomegaly scans was used to train and evaluate the performance of multi-atlas segmentation (MAS), 2D DCNN networks, and a 3-D DCNN network. From the experimental results, the DCNN methods achieved superior performance to the state-of-the-art MAS method. The proposed SS-Net method has achieved the highest median and mean Dice scores among the investigated baseline DCNN methods.