Multi-atlas segmentation of biomedical images: A survey

Genetic mapping of molar size relations identifies inhibitory locus for third molars in mice

Molar size in Mammals shows considerable disparity and exhibits variation similar to that predicted by the Inhibitory Cascade model. The importance of such developmental systems in favoring evolutionary trajectories is also underlined by the fact that this model can predict macroevolutionary patterns. Using backcross mice, we mapped QTL for molar sizes controlling for their sequential development. Genetic controls for upper and lower molars appear somewhat similar, and regions containing genes implied in dental defects drive this variation. We mapped three relationship QTLs (rQTL) modifying the control of the mesial molars on the focal third molar. These regions overlap Shh, Sostdc1, and Fst genes, which have pervasive roles in development and should be buffered against new variation. It has theoretically been shown that rQTL produces new variation channeled in the direction of adaptive changes. Our results provide evidence that evolutionary/disease patterns of tooth size variation could result from such a non-random generating process.

Deep Deconvolutional Neural Network for Target Segmentation of Nasopharyngeal Cancer in Planning Computed Tomography Images

BACKGROUND

Radiotherapy is one of the main treatment methods for nasopharyngeal carcinoma (NPC). It requires exact delineation of the nasopharynx gross tumor volume (GTVnx), the metastatic lymph node gross tumor volume (GTVnd), the clinical target volume (CTV), and organs at risk in the planning computed tomography images. However, this task is time-consuming and operator dependent. In the present study, we developed an end-to-end deep deconvolutional neural network (DDNN) for segmentation of these targets.

METHODS

The proposed DDNN is an end-to-end architecture enabling fast training and testing. It consists of two important components: an encoder network and a decoder network. The encoder network was used to extract the visual features of a medical image and the decoder network was used to recover the original resolution by deploying deconvolution. A total of 230 patients diagnosed with NPC stage I or stage II were included in this study. Data from 184 patients were chosen randomly as a training set to adjust the parameters of DDNN, and the remaining 46 patients were the test set to assess the performance of the model. The Dice similarity coefficient (DSC) was used to quantify the segmentation results of the GTVnx, GTVnd, and CTV. In addition, the performance of DDNN was compared with the VGG-16 model.

RESULTS

The proposed DDNN method outperformed the VGG-16 in all the segmentation. The mean DSC values of DDNN were 80.9% for GTVnx, 62.3% for the GTVnd, and 82.6% for CTV, whereas VGG-16 obtained 72.3, 33.7, and 73.7% for the DSC values, respectively.

CONCLUSION

DDNN can be used to segment the GTVnx and CTV accurately. The accuracy for the GTVnd segmentation was relatively low due to the considerable differences in its shape, volume, and location among patients. The accuracy is expected to increase with more training data and combination of MR images. In conclusion, DDNN has the potential to improve the consistency of contouring and streamline radiotherapy workflows, but careful human review and a considerable amount of editing will be required.

Recent advances in radiation oncology

Radiotherapy (RT) is very much a technology-driven treatment modality in the management of cancer. RT techniques have changed significantly over the past few decades, thanks to improvements in engineering and computing. We aim to highlight the recent developments in radiation oncology, focusing on the technological and biological advances. We will present state-of-the-art treatment techniques, employing photon beams, such as intensity-modulated RT, volumetric-modulated arc therapy, stereotactic body RT and adaptive RT, which make possible a highly tailored dose distribution with maximum normal tissue sparing. We will analyse all the steps involved in the treatment: imaging, delineation of the tumour and organs at risk, treatment planning and finally image-guidance for accurate tumour localisation before and during treatment delivery. Particular attention will be given to the crucial role that imaging plays throughout the entire process. In the case of adaptive RT, the precise identification of target volumes as well as the monitoring of tumour response/modification during the course of treatment is mainly based on multimodality imaging that integrates morphological, functional and metabolic information. Moreover, real-time imaging of the tumour is essential in breathing adaptive techniques to compensate for tumour motion due to respiration. Brief reference will be made to the recent spread of particle beam therapy, in particular to the use of protons, but also to the yet limited experience of using heavy particles such as carbon ions. Finally, we will analyse the latest biological advances in tumour targeting. Indeed, the effectiveness of RT has been improved not only by technological developments but also through the integration of radiobiological knowledge to produce more efficient and personalised treatment strategies.

Computational and mathematical methods in brain atlasing

Brain atlases have a wide range of use from education to research to clinical applications. Mathematical methods as well as computational methods and tools play a major role in the process of brain atlas building and developing atlas-based applications. Computational methods and tools cover three areas: dedicated editors for brain model creation, brain navigators supporting multiple platforms, and atlas-assisted specific applications. Mathematical methods in atlas building and developing atlas-aided applications deal with problems in image segmentation, geometric body modelling, physical modelling, atlas-to-scan registration, visualisation, interaction and virtual reality. Here I overview computational and mathematical methods in atlas building and developing atlas-assisted applications, and share my contribution to and experience in this field.

Comparison of image intensity, local, and multi-atlas priors in brain tissue classification

PURPOSE

Automated and accurate tissue classification in three-dimensional brain magnetic resonance images is essential in volumetric morphometry or as a preprocessing step for diagnosing brain diseases. However, noise, intensity in homogeneity, and partial volume effects limit the classification accuracy of existing methods. This paper provides a comparative study on the contributions of three commonly used image information priors for tissue classification in normal brains: image intensity, local, and multi-atlas priors.

METHODS

We compared the effectiveness of the three priors by comparing the four methods modeling them: K-Means (KM), KM combined with a Markov Random Field (KM-MRF), multi-atlas segmentation (MAS), and the combination of KM, MRF, and MAS (KM-MRF-MAS). The key parameters and factors in each of the four methods are analyzed, and the performance of all the models is compared quantitatively and qualitatively on both simulated and real data.

