MRI FLAIR lesion segmentation in multiple sclerosis: Does automated segmentation hold up with manual annotation?

Fully automated longitudinal segmentation of new or enlarged multiple sclerosis lesions using 3D convolutional neural networks

•

A fully automated segmentation of new or enlarged multiple sclerosis (MS) lesions.

•

3D convolutional neural network (CNN) with U-net-like encoder-decoder architecture.

•

Simultaneous processing of baseline and follow-up scan of the same patient.

•

Trained on 3253 patient data from over 103 different MR scanners.

•

Fast (<1min), robust algorithm with segmentation results in inter-rater variability.

The quantification of new or enlarged lesions from follow-up MRI scans is an important surrogate of clinical disease activity in patients with multiple sclerosis (MS). Not only is manual segmentation time consuming, but inter-rater variability is high. Currently, only a few fully automated methods are available. We address this gap in the field by employing a 3D convolutional neural network (CNN) with encoder-decoder architecture for fully automatic longitudinal lesion segmentation.

Input data consist of two fluid attenuated inversion recovery (FLAIR) images (baseline and follow-up) per patient. Each image is entered into the encoder and the feature maps are concatenated and then fed into the decoder. The output is a 3D mask indicating new or enlarged lesions (compared to the baseline scan). The proposed method was trained on 1809 single point and 1444 longitudinal patient data sets and then validated on 185 independent longitudinal data sets from two different scanners. From the two validation data sets, manual segmentations were available from three experienced raters, respectively. The performance of the proposed method was compared to the open source Lesion Segmentation Toolbox (LST), which is a current state-of-art longitudinal lesion segmentation method.

The mean lesion-wise inter-rater sensitivity was 62%, while the mean inter-rater number of false positive (FP) findings was 0.41 lesions per case. The two validated algorithms showed a mean sensitivity of 60% (CNN), 46% (LST) and a mean FP of 0.48 (CNN), 1.86 (LST) per case. Sensitivity and number of FP were not significantly different (p < 0.05) between the CNN and manual raters.

New or enlarged lesions counted by the CNN algorithm appeared to be comparable with manual expert ratings. The proposed algorithm seems to outperform currently available approaches, particularly LST. The high inter-rater variability in case of manual segmentation indicates the complexity of identifying new or enlarged lesions. An automated CNN-based approach can quickly provide an independent and deterministic assessment of new or enlarged lesions from baseline to follow-up scans with acceptable reliability.

Accuracy and practical aspects of semi- and fully automatic segmentation methods for resected brain areas

Purpose

Precise segmentation of brain lesions is essential for neurological research. Specifically, resection volume estimates can aid in the assessment of residual postoperative tissue, e.g. following surgery for glioma. Furthermore, behavioral lesion-symptom mapping in epilepsy relies on accurate delineation of surgical lesions. We sought to determine whether semi- and fully automatic segmentation methods can be applied to resected brain areas and which approach provides the most accurate and cost-efficient results.

Methods

We compared a semi-automatic (ITK-SNAP) with a fully automatic (lesion_GNB) method for segmentation of resected brain areas in terms of accuracy with manual segmentation serving as reference. Additionally, we evaluated processing times of all three methods. We used T1w, MRI-data of epilepsy patients (n = 27; 11 m; mean age 39 years, range 16–69) who underwent temporal lobe resections (17 left).

Results

The semi-automatic approach yielded superior accuracy (p < 0.001) with a median Dice similarity coefficient (mDSC) of 0.78 and a median average Hausdorff distance (maHD) of 0.44 compared with the fully automatic approach (mDSC 0.58, maHD 1.32). There was no significant difference between the median percent volume difference of the two approaches (p > 0.05). Manual segmentation required more human input (30.41 min/subject) and therefore inferring significantly higher costs than semi- (3.27 min/subject) or fully automatic approaches (labor and cost approaching zero).

Conclusion

Semi-automatic segmentation offers the most accurate results in resected brain areas with a moderate amount of human input, thus representing a viable alternative compared with manual segmentation, especially for studies with large patient cohorts.

