A GENERATIVE MODEL FOR MULTI-ATLAS SEGMENTATION ACROSS MODALITIES

Infant FreeSurfer: An automated segmentation and surface extraction pipeline for T1-weighted neuroimaging data of infants 0-2 years

The development of automated tools for brain morphometric analysis in infants has lagged significantly behind analogous tools for adults. This gap reflects the greater challenges in this domain due to: 1) a smaller-scaled region of interest, 2) increased motion corruption, 3) regional changes in geometry due to heterochronous growth, and 4) regional variations in contrast properties corresponding to ongoing myelination and other maturation processes. Nevertheless, there is a great need for automated image-processing tools to quantify differences between infant groups and other individuals, because aberrant cortical morphologic measurements (including volume, thickness, surface area, and curvature) have been associated with neuropsychiatric, neurologic, and developmental disorders in children. In this paper we present an automated segmentation and surface extraction pipeline designed to accommodate clinical MRI studies of infant brains in a population 0-2 year-olds. The algorithm relies on a single channel of T1-weighted MR images to achieve automated segmentation of cortical and subcortical brain areas, producing volumes of subcortical structures and surface models of the cerebral cortex. We evaluated the algorithm both qualitatively and quantitatively using manually labeled datasets, relevant comparator software solutions cited in the literature, and expert evaluations. The computational tools and atlases described in this paper will be distributed to the research community as part of the FreeSurfer image analysis package.

System-Specific Patterns of Thalamocortical Connectivity in Early Brain Development as Revealed by Structural and Functional MRI

The normal development of thalamocortical connections plays a critical role in shaping brain connectivity in the prenatal and postnatal periods. Recent studies using advanced magnetic resonance imaging (MRI) techniques in neonates and infants have shown that abnormal thalamocortical connectivity is associated with adverse neurodevelopmental outcomes. However, all these studies have focused on a single neuroimaging modality, overlooking the dynamic relationship between structure and function at this early stage. Here, we study the relationship between structural and functional thalamocortical connectivity patterns derived from healthy full-term infants scanned with diffusion-weighted MRI and resting-state functional MRI within the first weeks of life (mean gestational age = 39.3 ± 1.2 weeks; age at scan = 24.2 ± 7.9 days). Our results show that while there is, in general, good spatial agreement between both MRI modalities, there are regional variations that are system-specific: regions involving primary-sensory cortices exhibit greater structural/functional overlap, whereas higher-order association areas such as temporal and posterior parietal cortices show divergence in spatial patterns of each modality. This variability illustrates the complementarity of both modalities and highlights the importance of multimodal approaches.

Relating anthropometric indicators to brain structure in 2-month-old Bangladeshi infants growing up in poverty: A pilot study

Anthropometric indicators, including stunting, underweight, and wasting, have previously been associated with poor neurocognitive outcomes. This link may exist because malnutrition and infection, which are known to affect height and weight, also impact brain structure according to animal models. However, a relationship between anthropometric indicators and brain structural measures has not been tested yet, perhaps because stunting, underweight, and wasting are uncommon in higher-resource settings. Further, with diminished anthropometric growth prevalent in low-resource settings, where biological and psychosocial hazards are most severe, one might expect additional links between measures of poverty, anthropometry, and brain structure. To begin to examine these relationships, we conducted an MRI study in 2-3-month-old infants growing up in the extremely impoverished urban setting of Dhaka, Bangladesh. The sample size was relatively small because the challenges of investigating infant brain structure in a low-resource setting needed to be realized and resolved before introducing a larger cohort. Initially, fifty-four infants underwent T1 sequences using 3T MRI, and resulting structural images were segmented into gray and white matter maps, which were carefully evaluated for accurate tissue labeling by a pediatric neuroradiologist. Gray and white matter volumes from 29 infants (79 ​± ​10 days-of-age; F/M ​= ​12/17), whose segmentations were of relatively high quality, were submitted to semi-partial correlation analyses with stunting, underweight, and wasting, which were measured using height-for-age (HAZ), weight-for-age (WAZ), and weight-for-height (WHZ) scores. Positive semi-partial correlations (after adjusting for chronological age and sex and correcting for multiple comparisons) were observed between white matter volume and HAZ and WAZ; however, WHZ was not correlated with any measure of brain volume. No associations were observed between income-to-needs or maternal education and brain volumetric measures, suggesting that measures of poverty were not associated with total brain tissue volume in this sample. Overall, these results provide the first link between diminished anthropometric growth and white matter volume in infancy. Challenges of conducting a developmental neuroimaging study in a low-resource country are also described.