RESULTS

The KM-MRF-MAS model that combines the three image information priors performs best.

CONCLUSIONS

The image intensity prior is insufficient to generate reasonable results for a few images. Introducing local and multi-atlas priors results in improved brain tissue classification. This study provides a general guide on what image information priors can be used for effective brain tissue classification.

Detecting Anatomical Landmarks From Limited Medical Imaging Data Using Two-Stage Task-Oriented Deep Neural Networks

One of the major challenges in anatomical landmark detection, based on deep neural networks, is the limited availability of medical imaging data for network learning. To address this problem, we present a two-stage task-oriented deep learning method to detect large-scale anatomical landmarks simultaneously in real time, using limited training data. Specifically, our method consists of two deep convolutional neural networks (CNN), with each focusing on one specific task. Specifically, to alleviate the problem of limited training data, in the first stage, we propose a CNN based regression model using millions of image patches as input, aiming to learn inherent associations between local image patches and target anatomical landmarks. To further model the correlations among image patches, in the second stage, we develop another CNN model, which includes a) a fully convolutional network that shares the same architecture and network weights as the CNN used in the first stage and also b) several extra layers to jointly predict coordinates of multiple anatomical landmarks. Importantly, our method can jointly detect large-scale (e.g., thousands of) landmarks in real time. We have conducted various experiments for detecting 1200 brain landmarks from the 3D T1-weighted magnetic resonance images of 700 subjects, and also 7 prostate landmarks from the 3D computed tomography images of 73 subjects. The experimental results show the effectiveness of our method regarding both accuracy and efficiency in the anatomical landmark detection.

Primatologist: A modular segmentation pipeline for macaque brain morphometry

Because they bridge the genetic gap between rodents and humans, non-human primates (NHPs) play a major role in therapy development and evaluation for neurological disorders. However, translational research success from NHPs to patients requires an accurate phenotyping of the models. In patients, magnetic resonance imaging (MRI) combined with automated segmentation methods has offered the unique opportunity to assess in vivo brain morphological changes. Meanwhile, specific challenges caused by brain size and high field contrasts make existing algorithms hard to use routinely in NHPs. To tackle this issue, we propose a complete pipeline, Primatologist, for multi-region segmentation. Tissue segmentation is based on a modular statistical model that includes random field regularization, bias correction and denoising and is optimized by expectation-maximization. To deal with the broad variety of structures with different relaxing times at 7 T, images are segmented into 17 anatomical classes, including subcortical regions. Pre-processing steps insure a good initialization of the parameters and thus the robustness of the pipeline. It is validated on 10 T2-weighted MRIs of healthy macaque brains. Classification scores are compared with those of a non-linear atlas registration, and the impact of each module on classification scores is thoroughly evaluated.

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.

Segmentation and Quantification for Angle-Closure Glaucoma Assessment in Anterior Segment OCT

Angle-closure glaucoma is a major cause of irreversible visual impairment and can be identified by measuring the anterior chamber angle (ACA) of the eye. The ACA can be viewed clearly through anterior segment optical coherence tomography (AS-OCT), but the imaging characteristics and the shapes and locations of major ocular structures can vary significantly among different AS-OCT modalities, thus complicating image analysis. To address this problem, we propose a data-driven approach for automatic AS-OCT structure segmentation, measurement, and screening. Our technique first estimates initial markers in the eye through label transfer from a hand-labeled exemplar data set, whose images are collected over different patients and AS-OCT modalities. These initial markers are then refined by using a graph-based smoothing method that is guided by AS-OCT structural information. These markers facilitate segmentation of major clinical structures, which are used to recover standard clinical parameters. These parameters can be used not only to support clinicians in making anatomical assessments, but also to serve as features for detecting anterior angle closure in automatic glaucoma screening algorithms. Experiments on Visante AS-OCT and Cirrus high-definition-OCT data sets demonstrate the effectiveness of our approach.

Multiple Sclerosis Lesion Segmentation Using Joint Label Fusion

This paper adapts the joint label fusion (JLF) multi-atlas image segmentation algorithm to the problem of multiple sclerosis (MS) lesion segmentation in multi-modal MRI. Conventionally, JLF requires a set of atlas images to be co-registered to the target image using deformable registration. However, given the variable spatial distribution of lesions in the brain, whole-brain deformable registration is unlikely to line up lesions between atlases and the target image. As a solution, we propose to first pre-segment the target image using an intensity regression based technique, yielding a set of "candidate" lesions. Each "candidate" lesion is then matched to a set of similar lesions in the atlas based on location and size; and deformable registration and JLF are applied at the level of the "candidate" lesion. The approach is evaluated on a dataset of 74 subjects with MS and shown to improve Dice similarity coefficient with reference manual segmentation by 12% over intensity regression technique.

Brain Image Labeling Using Multi-atlas Guided 3D Fully Convolutional Networks

Automatic labeling of anatomical structures in brain images plays an important role in neuroimaging analysis. Among all methods, multi-atlas based segmentation methods are widely used, due to their robustness in propagating prior label information. However, non-linear registration is always needed, which is time-consuming. Alternatively, the patch-based methods have been proposed to relax the requirement of image registration, but the labeling is often determined independently by the target image information, without getting direct assistance from the atlases. To address these limitations, in this paper, we propose a multi-atlas guided 3D fully convolutional networks (FCN) for brain image labeling. Specifically, multi-atlas based guidance is incorporated during the network learning. Based on this, the discriminative of the FCN is boosted, which eventually contribute to accurate prediction. Experiments show that the use of multi-atlas guidance improves the brain labeling performance.