TAPAS: A Thresholding Approach for Probability Map Automatic Segmentation in Multiple Sclerosis

Total brain white matter lesion (WML) volume is the most widely established magnetic resonance imaging (MRI) outcome measure in studies of multiple sclerosis (MS). To estimate WML volume, there are a number of automatic segmentation methods available, yet manual delineation remains the gold standard approach. Automatic approaches often yield a probability map to which a threshold is applied to create lesion segmentation masks. Unfortunately, few approaches systematically determine the threshold employed; many methods use a manually selected threshold, thus introducing human error and bias into the automated procedure. In this study, we propose and validate an automatic thresholding algorithm, Thresholding Approach for Probability Map Automatic Segmentation in Multiple Sclerosis (TAPAS), to obtain subject-specific threshold estimates for probability map automatic segmentation of T2-weighted (T2) hyperintense WMLs. Using multimodal MRI, the proposed method applies an automatic segmentation algorithm to obtain probability maps. We obtain the true subject-specific threshold that maximizes the Sørensen-Dice similarity coefficient (DSC). Then the subject-specific thresholds are modeled on a naive estimate of volume using a generalized additive model. Applying this model, we predict a subject-specific threshold in data not used for training. We ran a Monte Carlo-resampled split-sample cross-validation (100 validation sets) using two data sets: the first obtained from the Johns Hopkins Hospital (JHH) on a Philips 3 Tesla (3T) scanner (n = 94) and a second collected at the Brigham and Women's Hospital (BWH) using a Siemens 3T scanner (n = 40). By means of the proposed automated technique, in the JHH data we found an average reduction in subject-level absolute error of 0.1 mL per one mL increase in manual volume. Using Bland-Altman analysis, we found that volumetric bias associated with group-level thresholding was mitigated when applying TAPAS. The BWH data showed similar absolute error estimates using group-level thresholding or TAPAS likely since Bland-Altman analyses indicated no systematic biases associated with group or TAPAS volume estimates. The current study presents the first validated fully automated method for subject-specific threshold prediction to segment brain lesions.

Multi-Disease Segmentation of Gliomas and White Matter Hyperintensities in the BraTS Data Using a 3D Convolutional Neural Network

An important challenge in segmenting real-world biomedical imaging data is the presence of multiple disease processes within individual subjects. Most adults above age 60 exhibit a variable degree of small vessel ischemic disease, as well as chronic infarcts, which will manifest as white matter hyperintensities (WMH) on brain MRIs. Subjects diagnosed with gliomas will also typically exhibit some degree of abnormal T2 signal due to WMH, rather than just due to tumor. We sought to develop a fully automated algorithm to distinguish and quantify these distinct disease processes within individual subjects’ brain MRIs. To address this multi-disease problem, we trained a 3D U-Net to distinguish between abnormal signal arising from tumors vs. WMH in the 3D multi-parametric MRI (mpMRI, i.e., native T1-weighted, T1-post-contrast, T2, T2-FLAIR) scans of the International Brain Tumor Segmentation (BraTS) 2018 dataset (n_training = 285, n_validation = 66). Our trained neuroradiologist manually annotated WMH on the BraTS training subjects, finding that 69% of subjects had WMH. Our 3D U-Net model had a 4-channel 3D input patch (80 × 80 × 80) from mpMRI, four encoding and decoding layers, and an output of either four [background, active tumor (AT), necrotic core (NCR), peritumoral edematous/infiltrated tissue (ED)] or five classes (adding WMH as the fifth class). For both the four- and five-class output models, the median Dice for whole tumor (WT) extent (i.e., union of AT, ED, NCR) was 0.92 in both training and validation sets. Notably, the five-class model achieved significantly (p = 0.002) lower/better Hausdorff distances for WT extent in the training subjects. There was strong positive correlation between manually segmented and predicted volumes for WT (r = 0.96) and WMH (r = 0.89). Larger lesion volumes were positively correlated with higher/better Dice scores for WT (r = 0.33), WMH (r = 0.34), and across all lesions (r = 0.89) on a log(10) transformed scale. While the median Dice for WMH was 0.42 across training subjects with WMH, the median Dice was 0.62 for those with at least 5 cm³ of WMH. We anticipate the development of computational algorithms that are able to model multiple diseases within a single subject will be a critical step toward translating and integrating artificial intelligence systems into the heterogeneous real-world clinical workflow.

Tsallis-Entropy Segmentation through MRF and Alzheimer anatomic reference for Brain Magnetic Resonance Parcellation

Quantifying the intracranial tissue volume changes in magnetic resonance imaging (MRI) assists specialists to analyze the effects of natural or pathological changes. Since these changes can be subtle, the accuracy of the automatic compartmentalization method is always criticized by specialists. We propose and then evaluate an automatic segmentation method based on modified q-entropy (Mqe) through a modified Markov Random Field (MMRF) enhanced by Alzheimer anatomic reference (AAR) to provide a high accuracy brain tissues parcellation approach (Mqe-MMRF). We underwent two strategies to evaluate Mqe-MMRF; a simulation of different levels of noise and non-uniformity effect on MRI data (7 subjects) and a set of twenty MRI data available from MRBrainS13 as patient brain tissue segmentation challenge. We accessed eleven quality metrics compared to reference tissues delineations to evaluate Mqe-MMRF. MRI segmentation scores decreased by only 4.6% on quality metrics after noise and non-uniformity simulations of 40% and 9%, respectively. We found significant mean improvements in the metrics of the five training subjects, for whole-brain 0.86%, White Matter 3.20%, Gray Matter 3.99%, and Cerebrospinal Fluid 4.16% (p-values < 0.02) when Mqe-MMRF compared to the other reference methods. We also processed the Mqe-MMRF on 15 evaluation subjects group from MRBrainS13 online challenge, and the results held a higher rank than the reference tools; FreeSurfer, SPM, and FSL. Since the proposed method improved the precision of brain segmentation, specifically, for GM, and thus one can use it in quantitative and morphological brain studies.