TRActs constrained by UnderLying INfant anatomy (TRACULInA): An automated probabilistic tractography tool with anatomical priors for use in the newborn brain

The ongoing myelination of white-matter fiber bundles plays a significant role in brain development. However, reliable and consistent identification of these bundles from infant brain MRIs is often challenging due to inherently low diffusion anisotropy, as well as motion and other artifacts. In this paper we introduce a new tool for automated probabilistic tractography specifically designed for newborn infants. Our tool incorporates prior information about the anatomical neighborhood of white-matter pathways from a training data set. In our experiments, we evaluate this tool on data from both full-term and prematurely born infants and demonstrate that it can reconstruct known white-matter tracts in both groups robustly, even in the presence of differences between the training set and study subjects. Additionally, we evaluate it on a publicly available large data set of healthy term infants (UNC Early Brain Development Program). This paves the way for performing a host of sophisticated analyses in newborns that we have previously implemented for the adult brain, such as pointwise analysis along tracts and longitudinal analysis, in both health and disease.

The relationship between biological and psychosocial risk factors and resting-state functional connectivity in 2-month-old Bangladeshi infants: A feasibility and pilot study

Childhood poverty has been associated with structural and functional alterations in the developing brain. However, poverty does not alter brain development directly, but acts through associated biological or psychosocial risk factors (e.g. malnutrition, family conflict). Yet few studies have investigated risk factors in the context of infant neurodevelopment, and none have done so in low-resource settings such as Bangladesh, where children are exposed to multiple, severe biological and psychosocial hazards. In this feasibility and pilot study, usable resting-state fMRI data were acquired in infants from extremely poor (n = 16) and (relatively) more affluent (n = 16) families in Dhaka, Bangladesh. Whole-brain intrinsic functional connectivity (iFC) was estimated using bilateral seeds in the amygdala, where iFC has shown susceptibility to early life stress, and in sensory areas, which have exhibited less susceptibility to early life hazards. Biological and psychosocial risk factors were examined for associations with iFC. Three resting-state networks were identified in within-group brain maps: medial temporal/striatal, visual, and auditory networks. Infants from extremely poor families compared with those from more affluent families exhibited greater (i.e. less negative) iFC in precuneus for amygdala seeds; however, no group differences in iFC were observed for sensory area seeds. Height-for-age, a proxy for malnutrition/infection, was not associated with amygdala/precuneus iFC, whereas prenatal family conflict was positively correlated. Findings suggest that it is feasible to conduct infant fMRI studies in low-resource settings. Challenges and practical steps for successful implementations are discussed.

Smaller hippocampal CA1 subfield volume in posttraumatic stress disorder

BACKGROUND

Smaller hippocampal volume in patients with posttraumatic stress disorder (PTSD) represents the most consistently reported structural alteration in the brain. Subfields of the hippocampus play distinct roles in encoding and processing of memories, which are disrupted in PTSD. We examined PTSD-associated alterations in 12 hippocampal subfields in relation to global hippocampal shape, and clinical features.

METHODS

Case-control cross-sectional studies of U.S. military veterans (n = 282) from the Iraq and Afghanistan era were grouped into PTSD (n = 142) and trauma-exposed controls (n = 140). Participants underwent clinical evaluation for PTSD and associated clinical parameters followed by MRI at 3 T. Segmentation with FreeSurfer v6.0 produced hippocampal subfield volumes for the left and right CA1, CA3, CA4, DG, fimbria, fissure, hippocampus-amygdala transition area, molecular layer, parasubiculum, presubiculum, subiculum, and tail, as well as hippocampal meshes. Covariates included age, gender, trauma exposure, alcohol use, depressive symptoms, antidepressant medication use, total hippocampal volume, and MRI scanner model.