4D Multi-atlas Label Fusion using Longitudinal Images

Longitudinal reproducibility is an essential concern in automated medical image segmentation, yet has proven to be an elusive objective as manual brain structure tracings have shown more than 10% variability. To improve reproducibility, longitudinal segmentation (4D) approaches have been investigated to reconcile temporal variations with traditional 3D approaches. In the past decade, multi-atlas label fusion has become a state-of-the-art segmentation technique for 3D image and many efforts have been made to adapt it to a 4D longitudinal fashion. However, the previous methods were either limited by using application specified energy function (e.g., surface fusion and multi model fusion) or only considered temporal smoothness on two consecutive time points (t and t+1) under sparsity assumption. Therefore, a 4D multi-atlas label fusion theory for general label fusion purpose and simultaneously considering temporal consistency on all time points is appealing. Herein, we propose a novel longitudinal label fusion algorithm, called 4D joint label fusion (4DJLF), to incorporate the temporal consistency modeling via non-local patch-intensity covariance models. The advantages of 4DJLF include: (1) 4DJLF is under the general label fusion framework by simultaneously incorporating the spatial and temporal covariance on all longitudinal time points. (2) The proposed algorithm is a longitudinal generalization of a leading joint label fusion method (JLF) that has proven adaptable to a wide variety of applications. (3) The spatial temporal consistency of atlases is modeled in a probabilistic model inspired from both voting based and statistical fusion. The proposed approach improves the consistency of the longitudinal segmentation while retaining sensitivity compared with original JLF approach using the same set of atlases. The method is available online in open-source.

Multiprotocol, multiatlas statistical fusion: theory and application

Multiatlas segmentation offers an exceedingly convenient process by which image segmentation tools can be created from a series of labeled atlases (i.e., raters). However, creation of the atlases is exceedingly time consuming and prone to shifts in clinical/research demands as anatomical definitions are refined, combined, or subdivided. Hence, a process by which atlases from distinct, but complementary, anatomical "protocols" could be combined would allow for greater innovation in structural analysis and efficiency of data (re)use. Recent innovation in protocol fusion has shown that propagation of information across distinct protocols is feasible. However, how to effectively include this information in simultaneous truth and performance level estimation (STAPLE) has been elusive. We present a generalization of the STAPLE framework to account for multiprotocol rater performance (i.e., accuracy of registered atlases). This approach, multiset STAPLE (MS-STAPLE), provides a statistical framework for combining label information from atlases that have been labeled with distinct protocols (i.e., whole brain versus subcortical) and is compatible with the current local, nonlocal, probabilistic, log-odds, and hierarchical innovations in STAPLE theory. Using the MS-STAPLE approach, information from a broad range of datasets can be combined so that each available dataset contributes in a spatially dependent manner to local labels. We evaluate the model in simulations and in the context of an experiment where an existing set of whole-brain labels (14 structures) is refined to include parcellation of subcortical structures (26 structures). In the empirical results, we see significant improvement in the Dice similarity coefficient when comparing MS-STAPLE to STAPLE and nonlocal MS-STAPLE to nonlocal STAPLE.

Reverse Classification Accuracy: Predicting Segmentation Performance in the Absence of Ground Truth

When integrating computational tools, such as automatic segmentation, into clinical practice, it is of utmost importance to be able to assess the level of accuracy on new data and, in particular, to detect when an automatic method fails. However, this is difficult to achieve due to the absence of ground truth. Segmentation accuracy on clinical data might be different from what is found through cross validation, because validation data are often used during incremental method development, which can lead to overfitting and unrealistic performance expectations. Before deployment, performance is quantified using different metrics, for which the predicted segmentation is compared with a reference segmentation, often obtained manually by an expert. But little is known about the real performance after deployment when a reference is unavailable. In this paper, we introduce the concept of reverse classification accuracy (RCA) as a framework for predicting the performance of a segmentation method on new data. In RCA, we take the predicted segmentation from a new image to train a reverse classifier, which is evaluated on a set of reference images with available ground truth. The hypothesis is that if the predicted segmentation is of good quality, then the reverse classifier will perform well on at least some of the reference images. We validate our approach on multi-organ segmentation with different classifiers and segmentation methods. Our results indicate that it is indeed possible to predict the quality of individual segmentations, in the absence of ground truth. Thus, RCA is ideal for integration into automatic processing pipelines in clinical routine and as a part of large-scale image analysis studies.

A competitive strategy for atrial and aortic tract segmentation based on deformable models

Multiple strategies have previously been described for atrial region (i.e. atrial bodies and aortic tract) segmentation. Although these techniques have proven their accuracy, inadequate results in the mid atrial walls are common, restricting their application for specific cardiac interventions. In this work, we introduce a novel competitive strategy to perform atrial region segmentation with correct delineation of the thin mid walls, and integrated it into the B-spline Explicit Active Surfaces framework. A double-stage segmentation process is used, which starts with a fast contour growing followed by a refinement stage with local descriptors. Independent functions are used to define each region, being afterward combined to compete for the optimal boundary. The competition locally constrains the surface evolution, prevents overlaps and allows refinement to the walls. Three different scenarios were used to demonstrate the advantages of the proposed approach, through the evaluation of its segmentation accuracy, and its performance for heterogeneous mid walls. Both computed tomography and magnetic resonance imaging datasets were used, presenting results similar to the state-of-the-art methods for both atria and aorta. The competitive strategy showed its superior performance with statistically significant differences against the traditional free-evolution approach in cases with bad image quality or missed atrial/aortic walls. Moreover, only the competitive approach was able to accurately segment the atrial/aortic wall. Overall, the proposed strategy showed to be suitable for atrial region segmentation with a correct segmentation of the mid thin walls, demonstrating its added value with respect to the traditional techniques.