Validation and comparison of two automated methods for quantifying brain white matter hyperintensities of presumed vascular origin

Objectives

White matter hyperintensities (WMH) are a common imaging finding indicative of cerebral small vessel disease. Lesion segmentation algorithms have been developed to overcome issues arising from visual rating scales. In this study, we evaluated two automated methods and compared them to visual and manual segmentation to determine the most robust algorithm provided by the open-source Lesion Segmentation Toolbox (LST).

Methods

We compared WMH data from visual ratings (Scheltens’ scale) with those derived from algorithms provided within LST. We then compared spatial and volumetric WMH data derived from manually-delineated lesion maps with WMH data and lesion maps provided by the LST algorithms.

Results

We identified optimal initial thresholds for algorithms provided by LST compared with visual ratings (Lesion Growth Algorithm (LGA): initial κ and lesion probability thresholds, 0.5; Lesion Probability Algorithm (LPA) lesion probability threshold, 0.65). LGA was found to perform better then LPA compared with manual segmentation.

Conclusion

LGA appeared to be the most suitable algorithm for quantifying WMH in relation to cerebral small vessel disease, compared with Scheltens’ score and manual segmentation. LGA offers a user-friendly, effective WMH segmentation method in the research environment.

Diffusion basis spectrum imaging for identifying pathologies in MS subtypes

Diffusion basis spectrum imaging (DBSI) combines discrete anisotropic diffusion tensors and the spectrum of isotropic diffusion tensors to model the underlying multiple sclerosis (MS) pathologies. We used clinical MS subtypes as a surrogate of underlying pathologies to assess DBSI as a biomarker of pathology in 55 individuals with MS. Restricted isotropic fraction (reflecting cellularity) and fiber fraction (representing apparent axonal density) were the most important DBSI metrics to classify MS using brain white matter lesions. These DBSI metrics outperformed lesion volume. When analyzing the normal‐appearing corpus callosum, the most significant DBSI metrics were fiber fraction, radial diffusivity (reflecting myelination), and nonrestricted isotropic fraction (representing edema). This study provides preliminary evidence supporting the ability of DBSI as a potential noninvasive biomarker of MS neuropathology.

White Matter Abnormalities in Multiple Sclerosis Evaluated by Quantitative Synthetic MRI, Diffusion Tensor Imaging, and Neurite Orientation Dispersion and Density Imaging

BACKGROUND AND PURPOSE

A number of MR-derived quantitative metrics have been suggested to assess the pathophysiology of MS, but the reports about combined analyses of these metrics are scarce. Our aim was to assess the spatial distribution of parameters for white matter myelin and axon integrity in patients with relapsing-remitting MS by multiparametric MR imaging.

MATERIALS AND METHODS

Twenty-four patients with relapsing-remitting MS and 24 age- and sex-matched controls were prospectively scanned by quantitative synthetic and 2-shell diffusion MR imaging. Synthetic MR imaging data were used to retrieve relaxometry parameters (R1 and R2 relaxation rates and proton density) and myelin volume fraction. Diffusion tensor metrics (fractional anisotropy and mean, axial, and radial diffusivity) and neurite orientation and dispersion index metrics (intracellular volume fraction, isotropic volume fraction, and orientation dispersion index) were retrieved from diffusion MR imaging data. These data were analyzed using Tract-Based Spatial Statistics.

RESULTS

Patients with MS showed significantly lower fractional anisotropy and myelin volume fraction and higher isotropic volume fraction in widespread white matter areas. Areas with different isotropic volume fractions were included within areas with lower fractional anisotropy. Myelin volume fraction showed no significant difference in some areas with significantly decreased fractional anisotropy in MS, including in the genu of the corpus callosum and bilateral anterior corona radiata, whereas myelin volume fraction was significantly decreased in some areas where fractional anisotropy showed no significant difference, including the bilateral posterior limb of the internal capsule, external capsule, sagittal striatum, fornix, and uncinate fasciculus.

CONCLUSIONS

We found differences in spatial distribution of abnormality in fractional anisotropy, isotropic volume fraction, and myelin volume fraction distribution in MS, which might be useful for characterizing white matter in patients with MS.

