Neonatal brain image segmentation in longitudinal MRI studies

In the study of early brain development, tissue segmentation of neonatal brain MR images remains challenging because of the insufficient image quality due to the properties of developing tissues. Among various brain tissue segmentation algorithms, atlas-based brain image segmentation can potentially achieve good segmentation results on neonatal brain images. However, their performances rely on both the quality of the atlas and the spatial correspondence between the atlas and the to-be-segmented image. Moreover, it is difficult to build a population atlas for neonates due to the requirement of a large set of tissue-segmented neonatal brain images. To combat these obstacles, we present a longitudinal neonatal brain image segmentation framework by taking advantage of the longitudinal data acquired at late time-point to build a subject-specific tissue probabilistic atlas. Specifically, tissue segmentation of the neonatal brain is formulated as two iterative steps of bias correction and probabilistic-atlas-based tissue segmentation, along with the longitudinal atlas reconstructed by the late time image of the same subject. The proposed method has been evaluated qualitatively through visual inspection and quantitatively by comparing with manual delineations and two population-atlas-based segmentation methods. Experimental results show that the utilization of a subject-specific probabilistic atlas can substantially improve tissue segmentation of neonatal brain images.     

A viscous fluid model for multimodal non-rigid image registration using mutual information

We propose a multimodal free-form registration algorithm based on maximization of mutual information. The warped image is modeled as a viscous fluid that deforms under the influence of forces derived from the gradient of the mutual information registration criterion. Parzen windowing is used to estimate the joint intensity probability of the images to be matched. The method is evaluated for non-rigid inter-subject registration of MR brain images. The accuracy of the method is verified using simulated multi-modal MR images with known ground truth deformation. The results show that the root mean square difference between the recovered and the ground truth deformation is smaller than 1 voxel. We illustrate the application of the method for atlas-based brain tissue segmentation in MR images in case of gross morphological differences between atlas and patient images.     

Automatic segmentation of MR images of the developing newborn brain

This paper describes an automatic tissue segmentation method for newborn brains from magnetic resonance images (MRI). The analysis and study of newborn brain MRI is of great interest due to its potential for studying early growth patterns and morphological changes in neurodevelopmental disorders. Automatic segmentation of newborn MRI is a challenging task mainly due to the low intensity contrast and the growth process of the white matter tissue. Newborn white matter tissue undergoes a rapid myelination process, where the nerves are covered in myelin sheathes. It is necessary to identify the white matter tissue as myelinated or non-myelinated regions. The degree of myelination is a fractional voxel property that represents regional changes of white matter as a function of age. Our method makes use of a registered probabilistic brain atlas. The method first uses robust graph clustering and parameter estimation to find the initial intensity distributions. The distribution estimates are then used together with the spatial priors to perform bias correction. Finally, the method refines the segmentation using training sample pruning and non-parametric kernel density estimation. Our results demonstrate that the method is able to segment the brain tissue and identify myelinated and non-myelinated white matter regions.     

Segmentation of MRI brain scans using non-uniform partial volume densities

We present an algorithm that provides a partial volume segmentation of a T1-weighted image of the brain into gray matter, white matter and cerebrospinal fluid. The algorithm incorporates a non-uniform partial volume density that takes the curved nature of the cortex into account. The pure gray and white matter intensities are estimated from the image, using scanner noise and cortical partial volume effects. Expected tissue fractions are subsequently computed in each voxel. The algorithm has been tested for reliability, correct estimation of the pure tissue intensities on both real (repeated) MRI data and on simulated (brain) images. Intra-class correlation coefficients (ICCs) were above 0.93 for all volumes of the three tissue types for repeated scans from the same scanner, as well as for scans with different voxel sizes from different scanners with different field strengths. The implementation of our non-uniform partial volume density provided more reliable volumes and tissue fractions, compared to a uniform partial volume density. Applying the algorithm to simulated images showed that the pure tissue intensities were estimated accurately. Variations in cortical thickness did not influence the accuracy of the volume estimates, which is a valuable property when studying (possible) group differences. In conclusion, we have presented a new partial volume segmentation algorithm that allows for comparisons over scanners and voxel sizes.     