Automated segmentation of human cervical-supraclavicular adipose tissue in magnetic resonance images

Human brown adipose tissue (BAT), with a major site in the cervical-supraclavicular depot, is a promising anti-obesity target. This work presents an automated method for segmenting cervical-supraclavicular adipose tissue for enabling time-efficient and objective measurements in large cohort research studies of BAT. Fat fraction (FF) and R₂^* maps were reconstructed from water-fat magnetic resonance imaging (MRI) of 25 subjects. A multi-atlas approach, based on atlases from nine subjects, was chosen as automated segmentation strategy. A semi-automated reference method was used to validate the automated method in the remaining subjects. Automated segmentations were obtained from a pipeline of preprocessing, affine registration, elastic registration and postprocessing. The automated method was validated with respect to segmentation overlap (Dice similarity coefficient, Dice) and estimations of FF, R₂^* and segmented volume. Bias in measurement results was also evaluated. Segmentation overlaps of Dice = 0.93 ± 0.03 (mean ± standard deviation) and correlation coefficients of r > 0.99 (P < 0.0001) in FF, R₂^* and volume estimates, between the methods, were observed. Dice and BMI were positively correlated (r = 0.54, P = 0.03) but no other significant bias was obtained (P ≥ 0.07). The automated method compared well with the reference method and can therefore be suitable for time-efficient and objective measurements in large cohort research studies of BAT.

Using clinically acquired MRI to construct age-specific ADC atlases: Quantifying spatiotemporal ADC changes from birth to 6-year old

Diffusion imaging is critical for detecting acute brain injury. However, normal apparent diffusion coefficient (ADC) maps change rapidly in early childhood, making abnormality detection difficult. In this article, we explored clinical PACS and electronic healthcare records (EHR) to create age-specific ADC atlases for clinical radiology reference. Using the EHR and three rounds of multiexpert reviews, we found ADC maps from 201 children 0-6 years of age scanned between 2006 and 2013 who had brain MRIs with no reported abnormalities and normal clinical evaluations 2+ years later. These images were grouped in 10 age bins, densely sampling the first 1 year of life (5 bins, including neonates and 4 quarters) and representing the 1-6 year age range (an age bin per year). Unbiased group-wise registration was used to construct ADC atlases for 10 age bins. We used the atlases to quantify (a) cross-sectional normative ADC variations; (b) spatiotemporal heterogeneous ADC changes; and (c) spatiotemporal heterogeneous volumetric changes. The quantified age-specific whole-brain and region-wise ADC values were compared to those from age-matched individual subjects in our study and in multiple existing independent studies. The significance of this study is that we have shown that clinically acquired images can be used to construct normative age-specific atlases. These first of their kind age-specific normative ADC atlases quantitatively characterize changes of myelination-related water diffusion in the first 6 years of life. The quantified voxel-wise spatiotemporal ADC variations provide standard references to assist radiologists toward more objective interpretation of abnormalities in clinical images. Our atlases are available at https://www.nitrc.org/projects/mgh_adcatlases. Hum Brain Mapp 38:3052-3068, 2017. © 2017 Wiley Periodicals, Inc.

Classification and evaluation strategies of auto-segmentation approaches for PET: Report of AAPM task group No. 211

PURPOSE

The purpose of this educational report is to provide an overview of the present state-of-the-art PET auto-segmentation (PET-AS) algorithms and their respective validation, with an emphasis on providing the user with help in understanding the challenges and pitfalls associated with selecting and implementing a PET-AS algorithm for a particular application.

APPROACH

A brief description of the different types of PET-AS algorithms is provided using a classification based on method complexity and type. The advantages and the limitations of the current PET-AS algorithms are highlighted based on current publications and existing comparison studies. A review of the available image datasets and contour evaluation metrics in terms of their applicability for establishing a standardized evaluation of PET-AS algorithms is provided. The performance requirements for the algorithms and their dependence on the application, the radiotracer used and the evaluation criteria are described and discussed. Finally, a procedure for algorithm acceptance and implementation, as well as the complementary role of manual and auto-segmentation are addressed.

FINDINGS

A large number of PET-AS algorithms have been developed within the last 20 years. Many of the proposed algorithms are based on either fixed or adaptively selected thresholds. More recently, numerous papers have proposed the use of more advanced image analysis paradigms to perform semi-automated delineation of the PET images. However, the level of algorithm validation is variable and for most published algorithms is either insufficient or inconsistent which prevents recommending a single algorithm. This is compounded by the fact that realistic image configurations with low signal-to-noise ratios (SNR) and heterogeneous tracer distributions have rarely been used. Large variations in the evaluation methods used in the literature point to the need for a standardized evaluation protocol.

CONCLUSIONS

Available comparison studies suggest that PET-AS algorithms relying on advanced image analysis paradigms provide generally more accurate segmentation than approaches based on PET activity thresholds, particularly for realistic configurations. However, this may not be the case for simple shape lesions in situations with a narrower range of parameters, where simpler methods may also perform well. Recent algorithms which employ some type of consensus or automatic selection between several PET-AS methods have potential to overcome the limitations of the individual methods when appropriately trained. In either case, accuracy evaluation is required for each different PET scanner and scanning and image reconstruction protocol. For the simpler, less robust approaches, adaptation to scanning conditions, tumor type, and tumor location by optimization of parameters is necessary. The results from the method evaluation stage can be used to estimate the contouring uncertainty. All PET-AS contours should be critically verified by a physician. A standard test, i.e., a benchmark dedicated to evaluating both existing and future PET-AS algorithms needs to be designed, to aid clinicians in evaluating and selecting PET-AS algorithms and to establish performance limits for their acceptance for clinical use. The initial steps toward designing and building such a standard are undertaken by the task group members.