MR g-ratio-weighted connectome analysis in patients with multiple sclerosis

Multiple sclerosis (MS) is a brain network disconnection syndrome. Although the brain network topology in MS has been evaluated using diffusion MRI tractography, the mechanism underlying disconnection in the disorder remains unclear. In this study, we evaluated the brain network topology in MS using connectomes with connectivity strengths based on the ratio of the inner to outer myelinated axon diameter (i.e., g-ratio), thereby providing enhanced sensitivity to demyelination compared with the conventional measures of connectivity. We mapped g-ratio-based connectomes in 14 patients with MS and compared them with those of 14 age- and sex-matched healthy controls. For comparison, probabilistic tractography was also used to map connectomes based on the number of streamlines (NOS). We found that g-ratio- and NOS-based connectomes comprised significant connectivity reductions in patients with MS, predominantly in the motor, somatosensory, visual, and limbic regions. However, only the g-ratio-based connectome enabled detection of significant increases in nodal strength in patients with MS. Finally, we found that the g-ratio-weighted nodal strength in motor, visual, and limbic regions significantly correlated with inter-individual variation in measures of disease severity. The g-ratio-based connectome can serve as a sensitive biomarker for diagnosing and monitoring disease progression.

Brain Amyloid PET Tracer Delivery is Related to White Matter Integrity in Patients with Mild Cognitive Impairment

BACKGROUND AND PURPOSE

Amyloid deposition, tau neurofibrillary tangles, and cerebrovascular dysfunction are important pathophysiologic features in Alzheimer's disease. Pittsburgh compound B ([¹¹ C]-PIB) is a positron emission tomography (PET) radiotracer used to quantify amyloid deposition in vivo. In addition, certain models of [¹¹ C]-PIB delivery reflect cerebral blood flow rather than amyloid plaques. As cerebral blood flow and perfusion deficits are associated with white matter pathology, we hypothesized that [¹¹ C]-PIB delivery in white matter regions may reflect white matter integrity.

METHODS

We obtained [¹¹ C]-PIB-PET scans and quantified white matter hyperintensities and global fractional anisotropy on magnetic resonance images as biomarkers of white matter pathology in 34 older participants with mild cognitive impairment with or without a history of major depressive disorder. We analyzed the [¹¹ C]-PIB time-activity curve data with models associated with cerebral blood flow: the early maximum standard uptake value and the relative delivery parameter R1. We used a global white matter region of interest.

RESULTS

Both of the partial-volume corrected PET parameters were correlated with white matter hyperintensities and fractional anisotropy.

CONCLUSION

Future studies are warranted to explore whether [¹¹ C]-PIB PET is a "triple biomarker" that may provide information about amyloid deposition, cerebral blood flow, and white matter pathology.

Segmenting brain tumors from FLAIR MRI using fully convolutional neural networks

BACKGROUND AND OBJECTIVE

Magnetic resonance imaging (MRI) is an indispensable tool in diagnosing brain-tumor patients. Automated tumor segmentation is being widely researched to accelerate the MRI analysis and allow clinicians to precisely plan treatment-accurate delineation of brain tumors is a critical step in assessing their volume, shape, boundaries, and other characteristics. However, it is still a very challenging task due to inherent MR data characteristics and high variability, e.g., in tumor sizes or shapes. We present a new deep learning approach for accurate brain tumor segmentation which can be trained from small and heterogeneous datasets annotated by a human reader (providing high-quality ground-truth segmentation is very costly in practice).

METHODS

In this paper, we present a new deep learning technique for segmenting brain tumors from fluid attenuation inversion recovery MRI. Our technique exploits fully convolutional neural networks, and it is equipped with a battery of augmentation techniques that make the algorithm robust against low data quality, and heterogeneity of small training sets. We train our models using only positive (tumorous) examples, due to the limited amount of available data.

RESULTS

Our algorithm was tested on a set of stage II-IV brain-tumor patients (image data collected using MAGNETOM Prisma 3T, Siemens). Rigorous experiments, backed up with statistical tests, revealed that our approach outperforms the state-of-the-art approach (utilizing hand-crafted features) in terms of segmentation accuracy, offers very fast training and instant segmentation (analysis of an image takes less than a second). Building our deep model is 1.3 times faster compared with extracting features for extremely randomized trees, and this training time can be controlled. Finally, we showed that too aggressive data augmentation may lead to deteriorated performance of the model, especially in the fixed-budget training (with maximum numbers of training epochs).

CONCLUSIONS

Our method yields the better performance when compared with the state of the art method which utilizes hand-crafted features. In addition, our deep network can be effectively applied to difficult (small, imbalanced, and heterogeneous) datasets, offers controllable training time, and infers in real-time.