Discriminative confidence estimation for probabilistic multi-atlas label fusion

Quantitative neuroimaging analyses often rely on the accurate segmentation of anatomical brain structures. In contrast to manual segmentation, automatic methods offer reproducible outputs and provide scalability to study large databases. Among existing approaches, multi-atlas segmentation has recently shown to yield state-of-the-art performance in automatic segmentation of brain images. It consists in propagating the labelmaps from a set of atlases to the anatomy of a target image using image registration, and then fusing these multiple warped labelmaps into a consensus segmentation on the target image. Accurately estimating the contribution of each atlas labelmap to the final segmentation is a critical step for the success of multi-atlas segmentation. Common approaches to label fusion either rely on local patch similarity, probabilistic statistical frameworks or a combination of both. In this work, we propose a probabilistic label fusion framework based on atlas label confidences computed at each voxel of the structure of interest. Maximum likelihood atlas confidences are estimated using a supervised approach, explicitly modeling the relationship between local image appearances and segmentation errors produced by each of the atlases. We evaluate different spatial pooling strategies for modeling local segmentation errors. We also present a novel type of label-dependent appearance features based on atlas labelmaps that are used during confidence estimation to increase the accuracy of our label fusion. Our approach is evaluated on the segmentation of seven subcortical brain structures from the MICCAI 2013 SATA Challenge dataset and the hippocampi from the ADNI dataset. Overall, our results indicate that the proposed label fusion framework achieves superior performance to state-of-the-art approaches in the majority of the evaluated brain structures and shows more robustness to registration errors.     

Multi-Scale hierarchical generation of PET parametric maps: application and testing on a [11C]DPN study

We propose a general approach to generate parametric maps. It consists in a multi-stage hierarchical scheme where, starting from the kinetic analysis of the whole brain, we then cascade the kinetic information to anatomical systems that are akin in terms of receptor densities, and then down to the voxel level. A-priori classes of voxels are generated either by anatomical atlas segmentation or by functional segmentation using unsupervised clustering. Kinetic properties are transmitted to the voxels in each class using maximum a posteriori (MAP) estimation method. We validate the novel method on a [¹¹C]diprenorphine (DPN) test-retest data-set that represents a challenge to estimation given [¹¹C]DPN's slow equilibration in tissue.The estimated parametric maps of volume of distribution (V_T) reflect the opioid receptor distributions known from previous [¹¹C]DPN studies. When priors are derived from the anatomical atlas, there is an excellent agreement and strong correlation among voxel MAP and ROI results and excellent test-retest reliability for all subjects but one. Voxel level results did not change when priors were defined through unsupervised clustering.This new method is fast (i.e. 15 min per subject) and applied to [¹¹C]DPN data achieves accurate quantification of V_T as well as high quality V_T images. Moreover, the way the priors are defined (i.e. using an anatomical atlas or unsupervised clustering) does not affect the estimates.     

Fast and robust parameter estimation for statistical partial volume models in brain MRI

Due to the finite spatial resolution of imaging devices, a single voxel in a medical image may be composed of mixture of tissue types, an effect known as partial volume effect (PVE). Partial volume estimation, that is, the estimation of the amount of each tissue type within each voxel, has received considerable interest in recent years. Much of this work has been focused on the mixel model, a statistical model of PVE. We propose a novel trimmed minimum covariance determinant (TMCD) method for the estimation of the parameters of the mixel PVE model. In this method, each voxel is first labeled according to the most dominant tissue type. Voxels that are prone to PVE are removed from this labeled set, following which robust location estimators with high breakdown points are used to estimate the mean and the covariance of each tissue class. Comparisons between different methods for parameter estimation based on classified images as well as expectation--maximization-like (EM-like) procedure for simultaneous parameter and partial volume estimation are reported. The robust estimators based on a pruned classification as presented here are shown to perform well even if the initial classification is of poor quality. The results obtained are comparable to those obtained using the EM-like procedure, but require considerably less computation time. Segmentation results of real data based on partial volume estimation are also reported. In addition to considering the parameter estimation problem, we discuss differences between different approximations to the complete mixel model. In summary, the proposed TMCD method allows for the accurate, robust, and efficient estimation of partial volume model parameters, which is crucial to a variety of brain MRI data analysis procedures such as the accurate estimation of tissue volumes and the accurate delineation of the cortical surface.     