Atlas-Based Mosaicing of Left Atrial 3-D Transesophageal Echocardiography Images

Transesophageal echocardiography (TEE) is a promising imaging modality used to guide cardiac interventions, such as catheter ablation for the treatment of cardiac arrhythmias. These procedures rely on good visualization of the left atrium and pulmonary veins. To visualize these structures in a single volume, the acquisition, registration and fusion of multiple TEE views of the left atrium are required. We introduce atlas-based mosaicing as a method for the registration of images that are acquired according to a standardized protocol. Inspired by atlas-based segmentation approaches, compounded data of other patients serve as atlases for the registration of new data. The performance of atlas-based mosaicing is studied on 3-D TEE data of the left atrium and compared with that of regular pairwise registration. This study indicates that improved registration robustness and smaller registration errors are achieved with atlas-based mosaicing compared with regular pairwise registration. This is an important step toward the use of TEE for interventional guidance of ablation procedures.

BrainSegNet: a convolutional neural network architecture for automated segmentation of human brain structures

Automated segmentation of cortical and noncortical human brain structures has been hitherto approached using nonrigid registration followed by label fusion. We propose an alternative approach for this using a convolutional neural network (CNN) which classifies a voxel into one of many structures. Four different kinds of two-dimensional and three-dimensional intensity patches are extracted for each voxel, providing local and global (context) information to the CNN. The proposed approach is evaluated on five different publicly available datasets which differ in the number of labels per volume. The obtained mean Dice coefficient varied according to the number of labels, for example, it is [Formula: see text] and [Formula: see text] for datasets with the least (32) and the most (134) number of labels, respectively. These figures are marginally better or on par with those obtained with the current state-of-the-art methods on nearly all datasets, at a reduced computational time. The consistently good performance of the proposed method across datasets and no requirement for registration make it attractive for many applications where reduced computational time is necessary.

Transformation diffusion reconstruction of three-dimensional histology volumes from two-dimensional image stacks

•

A method for 3D reconstruction of serial 2D histology image stacks is proposed.

•

Pre-alignment to an external pre-cut reference (blockface) prevents shape artifacts.

•

Formulated as diffusion of transformations from each slice to its neighbors.

•

Registrations replaced by much faster transformation operations.

Traditional histology is the gold standard for tissue studies, but it is intrinsically reliant on two-dimensional (2D) images. Study of volumetric tissue samples such as whole hearts produces a stack of misaligned and distorted 2D images that need to be reconstructed to recover a congruent volume with the original sample's shape. In this paper, we develop a mathematical framework called Transformation Diffusion (TD) for stack alignment refinement as a solution to the heat diffusion equation. This general framework does not require contour segmentation, is independent of the registration method used, and is trivially parallelizable. After the first stack sweep, we also replace registration operations by operations in the space of transformations, several orders of magnitude faster and less memory-consuming. Implementing TD with operations in the space of transformations produces our Transformation Diffusion Reconstruction (TDR) algorithm, applicable to general transformations that are closed under inversion and composition. In particular, we provide formulas for translation and affine transformations. We also propose an Approximated TDR (ATDR) algorithm that extends the same principles to tensor-product B-spline transformations. Using TDR and ATDR, we reconstruct a full mouse heart at pixel size 0.92 µm × 0.92 µm, cut 10 µm thick, spaced 20 µm (84G). Our algorithms employ only local information from transformations between neighboring slices, but the TD framework allows theoretical analysis of the refinement as applying a global Gaussian low-pass filter to the unknown stack misalignments. We also show that reconstruction without an external reference produces large shape artifacts in a cardiac specimen while still optimizing slice-to-slice alignment. To overcome this problem, we use a pre-cutting blockface imaging process previously developed by our group that takes advantage of Brewster's angle and a polarizer to capture the outline of only the topmost layer of wax in the block containing embedded tissue for histological sectioning.

Automatic segmentation and quantification of the cardiac structures from non-contrast-enhanced cardiac CT scans

Early structural changes to the heart, including the chambers and the coronary arteries, provide important information on pre-clinical heart disease like cardiac failure. Currently, contrast-enhanced cardiac computed tomography angiography (CCTA) is the preferred modality for the visualization of the cardiac chambers and the coronaries. In clinical practice not every patient undergoes a CCTA scan; many patients receive only a non-contrast-enhanced calcium scoring CT scan (CTCS), which has less radiation dose and does not require the administration of contrast agent. Quantifying cardiac structures in such images is challenging, as they lack the contrast present in CCTA scans. Such quantification would however be relevant, as it enables population based studies with only a CTCS scan. The purpose of this work is therefore to investigate the feasibility of automatic segmentation and quantification of cardiac structures viz whole heart, left atrium, left ventricle, right atrium, right ventricle and aortic root from CTCS scans. A fully automatic multi-atlas-based segmentation approach is used to segment the cardiac structures. Results show that the segmentation overlap between the automatic method and that of the reference standard have a Dice similarity coefficient of 0.91 on average for the cardiac chambers. The mean surface-to-surface distance error over all the cardiac structures is [Formula: see text] mm. The automatically obtained cardiac chamber volumes using the CTCS scans have an excellent correlation when compared to the volumes in corresponding CCTA scans, a Pearson correlation coefficient (R) of 0.95 is obtained. Our fully automatic method enables large-scale assessment of cardiac structures on non-contrast-enhanced CT scans.