Brain and lesion segmentation in multiple sclerosis using fully convolutional neural networks: A large-scale study

OBJECTIVE

To investigate the performance of deep learning (DL) based on fully convolutional neural network (FCNN) in segmenting brain tissues in a large cohort of multiple sclerosis (MS) patients.

METHODS

We developed a FCNN model to segment brain tissues, including T2-hyperintense MS lesions. The training, validation, and testing of FCNN were based on ~1000 magnetic resonance imaging (MRI) datasets acquired on relapsing-remitting MS patients, as a part of a phase 3 randomized clinical trial. Multimodal MRI data (dual-echo, FLAIR, and T1-weighted images) served as input to the network. Expert validated segmentation was used as the target for training the FCNN. We cross-validated our results using the leave-one-center-out approach.

RESULTS

We observed a high average (95% confidence limits) Dice similarity coefficient for all the segmented tissues: 0.95 (0.92-0.98) for white matter, 0.96 (0.93-0.98) for gray matter, 0.99 (0.98-0.99) for cerebrospinal fluid, and 0.82 (0.63-1.0) for T2 lesions. High correlations between the DL segmented tissue volumes and ground truth were observed (R² > 0.92 for all tissues). The cross validation showed consistent results across the centers for all tissues.

CONCLUSION

The results from this large-scale study suggest that deep FCNN can automatically segment MS brain tissues, including lesions, with high accuracy.

The Impact of MRI T1 Hypointense Brain Lesions on Cerebral Deep Gray Matter Volume Measures in Multiple Sclerosis

BACKGROUND AND PURPOSE

Deep gray matter (DGM) atrophy has been shown at early stages of multiple sclerosis (MS) and reported as an informative marker of cognitive dysfunction and clinical progression. Therefore, accurate measurement of DGM structure volume is a key priority in MS research. Findings from prior studies have shown that hypointense T1 lesions may impact the accuracy of global brain volume measures; however, literature on the effects of hypointense T1 lesions on DGM structure volumes is sparse.

METHODS

We explored the effects of hypointense T1 lesions on data from 54 relapsing remitting MS patients. Lesions were segmented both manually and with a freely available automatic lesion segmentation/in-painting algorithm (Lesion Segmentation Tool-LST). Volumes of 14 DGM structures were calculated from non-in-painted and in-painted images and compared via paired t-tests, intraclass correlation coefficient, and Dice similarity coefficient.

RESULTS

There were no significant differences in DGM structural volumes between non-in-painted and in-painted images. Automatic lesion-segmentation/in-painting tool provided similar results to manual segmentation/in-painting.

CONCLUSIONS

Our results suggest that lesion in-painting has a negligible impact on DGM structure volume measurement although some regions are more vulnerable to the impact of lesions than others. Furthermore, manual lesion segmentation/in-painting can be replaced by an automatic segmentation/in-painting process.

Metabolic syndrome alters relationships between cardiometabolic variables, cognition and white matter hyperintensity load

Cardiometabolic risk factors influence white matter hyperintensity (WMH) development: in metabolic syndrome (MetS), higher WMH load is often reported but the relationships between specific cardiometabolic variables, WMH load and cognitive performance are uncertain. We investigated these in a Brazilian sample (aged 50–85) with (N = 61) and without (N = 103) MetS. Stepwise regression models identified effects of cardiometabolic and demographic variables on WMH load (from FLAIR MRI) and verbal recall performance. WMH volume was greater in MetS, but verbal recall performance was not impaired. Age showed the strongest relationship with WMH load. Across all participants, systolic blood pressure (SBP) and fasting blood glucose were also contributors, and WMH volume was negatively associated with verbal recall performance. In non-MetS, higher HbA1c, SBP, and number of MetS components were linked to poorer recall performance while higher triglyceride levels appeared to be protective. In MetS only, these relationships were absent but education exerted a strongly protective effect on recall performance. Thus, results support MetS as a construct: the clustering of cardiometabolic variables in MetS alters their individual relationships with cognition; instead, MetS is characterised by a greater reliance on cognitive reserve mechanisms. In non-MetS, strategies to control HbA1c and SBP should be prioritised as these have the largest impact on cognition.

Blood-brain barrier permeability measured using dynamic contrast-enhanced magnetic resonance imaging: a validation study

KEY POINTS

The blood-brain barrier (BBB) is an important and dynamic structure which contributes to homeostasis in the central nervous system. BBB permeability changes occur in health and disease but measurement of BBB permeability in humans is not straightforward. Dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) can be used to model the movement of gadolinium contrast into the brain, expressed as the influx constant Ki . Here evidence is provided that Ki as measured by DCE-MRI behaves as expected for a marker of overall BBB leakage. These results support the use of DCE-MRI for in vivo studies of human BBB permeability in health and disease.