Automatic segmentation, internal classification, and follow-up of optic pathway gliomas in MRI

This paper presents an automatic method for the segmentation, internal classification and follow-up of optic pathway gliomas (OPGs) from multi-sequence MRI datasets. Our method starts with the automatic localization of the OPG and its core with an anatomical atlas followed by a binary voxel classification with a probabilistic tissue model whose parameters are estimated from the MR images. The method effectively incorporates prior location, tissue characteristics, and intensity information for the delineation of the OPG boundaries in a consistent and repeatable manner. Internal classification of the segmented OPG volume is then obtained with a robust method that overcomes grey-level differences between learning and testing datasets. Experimental results on 25 datasets yield a mean surface distance error of 0.73 mm as compared to manual segmentation by experienced radiologists. Our method exhibits reliable performance in OPG growth follow-up MR studies, which are crucial for monitoring disease progression. To the best of our knowledge, this is the first method that addresses automatic segmentation, internal classification, and follow-up of OPG.     

LoAd: a locally adaptive cortical segmentation algorithm

Thickness measurements of the cerebral cortex can aid diagnosis and provide valuable information about the temporal evolution of diseases such as Alzheimer's, Huntington's, and schizophrenia. Methods that measure the thickness of the cerebral cortex from in-vivo magnetic resonance (MR) images rely on an accurate segmentation of the MR data. However, segmenting the cortex in a robust and accurate way still poses a challenge due to the presence of noise, intensity non-uniformity, partial volume effects, the limited resolution of MRI and the highly convoluted shape of the cortical folds. Beginning with a well-established probabilistic segmentation model with anatomical tissue priors, we propose three post-processing refinements: a novel modification of the prior information to reduce segmentation bias; introduction of explicit partial volume classes; and a locally varying MRF-based model for enhancement of sulci and gyri. Experiments performed on a new digital phantom, on BrainWeb data and on data from the Alzheimer's Disease Neuroimaging Initiative (ADNI) show statistically significant improvements in Dice scores and PV estimation (p<10(-3)) and also increased thickness estimation accuracy when compared to three well established techniques.     

Estimation of the partial volume effect in MRI

The partial volume effect (PVE) arises in volumetric images when more than one tissue type occurs in a voxel. In such cases, the voxel intensity depends not only on the imaging sequence and tissue properties, but also on the proportions of each tissue type present in the voxel. We have demonstrated in previous work that ignoring this effect by establishing binary voxel-based segmentations introduces significant errors in quantitative measurements, such as estimations of the volumes of brain structures. In this paper, we provide a statistical estimation framework to quantify PVE and to propagate voxel-based estimates in order to compute global magnitudes, such as volume, with associated estimates of uncertainty. Validation is performed on ground truth synthetic images and MRI phantoms, and a clinical study is reported. Results show that the method allows for robust morphometric studies and provides resolution unattainable to date.     

Robust unsupervised segmentation of infarct lesion from diffusion tensor MR images using multiscale statistical classification and partial volume voxel reclassification

Manual region tracing method for segmentation of infarction lesions in images from diffusion tensor magnetic resonance imaging (DT-MRI) is usually used in clinical works, but it is time consuming. A new unsupervised method has been developed, which is a multistage procedure, involving image preprocessing, calculation of tensor field and measurement of diffusion anisotropy, segmentation of infarction volume based on adaptive multiscale statistical classification (MSSC), and partial volume voxel reclassification (PVVR). The method accounts for random noise, intensity overlapping, partial volume effect (PVE), and intensity shading artifacts, which always appear in DT-MR images. The proposed method was applied to 20 patients with clinically diagnosed brain infarction by DT-MRI scans. The accuracy and reproducibility in terms of identifying the infarction lesion have been confirmed by clinical experts. This automatic segmentation method is promising not only in detecting the location and the size of infarction lesion in stroke patient but also in quantitatively analyzing diffusion anisotropy of lesion to guide clinical diagnoses and therapy.     