RESULTS

Significantly lower subfield volumes were associated with PTSD in left CA1 (P = 0.01; d = 0.21; uncorrected), CA3 (P = 0.04; d = 0.08; uncorrected), and right CA3 (P = 0.02; d = 0.07; uncorrected) only if ipsilateral whole hippocampal volume was included as a covariate. A trend level association of L-CA1 with PTSD (F4, 221  = 3.32, P = 0.07) is present and the other subfield findings are nonsignificant if ipsilateral whole hippocampal volume is not included as a covariate. PTSD-associated differences in global hippocampal shape were nonsignificant.

CONCLUSIONS

The present finding of smaller hippocampal CA1 in PTSD is consistent with model systems in rodents that exhibit increased anxiety-like behavior from repeated exposure to acute stress. Behavioral correlations with hippocampal subfield volume differences in PTSD will elucidate their relevance to PTSD, particularly behaviors of associative fear learning, extinction training, and formation of false memories.

Generative diffeomorphic modelling of large MRI data sets for probabilistic template construction

In this paper we present a hierarchical generative model of medical image data, which can capture simultaneously the variability of both signal intensity and anatomical shapes across large populations. Such a model has a direct application for learning average-shaped probabilistic tissue templates in a fully automated manner. While in principle the generality of the proposed Bayesian approach makes it suitable to address a wide range of medical image computing problems, our work focuses primarily on neuroimaging applications. In particular we validate the proposed method on both real and synthetic brain MR scans including the cervical cord and demonstrate that it yields accurate alignment of brain and spinal cord structures, as compared to state-of-the-art tools for medical image registration. At the same time we illustrate how the resulting tissue probability maps can readily be used to segment, bias correct and spatially normalise unseen data, which are all crucial pre-processing steps for MR imaging studies.

•

We present a generative modelling framework to process large MRI data sets.

•

The proposed framework can serve to learn average-shaped tissue probability maps and empirical intensity priors.

•

We explore semi-supervised learning and variational inference schemes.

•

The method is validated against state-of-the-art tools using publicly available data.

Probabilistic tractography-based thalamic parcellation in healthy newborns and newborns with congenital heart disease

BACKGROUND

Given the central role of the thalamus in motor, sensory, and cognitive development, methods to study emerging thalamocortical connectivity in early infancy are of great interest.

PURPOSE

To determine the feasibility of performing probabilistic tractography-based thalamic parcellation (PTbTP) in typically developing (TD) neonates and to compare the results with a pilot sample of neonates with congenital heart disease (CHD).

STUDY TYPE

Institutional Review Board (IRB)-approved cross-sectional study.

MODEL

We prospectively recruited 20 TD neonates and five CHD neonates (imaged preoperatively).

FIELD STRENGTH/SEQUENCE

MRI was performed at 3.0T including diffusion-weighted imaging (DWI) and 3D magnetization prepared rapid gradient-echo (MPRAGE).

ASSESSMENT

A radiologist and trained research assistants segmented the thalamus and seven cortical targets for each hemisphere. Using the thalami as seeds and the cortical labels as targets, FSL library tools were used to generate probabilistic tracts. A Hierarchical Dirichlet Process algorithm was then used for clustering analysis. A radiologist qualitatively assessed the results of clustering. Quantitative analyses were also performed.

STATISTICAL TESTS

We summarized the demographic data and results of clustering with descriptive statistics. Linear regressions covarying for gestational age were used to compare groups.

RESULTS

In 17 of 20 TD neonates, we identified five connectivity-determined clusters, which correlate with known thalamic nuclei and subnuclei. In four neonates with CHD we observed a spectrum of abnormalities including fewer and disorganized clusters or small supernumerary clusters (up to seven per thalamus). After covarying for differences in corrected gestational age (cGA), the fractional anisotropy (FA), volume, and normalized thalamic volume were significantly lower in CHD neonates (P < 0.01).

DATA CONCLUSIONS

Using PTbTP clusters, correlating well with the location and connectivity of known thalamic nuclei, were identified in TD neonates. Differences in thalamic clustering outputs were identified in four neonates with CHD, raising concern for disordered thalamic connectivity. PTbTP is feasible in TD and CHD neonates. Preliminary findings suggest the prenatal origins of altered connectivity in CHD.