Reference standard space hippocampus labels according to the European Alzheimer's Disease Consortium-Alzheimer's Disease Neuroimaging Initiative harmonized protocol: Utility in automated volumetry

INTRODUCTION

A harmonized protocol (HarP) for manual hippocampal segmentation on magnetic resonance imaging (MRI) has recently been developed by an international European Alzheimer's Disease Consortium-Alzheimer's Disease Neuroimaging Initiative project. We aimed at providing consensual certified HarP hippocampal labels in Montreal Neurological Institute (MNI) standard space to serve as reference in automated image analyses.

METHODS

Manual HarP tracings on the high-resolution MNI152 standard space template of four expert certified HarP tracers were combined to obtain consensual bilateral hippocampus labels. Utility and validity of these reference labels is demonstrated in a simple atlas-based morphometry approach for automated calculation of HarP-compliant hippocampal volumes within SPM software.

RESULTS

Individual tracings showed very high agreement among the four expert tracers (pairwise Jaccard indices 0.82-0.87). Automatically calculated hippocampal volumes were highly correlated (r_L/R = 0.89/0.91) with gold standard volumes in the HarP benchmark data set (N = 135 MRIs), with a mean volume difference of 9% (standard deviation 7%).

CONCLUSION

The consensual HarP hippocampus labels in the MNI152 template can serve as a reference standard for automated image analyses involving MNI standard space normalization.

Multi-Modal and Targeted Imaging Improves Automated Mid-Brain Segmentation

The basal ganglia and limbic system, particularly the thalamus, putamen, internal and external globus pallidus, substantia nigra, and sub-thalamic nucleus, comprise a clinically relevant signal network for Parkinson's disease. In order to manually trace these structures, a combination of high-resolution and specialized sequences at 7T are used, but it is not feasible to scan clinical patients in those scanners. Targeted imaging sequences at 3T such as F-GATIR, and other optimized inversion recovery sequences, have been presented which enhance contrast in a select group of these structures. In this work, we show that a series of atlases generated at 7T can be used to accurately segment these structures at 3T using a combination of standard and optimized imaging sequences, though no one approach provided the best result across all structures. In the thalamus and putamen, a median Dice coefficient over 0.88 and a mean surface distance less than 1.0mm was achieved using a combination of T1 and an optimized inversion recovery imaging sequences. In the internal and external globus pallidus a Dice over 0.75 and a mean surface distance less than 1.2mm was achieved using a combination of T1 and F-GATIR imaging sequences. In the substantia nigra and sub-thalamic nucleus a Dice coefficient of over 0.6 and a mean surface distance of less than 1.0mm was achieved using the optimized inversion recovery imaging sequence. On average, using T1 and optimized inversion recovery together produced significantly improved segmentation results than any individual modality (p<0.05 wilcox sign-rank test).

A Combined Random Forests and Active Contour Model Approach for Fully Automatic Segmentation of the Left Atrium in Volumetric MRI

Segmentation of the left atrium (LA) from cardiac magnetic resonance imaging (MRI) datasets is of great importance for image guided atrial fibrillation ablation, LA fibrosis quantification, and cardiac biophysical modelling. However, automated LA segmentation from cardiac MRI is challenging due to limited image resolution, considerable variability in anatomical structures across subjects, and dynamic motion of the heart. In this work, we propose a combined random forests (RFs) and active contour model (ACM) approach for fully automatic segmentation of the LA from cardiac volumetric MRI. Specifically, we employ the RFs within an autocontext scheme to effectively integrate contextual and appearance information from multisource images together for LA shape inferring. The inferred shape is then incorporated into a volume-scalable ACM for further improving the segmentation accuracy. We validated the proposed method on the cardiac volumetric MRI datasets from the STACOM 2013 and HVSMR 2016 databases and showed that it outperforms other latest automated LA segmentation methods. Validation metrics, average Dice coefficient (DC) and average surface-to-surface distance (S2S), were computed as 0.9227 ± 0.0598 and 1.14 ± 1.205 mm, versus those of 0.6222-0.878 and 1.34-8.72 mm, obtained by other methods, respectively.

Validation of a method for retroperitoneal tumor segmentation

PURPOSE

In 2005, an application for surgical planning called AYRA[Formula: see text] was designed and validated by different surgeons and engineers at the Virgen del Rocío University Hospital, Seville (Spain). However, the segmentation methods included in AYRA and in other surgical planning applications are not able to segment accurately tumors that appear in soft tissue. The aims of this paper are to offer an exhaustive validation of an accurate semiautomatic segmentation tool to delimitate retroperitoneal tumors from CT images and to aid physicians in planning both radiotherapy doses and surgery.

METHODS

A panel of 6 experts manually segmented 11 cases of tumors, and the segmentation results were compared exhaustively with: the results provided by a surgical planning tool (AYRA), the segmentations obtained using a radiotherapy treatment planning system (Pinnacle[Formula: see text]), the segmentation results obtained by a group of experts in the delimitation of retroperitoneal tumors and the segmentation results using the algorithm under validation.

RESULTS

11 cases of retroperitoneal tumors were tested. The proposed algorithm provided accurate results regarding the segmentation of the tumor. Moreover, the algorithm requires minimal computational time-an average of 90.5% less than that required when manually contouring the same tumor.

CONCLUSION

A method developed for the semiautomatic selection of retroperitoneal tumor has been validated in depth. AYRA, as well as other surgical and radiotherapy planning tools, could be greatly improved by including this algorithm.

