Deconfounded Dimension Reduction via Partial Embeddings

Dimension reduction tools preserving similarity and graph structure such as t-SNE and UMAP can capture complex biological patterns in high-dimensional data. However, these tools typically are not designed to separate effects of interest from unwanted effects due to confounders. We introduce the partial embedding (PARE) framework, which enables removal of confounders from any distance-based dimension reduction method. We then develop partial t-SNE and partial UMAP and apply these methods to genomic and neuroimaging data. Our results show that the PARE framework can remove batch effects in single-cell sequencing data as well as separate clinical and technical variability in neuroimaging measures. We demonstrate that the PARE framework extends dimension reduction methods to highlight biological patterns of interest while effectively removing confounding effects.

Generalizing deep learning brain segmentation for skull removal and intracranial measurements

Total intracranial volume (TICV) and posterior fossa volume (PFV) are essential covariates for brain volumetric analyses with structural magnetic resonance imaging (MRI). Detailed whole brain segmentation provides a non-invasive way to measure brain regions. Furthermore, increasing neuroimaging data are distributed in a skull-stripped manner for privacy protection. Therefore, generalizing deep learning brain segmentation for skull removal and intracranial measurements is an appealing task. However, data availability is challenging due to a limited set of manually traced atlases with whole brain and TICV/PFV labels. In this paper, we employ U-Net tiles to achieve automatic TICV estimation and whole brain segmentation simultaneously on brains w/and w/o the skull. To overcome the scarcity of manually traced whole brain volumes, a transfer learning method is introduced to estimate additional TICV and PFV labels during whole brain segmentation in T1-weighted MRI. Specifically, U-Net tiles are first pre-trained using large-scale BrainCOLOR atlases without TICV and PFV labels, which are created by multi-atlas segmentation. Then the pre-trained models are refined by training the additional TICV and PFV labels using limited BrainCOLOR atlases. We also extend our method to handle skull-stripped brain MR images. From the results, our method provides promising whole brain segmentation and volume estimation results for both brains w/and w/o skull in terms of mean Dice similarity coefficients and mean surface distance and absolute volume similarity. This method has been made available in open source (https://github.com/MASILab/SLANTbrainSeg_skullstripped).

Five-year longitudinal changes in quantitative spinal cord MRI in multiple sclerosis

BACKGROUND

The spinal cord (SC) is highly relevant to disability in multiple sclerosis (MS), but few studies have evaluated longitudinal changes in quantitative spinal cord magnetic resonance imaging (SC-MRI).

OBJECTIVES

The aim of this study was to characterize the relationships between 5-year changes in SC-MRI with disability in MS.

METHODS

In total, 75 MS patients underwent 3 T SC-MRI and clinical assessment (expanded disability status scale (EDSS) and MS functional composite (MSFC)) at baseline, 2 and 5 years. SC-cross-sectional area (CSA) and diffusion-tensor indices (fractional anisotropy (FA), mean, perpendicular, parallel diffusivity (MD, λ⊥, λ||) and magnetization transfer ratio (MTR)) were extracted at C3-C4. Mixed-effects regression incorporating subject-specific slopes assessed longitudinal change in SC-MRI measures.

RESULTS

SC-CSA and MTR decreased (p = 0.009, p = 0.03) over 5.1 years. There were moderate correlations between 2- and 5-year subject-specific slopes of SC-MRI indices and follow-up EDSS scores (Pearson's r with FA = -0.23 (p < 0.001); MD = 0.31 (p < 0.001); λ⊥ = 0.34 (p < 0.001); λ|| = -0.12 (p = 0.05), MTR = -0.37 (p < 0.001); SC-CSA = -0.47 (p < 0.001) at 5 years); MSFC showed similar trends. The 2- and 5-year subject-specific slopes were robustly correlated (r = 0.93-0.97 for FA, λ⊥, SC-CSA and MTR, all ps < 0.001).

CONCLUSION

In MS, certain quantitative SC-MRI indices change over 5 years, reflecting ongoing tissue changes. Subject-specific trajectories of SC-MRI index change at 2 and 5 years are strongly correlated and highly relevant to follow-up disability. These findings suggest that individual dynamics of change should be accounted for when interpreting longitudinal SC-MRI measures and that measuring short-term change is predictive of long-term clinical disability.