ABSTRACT

Blood-brain barrier (BBB) leakage can be measured using dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) as the influx constant Ki . To validate this method we compared measured Ki with biological expectations, namely (1) higher Ki in healthy individual grey matter (GM) versus white matter (WM), (2) GM/WM cerebral blood volume (CBV) ratio close to the histologically established GM/WM vascular density ratio, (3) higher Ki in visibly enhancing multiple sclerosis (MS) lesions versus MS normal appearing white matter (NAWM), and (4) higher Ki in MS NAWM versus healthy individual NAWM. We recruited 13 healthy individuals and 12 patients with MS and performed whole-brain 3D DCE-MRI at 3 T. Ki and CBV were calculated using Patlak modelling for manual regions of interest (ROI) and segmented tissue masks. Ki was higher in control GM versus WM (P = 0.001). CBV was higher in GM versus WM (P = 0.005, mean ratio 1.9). Ki was higher in visibly enhancing MS lesions versus MS NAWM (P = 0.002), and in MS NAWM versus controls (P = 0.014). Bland-Altman analysis showed no significant difference between ROI and segmentation methods (P = 0.638) and an intra-class correlation coefficient showed moderate single measure consistency (0.610). Ki behaves as expected for a compound marker of permeability and surface area. The GM/WM CBV ratio measured by this technique is in agreement with the literature. This adds evidence to the validity of Ki measured by DCE-MRI as a marker of overall BBB leakage.

Improving the Quality of Synthetic FLAIR Images with Deep Learning Using a Conditional Generative Adversarial Network for Pixel-by-Pixel Image Translation

BACKGROUND AND PURPOSE

Synthetic FLAIR images are of lower quality than conventional FLAIR images. Here, we aimed to improve the synthetic FLAIR image quality using deep learning with pixel-by-pixel translation through conditional generative adversarial network training.

MATERIALS AND METHODS

Forty patients with MS were prospectively included and scanned (3T) to acquire synthetic MR imaging and conventional FLAIR images. Synthetic FLAIR images were created with the SyMRI software. Acquired data were divided into 30 training and 10 test datasets. A conditional generative adversarial network was trained to generate improved FLAIR images from raw synthetic MR imaging data using conventional FLAIR images as targets. The peak signal-to-noise ratio, normalized root mean square error, and the Dice index of MS lesion maps were calculated for synthetic and deep learning FLAIR images against conventional FLAIR images, respectively. Lesion conspicuity and the existence of artifacts were visually assessed.

RESULTS

The peak signal-to-noise ratio and normalized root mean square error were significantly higher and lower, respectively, in generated-versus-synthetic FLAIR images in aggregate intracranial tissues and all tissue segments (all P < .001). The Dice index of lesion maps and visual lesion conspicuity were comparable between generated and synthetic FLAIR images (P = 1 and .59, respectively). Generated FLAIR images showed fewer granular artifacts (P = .003) and swelling artifacts (in all cases) than synthetic FLAIR images.

CONCLUSIONS

Using deep learning, we improved the synthetic FLAIR image quality by generating FLAIR images that have contrast closer to that of conventional FLAIR images and fewer granular and swelling artifacts, while preserving the lesion contrast.

Automatic segmentation of the spinal cord and intramedullary multiple sclerosis lesions with convolutional neural networks

The spinal cord is frequently affected by atrophy and/or lesions in multiple sclerosis (MS) patients. Segmentation of the spinal cord and lesions from MRI data provides measures of damage, which are key criteria for the diagnosis, prognosis, and longitudinal monitoring in MS. Automating this operation eliminates inter-rater variability and increases the efficiency of large-throughput analysis pipelines. Robust and reliable segmentation across multi-site spinal cord data is challenging because of the large variability related to acquisition parameters and image artifacts. In particular, a precise delineation of lesions is hindered by a broad heterogeneity of lesion contrast, size, location, and shape. The goal of this study was to develop a fully-automatic framework - robust to variability in both image parameters and clinical condition - for segmentation of the spinal cord and intramedullary MS lesions from conventional MRI data of MS and non-MS cases. Scans of 1042 subjects (459 healthy controls, 471 MS patients, and 112 with other spinal pathologies) were included in this multi-site study (n = 30). Data spanned three contrasts (T1-, T2-, and T2∗-weighted) for a total of 1943 vol and featured large heterogeneity in terms of resolution, orientation, coverage, and clinical conditions. The proposed cord and lesion automatic segmentation approach is based on a sequence of two Convolutional Neural Networks (CNNs). To deal with the very small proportion of spinal cord and/or lesion voxels compared to the rest of the volume, a first CNN with 2D dilated convolutions detects the spinal cord centerline, followed by a second CNN with 3D convolutions that segments the spinal cord and/or lesions. CNNs were trained independently with the Dice loss. When compared against manual segmentation, our CNN-based approach showed a median Dice of 95% vs. 88% for PropSeg (p ≤ 0.05), a state-of-the-art spinal cord segmentation method. Regarding lesion segmentation on MS data, our framework provided a Dice of 60%, a relative volume difference of -15%, and a lesion-wise detection sensitivity and precision of 83% and 77%, respectively. In this study, we introduce a robust method to segment the spinal cord and intramedullary MS lesions on a variety of MRI contrasts. The proposed framework is open-source and readily available in the Spinal Cord Toolbox.