Multi-atlas Segmentation Enables Robust Multi-contrast MRI Spleen Segmentation for Splenomegaly

Non-invasive spleen volume estimation is essential in detecting splenomegaly. Magnetic resonance imaging (MRI) has been used to facilitate splenomegaly diagnosis in vivo. However, achieving accurate spleen volume estimation from MR images is challenging given the great inter-subject variance of human abdomens and wide variety of clinical images/modalities. Multi-atlas segmentation has been shown to be a promising approach to handle heterogeneous data and difficult anatomical scenarios. In this paper, we propose to use multi-atlas segmentation frameworks for MRI spleen segmentation for splenomegaly. To the best of our knowledge, this is the first work that integrates multi-atlas segmentation for splenomegaly as seen on MRI. To address the particular concerns of spleen MRI, automated and novel semi-automated atlas selection approaches are introduced. The automated approach interactively selects a subset of atlases using selective and iterative method for performance level estimation (SIMPLE) approach. To further control the outliers, semi-automated craniocaudal length based SIMPLE atlas selection (L-SIMPLE) is proposed to introduce a spatial prior in a fashion to guide the iterative atlas selection. A dataset from a clinical trial containing 55 MRI volumes (28 T1 weighted and 27 T2 weighted) was used to evaluate different methods. Both automated and semi-automated methods achieved median DSC > 0.9. The outliers were alleviated by the L-SIMPLE (≈1 min manual efforts per scan), which achieved 0.9713 Pearson correlation compared with the manual segmentation. The results demonstrated that the multi-atlas segmentation is able to achieve accurate spleen segmentation from the multi-contrast splenomegaly MRI scans.

3D-SIFT-Flow for atlas-based CT liver image segmentation

PURPOSE

In this paper, the authors proposed a new 3D registration algorithm, 3D-scale invariant feature transform (SIFT)-Flow, for multiatlas-based liver segmentation in computed tomography (CT) images.

METHODS

In the registration work, the authors developed a new registration method that takes advantage of dense correspondence using the informative and robust SIFT feature. The authors computed the dense SIFT features for the source image and the target image and designed an objective function to obtain the correspondence between these two images. Labeling of the source image was then mapped to the target image according to the former correspondence, resulting in accurate segmentation. In the fusion work, the 2D-based nonparametric label transfer method was extended to 3D for fusing the registered 3D atlases.

RESULTS

Compared with existing registration algorithms, 3D-SIFT-Flow has its particular advantage in matching anatomical structures (such as the liver) that observe large variation/deformation. The authors observed consistent improvement over widely adopted state-of-the-art registration methods such as ELASTIX, ANTS, and multiatlas fusion methods such as joint label fusion. Experimental results of liver segmentation on the MICCAI 2007 Grand Challenge are encouraging, e.g., Dice overlap ratio 96.27% ± 0.96% by our method compared with the previous state-of-the-art result of 94.90% ± 2.86%.

CONCLUSIONS

Experimental results show that 3D-SIFT-Flow is robust for segmenting the liver from CT images, which has large tissue deformation and blurry boundary, and 3D label transfer is effective and efficient for improving the registration accuracy.

Efficient Multi-Atlas Registration using an Intermediate Template Image

Multi-atlas label fusion is an accurate but time-consuming method of labeling the human brain. Using an intermediate image as a registration target can allow researchers to reduce time constraints by storing the deformations required of the atlas images. In this paper, we investigate the effect of registration through an intermediate template image on multi-atlas label fusion and propose a novel registration technique to counteract the negative effects of through-template registration. We show that overall computation time can be decreased dramatically with minimal impact on final label accuracy and time can be exchanged for improved results in a predictable manner. We see almost complete recovery of Dice similarity over a simple through-template registration using the corrected method and still maintain a 3-4 times speed increase. Further, we evaluate the effectiveness of this method on brains of patients with normal-pressure hydrocephalus, where abnormal brain shape presents labeling difficulties, specifically the ventricular labels. Our correction method creates substantially better ventricular labeling than traditional methods and maintains the speed increase seen in healthy subjects.

SEGMA: An Automatic SEGMentation Approach for Human Brain MRI Using Sliding Window and Random Forests

Quantitative volumes from brain magnetic resonance imaging (MRI) acquired across the life course may be useful for investigating long term effects of risk and resilience factors for brain development and healthy aging, and for understanding early life determinants of adult brain structure. Therefore, there is an increasing need for automated segmentation tools that can be applied to images acquired at different life stages. We developed an automatic segmentation method for human brain MRI, where a sliding window approach and a multi-class random forest classifier were applied to high-dimensional feature vectors for accurate segmentation. The method performed well on brain MRI data acquired from 179 individuals, analyzed in three age groups: newborns (38–42 weeks gestational age), children and adolescents (4–17 years) and adults (35–71 years). As the method can learn from partially labeled datasets, it can be used to segment large-scale datasets efficiently. It could also be applied to different populations and imaging modalities across the life course.

Metric Learning for Multi-atlas based Segmentation of Hippocampus

Automatic and reliable segmentation of hippocampus from MR brain images is of great importance in studies of neurological diseases, such as epilepsy and Alzheimer's disease. In this paper, we proposed a novel metric learning method to fuse segmentation labels in multi-atlas based image segmentation. Different from current label fusion methods that typically adopt a predefined distance metric model to compute a similarity measure between image patches of atlas images and the image to be segmented, we learn a distance metric model from the atlases to keep image patches of the same structure close to each other while those of different structures are separated. The learned distance metric model is then used to compute the similarity measure between image patches in the label fusion. The proposed method has been validated for segmenting hippocampus based on the EADC-ADNI dataset with manually labelled hippocampus of 100 subjects. The experiment results demonstrated that our method achieved statistically significant improvement in segmentation accuracy, compared with state-of-the-art multi-atlas image segmentation methods.