An information theoretic approach for non-rigid image registration using voxel class probabilities

We propose two information theoretic similarity measures that allow to incorporate tissue class information in non-rigid image registration. The first measure assumes that tissue class probabilities have been assigned to each of the images to be registered by prior segmentation of both of them. One image is then non-rigidly deformed to match the other such that the fuzzy overlap of corresponding voxel object labels becomes similar to the ideal case whereby the tissue probability maps of both images are identical. Image similarity is assessed during registration by the divergence between the ideal and actual joint class probability distributions of both images. A second registration measure is proposed that applies in case a segmentation is available for only one of the images, for instance an atlas image that is to be matched to a study image to guide the segmentation thereof. Intensities in one image are matched to the fuzzy class labels in the other image by minimizing the conditional entropy of the intensities in the first image given the class labels in the second image. We derive analytic expressions for the gradient of each measure with respect to individual voxel displacements to derive a force field that drives the registration process, which is regularized by a viscous fluid model. The performance of the class-based measures is evaluated in the context of non-rigid inter-subject registration and atlas-based segmentation of MR brain images and compared with maximization of mutual information using only intensity information. Our results demonstrate that incorporation of class information in the registration measure significantly improves the overlap between corresponding tissue classes after non-rigid matching. The methods proposed here open new perspectives for integrating segmentation and registration in a single process, whereby the output of one is used to guide the other.     

Unbiased diffeomorphic atlas construction for computational anatomy

Construction of population atlases is a key issue in medical image analysis, and particularly in brain mapping. Large sets of images are mapped into a common coordinate system to study intra-population variability and inter-population differences, to provide voxel-wise mapping of functional sites, and help tissue and object segmentation via registration of anatomical labels. Common techniques often include the choice of a template image, which inherently introduces a bias. This paper describes a new method for unbiased construction of atlases in the large deformation diffeomorphic setting. A child neuroimaging autism study serves as a driving application. There is lack of normative data that explains average brain shape and variability at this early stage of development. We present work in progress toward constructing an unbiased MRI atlas of 2 years of children and the building of a probabilistic atlas of anatomical structures, here the caudate nucleus. Further, we demonstrate the segmentation of new subjects via atlas mapping. Validation of the methodology is performed by comparing the deformed probabilistic atlas with existing manual segmentations.     

Advanced neonatal NeuroMRI

This article describes the potentials and challenges of quantitative analyses of human neonatal brain images using structural magnetic resonance (MR) imaging and diffusion tensor imaging. To maximize the potential of MR imaging for neonatal brain studies, the combination of both contrasts is highly beneficial. Based on the multicontrast data, a neonate brain atlas was created, which allows automated segmentation of neonate brain MR images. The accuracy, advantages, and potential pitfalls of this atlas-based segmentation approach are discussed. The accurate and reproducible MR imaging quantification achieved by this approach could be an initial step toward the successful clinical evaluation of the neonatal brain.     