Value of fluid-attenuated inversion recovery MRI data analyzed by the lesion segmentation toolbox in amyotrophic lateral sclerosis

Background

MRI fluid‐attenuated inversion recovery (FLAIR) studies reported hyperintensity in the corticospinal tract and corpus callosum of patients with amyotrophic lateral sclerosis (ALS).

Purpose

To evaluate the lesion segmentation toolbox (LST) for the objective quantification of FLAIR lesions in ALS patients.

Study Type

Retrospective.

Population

Twenty‐eight ALS patients (eight females, mean age: 50 range: 24–73, mean ALSFRS‐R sum score: 36) were compared with 31 age‐matched healthy controls (12 females, mean age: 45, range: 25–67). ALS patients were treated with riluzole and additional G‐CSF (granulocyte‐colony stimulating factor) on a named patient basis.

Field Strength/Sequence

1.5 T, FLAIR, T₁‐weighted MRI.

Assessment

The lesion prediction algorithm (LPA) of the LST enabled the extraction of individual binary lesion maps, total lesion volume (TLV), and number (TLN). Location and overlap of FLAIR lesions across patients were investigated by registration to FLAIR average space and an atlas. ALS‐specific functional rating scale revised (ALSFRS‐R), disease progression, and survival since diagnosis served as clinical correlates.

Statistical Tests

Univariate analysis of variance (ANOVA), repeated‐measures ANOVA, t‐test, Bravais‐Pearson correlation, Chi‐square test of independence, Kaplan–Meier analysis, Cox‐regression analysis.

Results

Both ALS patients and healthy controls exhibited FLAIR alterations. TLN significantly depended on age (F(1,54) = 24.659, P < 0.001) and sex (F(1,54) = 5.720, P = 0.020). ALS patients showed higher TLN than healthy controls depending on sex (F(1, 54) = 5.076, P = 0.028). FLAIR lesions were small and most pronounced in male ALS patients. FLAIR alterations were predominantly detected in the superior and posterior corona radiata, anterior capsula interna, and posterior thalamic radiation. Patients with pyramidal tract (PT) lesions exhibited significantly inferior survival than patients without PT lesions (P = 0.013). Covariate age exhibited strong prognostic value for survival (P = 0.015).

Data Conclusion

LST enables the objective quantification of FLAIR alterations and is a potential prognostic biomarker for ALS.

Level of Evidence: 3

Technical Efficacy: Stage 2

J. Magn. Reson. Imaging 2019;50:552–559.

Compromised prefrontal structure and function are associated with slower walking in older adults

Our previous work demonstrates that reduced activation of the executive network is associated with slow walking speed in a cohort of older adults from the MOBILIZE Boston Study. However, the influence of underlying white matter integrity on the activation of this network and walking speed is unknown. Thus, we used diffusion-weighted imaging and fMRI during an n-back task to assess associations between executive network structure, function, and walking speed. Whole-brain tract-based spatial statistics (TBSS) were used to identify regions of white matter microstructural integrity that were associated with walking speed. The integrity of these regions was then entered into multiple regression models to predict task performance and executive network activation during the n-back task. Among the significant associations of FA with walking speed, we observed the anterior thalamic radiation and superior longitudinal fasciculus were further associated with both n-back response speed and executive network activation. These findings suggest that subtle damage to frontal white matter may contribute to altered executive network activation and slower walking in older adults.

•

Older adult walking speed was not associated with white matter lesion burden.

•

Walking speed was associated with microstructural white matter integrity.

•

The integrity of prefrontal areas was associated with executive network activation.

•

Low executive network activation also corresponded to slower walking.

•

Interventions targeting the executive network may preserve older adult mobility.