Cognitive impairment, the central vein sign, and paramagnetic rim lesions in RIS

Objective:

The central vein sign (CVS) and “paramagnetic rim lesions” (PRL) are emerging imaging biomarkers in multiple sclerosis (MS) reflecting perivenular demyelination and chronic, smoldering inflammation. The objective of this study was to assess relationships between cognitive impairment (CI) and the CVS and PRL in radiologically isolated syndrome (RIS).

Methods:

Twenty-seven adults with RIS underwent 3.0 T MRI of the brain and cervical spinal cord (SC) and cognitive assessment using the minimal assessment of cognitive function in MS battery. The CVS and PRL were assessed in white-matter lesions (WMLs) on T2*-weighted segmented echo-planar magnitude and phase images. Multivariable linear regression evaluated relationships between CI and MRI measures.

Results:

Global CI was present in 9 (33%) participants with processing speed and visual memory most frequently affected. Most participants (93%) had ⩾ 40% CVS + WML (a threshold distinguishing MS from other WM disorders); 63% demonstrated PRL. Linear regression revealed that CVS + WML predicted performance on verbal memory(β =-0.024, p = 0.03) while PRL predicted performance on verbal memory (β = -0.040, p = 0.04) and processing speed (β = -0.039, p = 0.03).

Conclusions:

CI is common in RIS and is associated with markers of perivenular demyelination and chronic inflammation in WML, such as CVS + WML and PRL. A prospective follow-up of this cohort will ascertain the importance of CI, CVS, and PRL as risk factors for conversion from RIS to MS.

Matrix decomposition for modeling lesion development processes in multiple sclerosis

Our main goal is to study and quantify the evolution of multiple sclerosis lesions observed longitudinally over many years in multi-sequence structural magnetic resonance imaging (sMRI). To achieve that, we propose a class of functional models for capturing the temporal dynamics and spatial distribution of the voxel-specific intensity trajectories in all sMRI sequences. To accommodate the hierarchical data structure (observations nested within voxels, which are nested within lesions, which, in turn, are nested within study participants), we use structured functional principal component analysis. We propose and evaluate the finite sample properties of hypothesis tests of therapeutic intervention effects on lesion evolution while accounting for the multilevel structure of the data. Using this novel testing strategy, we found statistically significant differences in lesion evolution between treatment groups.

Abstract 37: Increased Mean Transit Time and Blood Brain Permeability in Asymptomatic White Matter Lesions of Ischemic Origin

Introduction: White matter lesions of presumed ischemic origin (WMH) have been associated with increased risk of stroke, cognitive and motor decline, and are a subject of public health research. Engineering new MRI pipelines allowing for determination of mean transit time (MTT), and blood brain barrier permeability (BBBP), within WMH lesions is required for long-term population-based studies of lesion progression in patients with dementia and vascular cognitive impairment.

Methods: WMH lesion volumes in 24 asymptomatic individuals was determined using an automated segmentation methodology, S3DL, with manual correction to remove false positives. A double contrast injection scheme was used to measure both K trans using dynamic contrast enhanced (DCE) imaging and K 2 using dynamic susceptibility contrast (DSC) imaging which also provided perfusion-related measures. BBBP was measured as k 2 within segmented WMH lesions and compared with normal white and gray matter.

Results: The mean transit time (MTT) was found to be significantly prolonged (8.11, p<0.001Wilcoxon Signed-Rank Test) in WMH lesions when compared to normal appearing white and gray matter. There was no significant difference in DCE-K trans (0.018, p=0.351) between the lesions and the white/gray matter. Permeability measured in the WMH lesions using the DSC-K 2 method was increased and was correlated withincreasing total WMH lesion volume (spearman correlation 0.44; p< 0.046).

Conclusion: In this first study using an advanced WMH lesion automated segmentation pipeline, we measured DCE and DSC perfusion and permeability variables within WMH lesions and compared them to normal white and grey matter in healthy people. We observed increasing MTT, within WMH lesions as compared to unaffected white and gray matter. Using the DSC-K 2 method, BBBP was higher within WMH lesions in these asymptomatic people, and correlated with increasing total lesion volume.