Construction of a 3D probabilistic atlas of human cortical structures

We describe the construction of a digital brain atlas composed of data from manually delineated MRI data. A total of 56 structures were labeled in MRI of 40 healthy, normal volunteers. This labeling was performed according to a set of protocols developed for this project. Pairs of raters were assigned to each structure and trained on the protocol for that structure. Each rater pair was tested for concordance on 6 of the 40 brains; once they had achieved reliability standards, they divided the task of delineating the remaining 34 brains. The data were then spatially normalized to well-known templates using 3 popular algorithms: AIR5.2.5's nonlinear warp (Woods et al., 1998) paired with the ICBM452 Warp 5 atlas (Rex et al., 2003), FSL's FLIRT (Smith et al., 2004) was paired with its own template, a skull-stripped version of the ICBM152 T1 average; and SPM5's unified segmentation method (Ashburner and Friston, 2005) was paired with its canonical brain, the whole head ICBM152 T1 average. We thus produced 3 variants of our atlas, where each was constructed from 40 representative samples of a data processing stream that one might use for analysis. For each normalization algorithm, the individual structure delineations were then resampled according to the computed transformations. We next computed averages at each voxel location to estimate the probability of that voxel belonging to each of the 56 structures. Each version of the atlas contains, for every voxel, probability densities for each region, thus providing a resource for automated probabilistic labeling of external data types registered into standard spaces; we also computed average intensity images and tissue density maps based on the three methods and target spaces. These atlases will serve as a resource for diverse applications including meta-analysis of functional and structural imaging data and other bioinformatics applications where display of arbitrary labels in probabilistically defined anatomic space will facilitate both knowledge-based development and visualization of findings from multiple disciplines.     

Unsupervised quality control of segmentations based on a smoothness and intensity probabilistic model

Monitoring the quality of image segmentation is key to many clinical applications. This quality assessment can be carried out by a human expert when the number of cases is limited. However, it becomes onerous when dealing with large image databases, so partial automation of this process is preferable. Previous works have proposed both supervised and unsupervised methods for the automated control of image segmentations. The former assume the availability of a subset of trusted segmented images on which supervised learning is performed, while the latter does not. In this paper, we introduce a novel unsupervised approach for quality assessment of segmented images based on a generic probabilistic model. Quality estimates are produced by comparing each segmentation with the output of a probabilistic segmentation model that relies on intensity and smoothness assumptions. Ranking cases with respect to these two assumptions allows the most challenging cases in a dataset to be detected. Furthermore, unlike prior work, our approach enables possible segmentation errors to be localized within an image. The proposed generic probabilistic segmentation method combines intensity mixture distributions with spatial regularization prior models whose parameters are estimated with variational Bayesian techniques. We introduce a novel smoothness prior based on the penalization of the derivatives of label maps which allows an automatic estimation of its hyperparameter in a fully data-driven way. Extensive evaluation of quality control on medical and COCO datasets is conducted, showing the ability to isolate atypical segmentations automatically and to predict, in some cases, the performance of segmentation algorithms.     

Probabilistic segmentation of brain tissue in MR imaging

A new method has been developed for probabilistic segmentation of five different types of brain structures: white matter, gray matter, cerebro-spinal fluid without ventricles, ventricles and white matter lesion in cranial MR imaging. The algorithm is based on information from T1-weighted (T1-w), inversion recovery (IR), proton density-weighted (PD), T2-weighted (T2-w) and fluid attenuation inversion recovery (FLAIR) scans. It uses the K-Nearest Neighbor classification technique that builds a feature space from spatial information and voxel intensities. The technique generates for each tissue type an image representing the probability per voxel being part of it. By application of thresholds on these probability maps, binary segmentations can be obtained. A similarity index (SI) and a probabilistic SI (PSI) were calculated for quantitative evaluation of the results. The influence of each image type on the performance was investigated by alternately leaving out one of the five scan types. This procedure showed that the incorporation of the T1-w, PD or T2-w did not significantly improve the segmentation results. Further investigation indicated that the combination of IR and FLAIR was optimal for segmentation of the five brain tissue types. Evaluation with respect to the gold standard showed that the SI-values for all tissues exceeded 0.8 and all PSI-values exceeded 0.7, implying an excellent agreement.     