Voxel-Wise Logistic Regression and Leave-One-Source-Out Cross Validation for white matter hyperintensity segmentation

Many algorithms have been proposed for automated segmentation of white matter hyperintensities (WMH) in brain MRI. Yet, broad uptake of any particular algorithm has not been observed. In this work, we argue that this may be due to variable and suboptimal validation data and frameworks, precluding direct comparison of methods on heterogeneous data. As a solution, we present Leave-One-Source-Out Cross Validation (LOSO-CV), which leverages all available data for performance estimation, and show that this gives more realistic (lower) estimates of segmentation algorithm performance on data from different scanners. We also develop a FLAIR-only WMH segmentation algorithm: Voxel-Wise Logistic Regression (VLR), inspired by the open-source Lesion Prediction Algorithm (LPA). Our variant facilitates more accurate parameter estimation, and permits intuitive interpretation of model parameters. We illustrate the performance of the VLR algorithm using the LOSO-CV framework with a dataset comprising freely available data from several recent competitions (96 images from 7 scanners). The performance of the VLR algorithm (median Similarity Index of 0.69) is compared to its LPA predecessor (0.58), and the results of the VLR algorithm in the 2017 WMH Segmentation Competition are also presented.

Different MRI-defined tuber types in tuberous sclerosis complex: Quantitative evaluation and association with disease manifestations

BACKGROUND

Tuberous sclerosis complex (TSC) is a rare genetic disorder with multisystem involvement. A magnetic-resonance (MRI) based classification of tubers into types A, B and C has been proposed. However, the relationship between different tuber types and their quantitative characteristics, also the non-neurological manifestations of TSC remains unknown.

AIMS

To quantitatively evaluate different MRI-defined tuber types and to explore their relationships with major disease manifestations in patients with tuberous sclerosis complex.

METHODS

We performed quantitative manual assessment of tubers visible on T1W, T2W/FLAIR images and DW/ADC maps of 20 patients with TSC. Tubers were classified into types A, B and C based on their signal intensity on MRI. General clinical information and quantitative tuber characteristics were evaluated. Between-group comparisons were made using the nonparametric Mann-Whitney U test with Bonferroni correction.

RESULTS

In total, 20 patients with 770 tubers were evaluated. Type A tubers were most numerous followed closely by Type B tubers, whereas Type C tubers were relatively rare. Tuber size was markedly different among the three tuber types: it increased from Type A to Type B to Type C. Infantile spasms, generalized-tonic clonic seizures, poor seizure control, cardiac rhabdomyomas, SEGA and developmental delay were not associated with quantitative tuber characteristics. Increased total Type B tuber load was associated with early onset epilepsy, while individually larger Type A and Type B tubers were associated with the presence angiomyolipoma (AML) and renal cysts.

CONCLUSIONS

MRI-defined tuber types differ significantly in their size and number. Larger total Type B tuber load and larger individual Type A and Type B tubers were found to be most associated with early seizure onset and renal angiomyolipomas, respectively. One possible explanation for the observed differences in the clinical phenotype based on MRI-defined tuber types is not the intrinsic qualitative distinctions between different tuber types, but rather their individual size and total tuber load.

MRI Analysis of White Matter Myelin Water Content in Multiple Sclerosis: A Novel Approach Applied to Finding Correlates of Cortical Thinning

A novel lesion-mask free method based on a gamma mixture model was applied to myelin water fraction (MWF) maps to estimate the association between cortical thickness and myelin content, and how it differs between relapsing-remitting (RRMS) and secondary-progressive multiple sclerosis (SPMS) groups (135 and 23 patients, respectively). It was compared to an approach based on lesion masks. The gamma mixture distribution of whole brain, white matter (WM) MWF was characterized with three variables: the mode (most frequent value) m₁ of the gamma component shown to relate to lesion, the mode m₂ of the component shown to be associated with normal appearing (NA) WM, and the mixing ratio (λ) between the two distributions. The lesion-mask approach relied on the mean MWF within lesion and within NAWM. A multivariate regression analysis was carried out to find the best predictors of cortical thickness for each group and for each approach. The gamma-mixture method was shown to outperform the lesion-mask approach in terms of adjusted R², both for the RRMS and SPMS groups. The predictors of the final gamma-mixture models were found to be m₁ (β = 1.56, p < 0.005), λ (β = −0.30, p < 0.0005) and age (β = −0.0031, p < 0.005) for the RRMS group (adjusted R² = 0.16), and m₂ (β = 4.72, p < 0.0005) for the SPMS group (adjusted R² = 0.45). Further, a DICE coefficient analysis demonstrated that the lesion mask had more overlap to an ROI associated with m₁, than to an ROI associated with m₂ (p < 0.00001), and vice versa for the NAWM mask (p < 0.00001). These results suggest that during the relapsing phase, focal WM damage is associated with cortical thinning, yet in SPMS patients, global WM deterioration has a much stronger influence on secondary degeneration. Through these findings, we demonstrate the potential contribution of myelin loss on neuronal degeneration at different disease stages and the usefulness of our statistical reduction technique which is not affected by the typical bias associated with approaches based on lesion masks.