LEVEL OF EVIDENCE

2 Technical Efficacy: Stage 4 J. Magn. Reson. Imaging 2018;47:1626-1637.

Fast and sequence-adaptive whole-brain segmentation using parametric Bayesian modeling

Quantitative analysis of magnetic resonance imaging (MRI) scans of the brain requires accurate automated segmentation of anatomical structures. A desirable feature for such segmentation methods is to be robust against changes in acquisition platform and imaging protocol. In this paper we validate the performance of a segmentation algorithm designed to meet these requirements, building upon generative parametric models previously used in tissue classification. The method is tested on four different datasets acquired with different scanners, field strengths and pulse sequences, demonstrating comparable accuracy to state-of-the-art methods on T1-weighted scans while being one to two orders of magnitude faster. The proposed algorithm is also shown to be robust against small training datasets, and readily handles images with different MRI contrast as well as multi-contrast data.

Improved segmentation of cerebellar structures in children

BACKGROUND

Consistent localization of cerebellar cortex in a standard coordinate system is important for functional studies and detection of anatomical alterations in studies of morphometry. To date, no pediatric cerebellar atlas is available.

NEW METHOD

The probabilistic Cape Town Pediatric Cerebellar Atlas (CAPCA18) was constructed in the age-appropriate National Institute of Health Pediatric Database asymmetric template space using manual tracings of 16 cerebellar compartments in 18 healthy children (9-13 years) from Cape Town, South Africa. The individual atlases of the training subjects were also used to implement multi atlas label fusion using multi atlas majority voting (MAMV) and multi atlas generative model (MAGM) approaches. Segmentation accuracy in 14 test subjects was compared for each method to 'gold standard' manual tracings.

RESULTS

Spatial overlap between manual tracings and CAPCA18 automated segmentation was 73% or higher for all lobules in both hemispheres, except VIIb and X. Automated segmentation using MAGM yielded the best segmentation accuracy over all lobules (mean Dice Similarity Coefficient 0.76; range 0.55-0.91; mean Hausdorff distance 0.9 mm; range 0.8-2.7 mm).

COMPARISON WITH EXISTING METHODS

In all lobules, spatial overlap of CAPCA18 segmentations with manual tracings was similar or higher than those obtained with SUIT (spatially unbiased infra-tentorial template), providing additional evidence of the benefits of an age appropriate atlas. MAGM segmentation accuracy was comparable to values reported recently by Park et al. (Neuroimage 2014;95(1):217) in adults (across all lobules mean DSC=0.73, range 0.40-0.89).

CONCLUSIONS

CAPCA18 and the associated multi-subject atlases of the training subjects yield improved segmentation of cerebellar structures in children.

Multi-atlas segmentation of biomedical images: A survey

Multi-atlas segmentation (MAS), first introduced and popularized by the pioneering work of Rohlfing, et al. (2004), Klein, et al. (2005), and Heckemann, et al. (2006), is becoming one of the most widely-used and successful image segmentation techniques in biomedical applications. By manipulating and utilizing the entire dataset of "atlases" (training images that have been previously labeled, e.g., manually by an expert), rather than some model-based average representation, MAS has the flexibility to better capture anatomical variation, thus offering superior segmentation accuracy. This benefit, however, typically comes at a high computational cost. Recent advancements in computer hardware and image processing software have been instrumental in addressing this challenge and facilitated the wide adoption of MAS. Today, MAS has come a long way and the approach includes a wide array of sophisticated algorithms that employ ideas from machine learning, probabilistic modeling, optimization, and computer vision, among other fields. This paper presents a survey of published MAS algorithms and studies that have applied these methods to various biomedical problems. In writing this survey, we have three distinct aims. Our primary goal is to document how MAS was originally conceived, later evolved, and now relates to alternative methods. Second, this paper is intended to be a detailed reference of past research activity in MAS, which now spans over a decade (2003-2014) and entails novel methodological developments and application-specific solutions. Finally, our goal is to also present a perspective on the future of MAS, which, we believe, will be one of the dominant approaches in biomedical image segmentation.