Comparative assessment of statistical brain MR image segmentation algorithms and their impact on partial volume correction in PET

Magnetic resonance imaging (MRI)-guided partial volume effect correction (PVC) in brain positron emission tomography (PET) is now a well-established approach to compensate the large bias in the estimate of regional radioactivity concentration, especially for small structures. The accuracy of the algorithms developed so far is, however, largely dependent on the performance of segmentation methods partitioning MRI brain data into its main classes, namely gray matter (GM), white matter (WM), and cerebrospinal fluid (CSF). A comparative evaluation of three brain MRI segmentation algorithms using simulated and clinical brain MR data was performed, and subsequently their impact on PVC in 18F-FDG and 18F-DOPA brain PET imaging was assessed. Two algorithms, the first is bundled in the Statistical Parametric Mapping (SPM2) package while the other is the Expectation Maximization Segmentation (EMS) algorithm, incorporate a priori probability images derived from MR images of a large number of subjects. The third, here referred to as the HBSA algorithm, is a histogram-based segmentation algorithm incorporating an Expectation Maximization approach to model a four-Gaussian mixture for both global and local histograms. Simulated under different combinations of noise and intensity non-uniformity, MR brain phantoms with known true volumes for the different brain classes were generated. The algorithms' performance was checked by calculating the kappa index assessing similarities with the "ground truth" as well as multiclass type I and type II errors including misclassification rates. The impact of image segmentation algorithms on PVC was then quantified using clinical data. The segmented tissues of patients' brain MRI were given as input to the region of interest (RoI)-based geometric transfer matrix (GTM) PVC algorithm, and quantitative comparisons were made. The results of digital MRI phantom studies suggest that the use of HBSA produces the best performance for WM classification. For GM classification, it is suggested to use the EMS. Segmentation performed on clinical MRI data show quite substantial differences, especially when lesions are present. For the particular case of PVC, SPM2 and EMS algorithms show very similar results and may be used interchangeably. The use of HBSA is not recommended for PVC. The partial volume corrected activities in some regions of the brain show quite large relative differences when performing paired analysis on 2 algorithms, implying a careful choice of the segmentation algorithm for GTM-based PVC.     

Sequence-independent segmentation of magnetic resonance images

We present a set of techniques for embedding the physics of the imaging process that generates a class of magnetic resonance images (MRIs) into a segmentation or registration algorithm. This results in substantial invariance to acquisition parameters, as the effect of these parameters on the contrast properties of various brain structures is explicitly modeled in the segmentation. In addition, the integration of image acquisition with tissue classification allows the derivation of sequences that are optimal for segmentation purposes. Another benefit of these procedures is the generation of probabilistic models of the intrinsic tissue parameters that cause MR contrast (e.g., T1, proton density, T2*), allowing access to these physiologically relevant parameters that may change with disease or demographic, resulting in nonmorphometric alterations in MR images that are otherwise difficult to detect. Finally, we also present a high band width multiecho FLASH pulse sequence that results in high signal-to-noise ratio with minimal image distortion due to B0 effects. This sequence has the added benefit of allowing the explicit estimation of T2* and of reducing test-retest intensity variability.     

Morphology-driven automatic segmentation of MR images of the neonatal brain

The segmentation of MR images of the neonatal brain is an essential step in the study and evaluation of infant brain development. State-of-the-art methods for adult brain MRI segmentation are not applicable to the neonatal brain, due to large differences in structure and tissue properties between newborn and adult brains. Existing newborn brain MRI segmentation methods either rely on manual interaction or require the use of atlases or templates, which unavoidably introduces a bias of the results towards the population that was used to derive the atlases. We propose a different approach for the segmentation of neonatal brain MRI, based on the infusion of high-level brain morphology knowledge, regarding relative tissue location, connectivity and structure. Our method does not require manual interaction, or the use of an atlas, and the generality of its priors makes it applicable to different neonatal populations, while avoiding atlas-related bias. The proposed algorithm segments the brain both globally (intracranial cavity, cerebellum, brainstem and the two hemispheres) and at tissue level (cortical and subcortical gray matter, myelinated and unmyelinated white matter, and cerebrospinal fluid). We validate our algorithm through visual inspection by medical experts, as well as by quantitative comparisons that demonstrate good agreement with expert manual segmentations. The algorithm’s robustness is verified by testing on variable quality images acquired on different machines, and on subjects with variable anatomy (enlarged ventricles, preterm- vs. term-born).