Reproducibility of superficial white matter tracts using diffusion-weighted imaging tractography

Human brain connection map is far from being complete. In particular the study of the superficial white matter (SWM) is an unachieved task. Its description is essential for the understanding of human brain function and the study of pathogenesis triggered by abnormal connectivity. In this work we automatically created a multi-subject atlas of SWM diffusion-based bundles of the whole brain. For each subject, the complete cortico-cortical tractogram is first split into sub-tractograms connecting pairs of gyri. Then intra-subject shape-based fiber clustering performs compression of each sub-tractogram into a set of bundles. Proceeding further with shape-based clustering provides a match of the bundles across subjects. Bundles found in most of the subjects are instantiated in the atlas. To increase robustness, this procedure was performed with two independent groups of subjects, in order to discard bundles without match across the two independent atlases. Finally, the resulting intersection atlas was projected on a third independent group of subjects in order to filter out bundles without reproducible and reliable projection. The final multi-subject diffusion-based U-fiber atlas is composed of 100 bundles in total, 50 per hemisphere, from which 35 are common to both hemispheres.

Extended Modality Propagation: Image Synthesis of Pathological Cases

This paper describes a novel generative model for the synthesis of multi-modal medical images of pathological cases based on a single label map. Our model builds upon i) a generative model commonly used for label fusion and multi-atlas patch-based segmentation of healthy anatomical structures, ii) the Modality Propagation iterative strategy used for a spatially-coherent synthesis of subject-specific scans of desired image modalities. The expression Extended Modality Propagation is coined to refer to the extension of Modality Propagation to the synthesis of images of pathological cases. Moreover, image synthesis uncertainty is estimated. An application to Magnetic Resonance Imaging synthesis of glioma-bearing brains is i) validated on the training dataset of a Multimodal Brain Tumor Image Segmentation challenge, ii) compared to the state-of-the-art in glioma image synthesis, and iii) illustrated using the output of two different tumor growth models. Such a generative model allows the generation of a large dataset of synthetic cases, which could prove useful for the training, validation, or benchmarking of image processing algorithms.

Simultaneous total intracranial volume and posterior fossa volume estimation using multi-atlas label fusion

Total intracranial volume (TICV) is an essential covariate in brain volumetric analyses. The prevalent brain imaging software packages provide automatic TICV estimates. FreeSurfer and FSL estimate TICV using a scaling factor while SPM12 accumulates probabilities of brain tissues. None of the three provide explicit skull/CSF boundary (SCB) since it is challenging to distinguish these dark structures in a T1-weighted image. However, explicit SCB not only leads to a natural way of obtaining TICV (i.e., counting voxels inside the skull) but also allows sub-definition of TICV, for example, the posterior fossa volume (PFV). In this article, they proposed to use multi-atlas label fusion to obtain TICV and PFV simultaneously. The main contributions are: (1) TICV and PFV are simultaneously obtained with explicit SCB from a single T1-weighted image. (2) TICV and PFV labels are added to the widely used BrainCOLOR atlases. (3) Detailed mathematical derivation of non-local spatial STAPLE (NLSS) label fusion is presented. As the skull is clearly distinguished in CT images, we use a semi-manual procedure to obtain atlases with TICV and PFV labels using 20 subjects who both have a MR and CT scan. The proposed method provides simultaneous TICV and PFV estimation while achieving more accurate TICV estimation compared with FreeSurfer, FSL, SPM12, and the previously proposed STAPLE based approach. The newly developed TICV and PFV labels for the OASIS BrainCOLOR atlases provide acceptable performance, which enables simultaneous TICV and PFV estimation during whole brain segmentation. The NLSS method and the new atlases have been made freely available. Hum Brain Mapp 38:599-616, 2017. © 2016 Wiley Periodicals, Inc.

Globally Optimal Label Fusion with Shape Priors

Multi-atlas label fusion methods have gained popularity in a variety of segmentation tasks given their attractive performance. Graph-based segmentation methods are widely used given their global optimality guarantee. We propose a novel approach, GOLF, that combines the strengths of these two approaches. GOLF incorporates shape priors to the label-fusion problem and provides a globally optimal solution even for the multi-label scenario, while also leveraging the highly accurate posterior maps from a multi-atlas label fusion approach. We demonstrate GOLF for the joint segmentation of the left and right pairs of caudate, putamen, globus pallidus and nucleus accumbens. Compared to the FreeSurfer and FIRST approaches, GOLF is significantly more accurate on all reported indices for all 8 structures. We also present comparisons to a multi-atlas approach, which reveals further insights on the contributions of the different components of the proposed framework.

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.

Abdomen and spinal cord segmentation with augmented active shape models

Active shape models (ASMs) have been widely used for extracting human anatomies in medical images given their capability for shape regularization of topology preservation. However, sensitivity to model initialization and local correspondence search often undermines their performances, especially around highly variable contexts in computed-tomography (CT) and magnetic resonance (MR) images. In this study, we propose an augmented ASM (AASM) by integrating the multiatlas label fusion (MALF) and level set (LS) techniques into the traditional ASM framework. Using AASM, landmark updates are optimized globally via a region-based LS evolution applied on the probability map generated from MALF. This augmentation effectively extends the searching range of correspondent landmarks while reducing sensitivity to the image contexts and improves the segmentation robustness. We propose the AASM framework as a two-dimensional segmentation technique targeting structures with one axis of regularity. We apply AASM approach to abdomen CT and spinal cord (SC) MR segmentation challenges. On 20 CT scans, the AASM segmentation of the whole abdominal wall enables the subcutaneous/visceral fat measurement, with high correlation to the measurement derived from manual segmentation. On 28 3T MR scans, AASM yields better performances than other state-of-the-art approaches in segmenting white/gray matter in SC.