Manganese-Enhanced MRI of the Brain in Healthy Volunteers

BACKGROUND AND PURPOSE

The manganese ion is used as an intracellular MR imaging contrast agent to study neuronal function in animal models, but it remains unclear whether manganese-enhanced MR imaging can be similarly useful in humans. Using mangafodipir (Teslascan, a chelated manganese-based contrast agent that is FDA-approved), we evaluated the dynamics of manganese enhancement of the brain and glandular structures in the rostral head and neck in healthy volunteers.

MATERIALS AND METHODS

We administered mangafodipir intravenously at a rate of 1 mL/minute for a total dose of 5 μmol/kg body weight. Nine healthy adult volunteers (6 men/3 women; median age, 43 years) completed baseline history and physical examination, 3T MR imaging, and blood work. MR imaging also followed mangafodipir administration at various time points from immediate to 7 days, with delayed scans at 1-3 months.

RESULTS

The choroid plexus and anterior pituitary gland enhanced within 10 minutes of infusion, with enhancement persisting up to 7 and 30 days, respectively. Exocrine (parotid, submandibular, sublingual, and lacrimal) glands also enhanced avidly as early as 1 hour postadministration, generally resolving by 1 month; 3 volunteers had residual exocrine gland enhancement, which resolved by 2 months in 1 and by 3 months in the other 2. Mangafodipir did not affect clinical parameters, laboratory values, or T1-weighted signal in the basal ganglia.

CONCLUSIONS

Manganese ions released from mangafodipir successfully enable noninvasive visualization of intra- and extracranial structures that lie outside the blood-brain barrier without adverse clinical effects, setting the stage for future neuroradiologic investigation in disease.

Imaging outcome measures of neuroprotection and repair in MS: A consensus statement from NAIMS

OBJECTIVE

To summarize current and emerging imaging techniques that can be used to assess neuroprotection and repair in multiple sclerosis (MS), and to provide a consensus opinion on the potential utility of each technique in clinical trial settings.

METHODS

Clinicians and scientists with expertise in the use of MRI in MS convened in Toronto, Canada, in November 2016 at a North American Imaging in Multiple Sclerosis (NAIMS) Cooperative workshop meeting. The discussion was compiled into a manuscript and circulated to all NAIMS members in attendance. Edits and feedback were incorporated until all authors were in agreement.

RESULTS

A wide spectrum of imaging techniques and analysis methods in the context of specific study designs were discussed, with a focus on the utility and limitations of applying each technique to assess neuroprotection and repair. Techniques were discussed under specific themes, and included conventional imaging, magnetization transfer ratio, diffusion tensor imaging, susceptibility-weighted imaging, imaging cortical lesions, magnetic resonance spectroscopy, PET, advanced diffusion imaging, sodium imaging, multimodal techniques, imaging of special regions, statistical considerations, and study design.

CONCLUSIONS

Imaging biomarkers of neuroprotection and repair are an unmet need in MS. There are a number of promising techniques with different strengths and limitations, and selection of a specific technique will depend on a number of factors, notably the question the trial seeks to answer. Ongoing collaborative efforts will enable further refinement and improved methods to image the effect of novel therapeutic agents that exert benefit in MS predominately through neuroprotective and reparative mechanisms.

Improved Visualization of Cortical Lesions in Multiple Sclerosis Using 7T MP2RAGE

BACKGROUND AND PURPOSE

Cortical lesions are common and often extensive in multiple sclerosis but are difficult to visualize by MRI, leaving important questions about their clinical implications and response to therapy unanswered. Our aim was to determine whether cortical lesions are better visualized using magnetization prepared 2 rapid acquisition gradient echoes (MP2RAGE) than T2*-weighted imaging on 7T MR imaging.

MATERIALS AND METHODS

Brain MR imaging using T1-weighted MP2RAGE at 500-μm isotropic resolution, T2*-weighted gradient-echo, and T2*-weighted segmented echo-planar imaging sequences were collected for 13 patients with MS and 5 age-matched neurologically healthy controls on a 7T research system. One MS case underwent postmortem MR imaging including gradient-echo and MP2RAGE sequences, after which cortical lesions seen on MR imaging were assessed with immunohistochemistry.