An algorithm for optimal fusion of atlases with different labeling protocols

In this paper we present a novel label fusion algorithm suited for scenarios in which different manual delineation protocols with potentially disparate structures have been used to annotate the training scans (hereafter referred to as "atlases"). Such scenarios arise when atlases have missing structures, when they have been labeled with different levels of detail, or when they have been taken from different heterogeneous databases. The proposed algorithm can be used to automatically label a novel scan with any of the protocols from the training data. Further, it enables us to generate new labels that are not present in any delineation protocol by defining intersections on the underling labels. We first use probabilistic models of label fusion to generalize three popular label fusion techniques to the multi-protocol setting: majority voting, semi-locally weighted voting and STAPLE. Then, we identify some shortcomings of the generalized methods, namely the inability to produce meaningful posterior probabilities for the different labels (majority voting, semi-locally weighted voting) and to exploit the similarities between the atlases (all three methods). Finally, we propose a novel generative label fusion model that can overcome these drawbacks. We use the proposed method to combine four brain MRI datasets labeled with different protocols (with a total of 102 unique labeled structures) to produce segmentations of 148 brain regions. Using cross-validation, we show that the proposed algorithm outperforms the generalizations of majority voting, semi-locally weighted voting and STAPLE (mean Dice score 83%, vs. 77%, 80% and 79%, respectively). We also evaluated the proposed algorithm in an aging study, successfully reproducing some well-known results in cortical and subcortical structures.

A unified framework for cross-modality multi-atlas segmentation of brain MRI

Multi-atlas label fusion is a powerful image segmentation strategy that is becoming increasingly popular in medical imaging. A standard label fusion algorithm relies on independently computed pairwise registrations between individual atlases and the (target) image to be segmented. These registrations are then used to propagate the atlas labels to the target space and fuse them into a single final segmentation. Such label fusion schemes commonly rely on the similarity between intensity values of the atlases and target scan, which is often problematic in medical imaging - in particular, when the atlases and target images are obtained via different sensor types or imaging protocols. In this paper, we present a generative probabilistic model that yields an algorithm for solving the atlas-to-target registrations and label fusion steps simultaneously. The proposed model does not directly rely on the similarity of image intensities. Instead, it exploits the consistency of voxel intensities within the target scan to drive the registration and label fusion, hence the atlases and target image can be of different modalities. Furthermore, the framework models the joint warp of all the atlases, introducing interdependence between the registrations. We use variational expectation maximization and the Demons registration framework in order to efficiently identify the most probable segmentation and registrations. We use two sets of experiments to illustrate the approach, where proton density (PD) MRI atlases are used to segment T1-weighted brain scans and vice versa. Our results clearly demonstrate the accuracy gain due to exploiting within-target intensity consistency and integrating registration into label fusion.

A probabilistic patch-based label fusion model for multi-atlas segmentation with registration refinement: application to cardiac MR images

The evaluation of ventricular function is important for the diagnosis of cardiovascular diseases. It typically involves measurement of the left ventricular (LV) mass and LV cavity volume. Manual delineation of the myocardial contours is time-consuming and dependent on the subjective experience of the expert observer. In this paper, a multi-atlas method is proposed for cardiac magnetic resonance (MR) image segmentation. The proposed method is novel in two aspects. First, it formulates a patch-based label fusion model in a Bayesian framework. Second, it improves image registration accuracy by utilizing label information, which leads to improvement of segmentation accuracy. The proposed method was evaluated on a cardiac MR image set of 28 subjects. The average Dice overlap metric of our segmentation is 0.92 for the LV cavity, 0.89 for the right ventricular cavity and 0.82 for the myocardium. The results show that the proposed method is able to provide accurate information for clinical diagnosis.

A Generative Model for Probabilistic Label Fusion of Multimodal Data

The maturity of registration methods, in combination with the increasing processing power of computers, has made multi-atlas segmentation methods practical. The problem of merging the deformed label maps from the atlases is known as label fusion. Even though label fusion has been well studied for intramodality scenarios, it remains relatively unexplored when the nature of the target data is multimodal or when its modality is different from that of the atlases. In this paper, we review the literature on label fusion methods and also present an extension of our previously published algorithm to the general case in which the target data are multimodal. The method is based on a generative model that exploits the consistency of voxel intensities within the target scan based on the current estimate of the segmentation. Using brain MRI scans acquired with a multiecho FLASH sequence, we compare the method with majority voting, statistical-atlas-based segmentation, the popular package FreeSurfer and an adaptive local multi-atlas segmentation method. The results show that our approach produces highly accurate segmentations (Dice 86.3% across 22 brain structures of interest), outperforming the competing methods.