RESULTS

MP2RAGE detected 203 cortical lesions (median, 16 lesions/case; interquartile range, 15), compared to 92 with T2*gradient-echo (median, 7; interquartile range, 8; P < .001) and 81 with T2*EPI (median, 7; interquartile range, 5; P < .001). This increase in lesion number detected on MP2RAGE versus T2* was observed for juxtacortical, leukocortical, and intracortical lesions. Forty-three percent of all cortical lesions were identified only on MP2RAGE. White matter lesion volume correlated with total juxtacortical (r = 0.86, P < .001) and leukocortical lesion volume (r = 0.70, P < .01) but not intracortical lesion volume, suggesting that pathophysiology may differ by lesion type. Of 4 suspected lesions seen on postmortem imaging, 3 were found to be true cortical lesions while 1 represented postmortem tissue damage.

CONCLUSIONS

A combination of MP2RAGE and T2*-weighted imaging at 7T improved detection of cortical lesions and should enable longitudinal studies to elucidate their spatiotemporal dynamics and clinical implications.

Statistical estimation of white matter microstructure from conventional MRI

Diffusion tensor imaging (DTI) has become the predominant modality for studying white matter integrity in multiple sclerosis (MS) and other neurological disorders. Unfortunately, the use of DTI-based biomarkers in large multi-center studies is hindered by systematic biases that confound the study of disease-related changes. Furthermore, the site-to-site variability in multi-center studies is significantly higher for DTI than that for conventional MRI-based markers. In our study, we apply the Quantitative MR Estimation Employing Normalization (QuEEN) model to estimate the four DTI measures: MD, FA, RD, and AD. QuEEN uses a voxel-wise generalized additive regression model to relate the normalized intensities of one or more conventional MRI modalities to a quantitative modality, such as DTI. We assess the accuracy of the models by comparing the prediction error of estimated DTI images to the scan-rescan error in subjects with two sets of scans. Across the four DTI measures, the performance of the models is not consistent: Both MD and RD estimations appear to be quite accurate, while AD estimation is less accurate than MD and RD; the accuracy of FA estimation is poor. Thus, in some cases when assessing white matter integrity, it may be sufficient to acquire conventional MRI sequences alone.

Statistical estimation of T1 relaxation times using conventional magnetic resonance imaging

Quantitative T1 maps estimate T1 relaxation times and can be used to assess diffuse tissue abnormalities within normal-appearing tissue. T1 maps are popular for studying the progression and treatment of multiple sclerosis (MS). However, their inclusion in standard imaging protocols remains limited due to the additional scanning time and expert calibration required and susceptibility to bias and noise. Here, we propose a new method of estimating T1 maps using four conventional MR images, which are intensity-normalized using cerebellar gray matter as a reference tissue and related to T1 using a smooth regression model. Using cross-validation, we generate statistical T1 maps for 61 subjects with MS. The statistical maps are less noisy than the acquired maps and show similar reproducibility. Tests of group differences in normal-appearing white matter across MS subtypes give similar results using both methods.

Litter decomposition in western red cedar and western hemlock forests on northern Vancouver Island, British Columbia

Litter decomposition and changes in N and organic chemicals were studied for 2 years in two forest types: old-growth western red cedar (Thuja plicata Donn) and western hemlock (Tsuga heterophylla (Raf.) Sarge) and 85-year-old stands of western hemlock and amabilis fir (Abies amabilis (Dougl.) Forbes) that developed after a major windstorm. We tested the hypothesis that lower rates of mass loss and different patterns of nutrient release in decomposing litter could explain lower nutrient availability in the cedar–hemlock type. Decomposition rate of a standard litter substrate, lodgepole pine needles, was almost identical in the two forest types indicating that each type had similar microenvironmental conditions for decomposers. Salal leaves had a lower lignin to N ratio and decomposed and released N more rapidly than the conifer litters. Among the conifers, cedar had poorer litter quality (higher lignin to N ratio), decomposed more slowly, and released considerably less N during the study. Cedar litter contributes to lower N availability in cedar–hemlock forests, but other factors, such as lower external N cycling and complexing of N with secondary carbon compounds during later stages of decomposition, are also likely to have a major influence on N availability. Keywords: Thuja plicata, Tsuga heterophylla, decomposition, litter quality, N cycling.