Multi-atlas segmentation of subcortical brain structures via the AutoSeg software pipeline.

Automated segmenting and labeling of individual brain anatomical regions, in MRI are challenging, due to the issue of individual structural variability. Although atlas-based segmentation has shown its potential for both tissue and structure segmentation, due to the inherent natural variability as well as disease-related changes in MR appearance, a single atlas image is often inappropriate to represent the full population of datasets processed in a given neuroimaging study. As an alternative for the case of single atlas segmentation, the use of multiple atlases alongside label fusion techniques has been introduced using a set of individual "atlases" that encompasses the expected variability in the studied population. In our study, we proposed a multi-atlas segmentation scheme with a novel graph-based atlas selection technique. We first paired and co-registered all atlases and the subject MR scans. A directed graph with edge weights based on intensity and shape similarity between all MR scans is then computed. The set of neighboring templates is selected via clustering of the graph. Finally, weighted majority voting is employed to create the final segmentation over the selected atlases. This multi-atlas segmentation scheme is used to extend a single-atlas-based segmentation toolkit entitled AutoSeg, which is an open-source, extensible C++ based software pipeline employing BatchMake for its pipeline scripting, developed at the Neuro Image Research and Analysis Laboratories of the University of North Carolina at Chapel Hill. AutoSeg performs N4 intensity inhomogeneity correction, rigid registration to a common template space, automated brain tissue classification based skull-stripping, and the multi-atlas segmentation. The multi-atlas-based AutoSeg has been evaluated on subcortical structure segmentation with a testing dataset of 20 adult brain MRI scans and 15 atlas MRI scans. The AutoSeg achieved mean Dice coefficients of 81.73% for the subcortical structures.

ITK: enabling reproducible research and open science.

Reproducibility verification is essential to the practice of the scientific method. Researchers report their findings, which are strengthened as other independent groups in the scientific community share similar outcomes. In the many scientific fields where software has become a fundamental tool for capturing and analyzing data, this requirement of reproducibility implies that reliable and comprehensive software platforms and tools should be made available to the scientific community. The tools will empower them and the public to verify, through practice, the reproducibility of observations that are reported in the scientific literature. Medical image analysis is one of the fields in which the use of computational resources, both software and hardware, are an essential platform for performing experimental work. In this arena, the introduction of the Insight Toolkit (ITK) in 1999 has transformed the field and facilitates its progress by accelerating the rate at which algorithmic implementations are developed, tested, disseminated and improved. By building on the efficiency and quality of open source methodologies, ITK has provided the medical image community with an effective platform on which to build a daily workflow that incorporates the true scientific practices of reproducibility verification. This article describes the multiple tools, methodologies, and practices that the ITK community has adopted, refined, and followed during the past decade, in order to become one of the research communities with the most modern reproducibility verification infrastructure. For example, 207 contributors have created over 2400 unit tests that provide over 84% code line test coverage. The Insight Journal, an open publication journal associated with the toolkit, has seen over 360,000 publication downloads. The median normalized closeness centrality, a measure of knowledge flow, resulting from the distributed peer code review system was high, 0.46.

An efficient, scalable, and adaptable framework for solving generic systems of level-set PDEs.

In the last decade, level-set methods have been actively developed for applications in image registration, segmentation, tracking, and reconstruction. However, the development of a wide variety of level-set PDEs and their numerical discretization schemes, coupled with hybrid combinations of PDE terms, stopping criteria, and reinitialization strategies, has created a software logistics problem. In the absence of an integrative design, current toolkits support only specific types of level-set implementations which restrict future algorithm development since extensions require significant code duplication and effort. In the new NIH/NLM Insight Toolkit (ITK) v4 architecture, we implemented a level-set software design that is flexible to different numerical (continuous, discrete, and sparse) and grid representations (point, mesh, and image-based). Given that a generic PDE is a summation of different terms, we used a set of linked containers to which level-set terms can be added or deleted at any point in the evolution process. This container-based approach allows the user to explore and customize terms in the level-set equation at compile-time in a flexible manner. The framework is optimized so that repeated computations of common intensity functions (e.g., gradient and Hessians) across multiple terms is eliminated. The framework further enables the evolution of multiple level-sets for multi-object segmentation and processing of large datasets. For doing so, we restrict level-set domains to subsets of the image domain and use multithreading strategies to process groups of subdomains or level-set functions. Users can also select from a variety of reinitialization policies and stopping criteria. Finally, we developed a visualization framework that shows the evolution of a level-set in real-time to help guide algorithm development and parameter optimization. We demonstrate the power of our new framework using confocal microscopy images of cells in a developing zebrafish embryo.

Multi-atlas segmentation with joint label fusion and corrective learning-an open source implementation.

Label fusion based multi-atlas segmentation has proven to be one of the most competitive techniques for medical image segmentation. This technique transfers segmentations from expert-labeled images, called atlases, to a novel image using deformable image registration. Errors produced by label transfer are further reduced by label fusion that combines the results produced by all atlases into a consensus solution. Among the proposed label fusion strategies, weighted voting with spatially varying weight distributions derived from atlas-target intensity similarity is a simple and highly effective label fusion technique. However, one limitation of most weighted voting methods is that the weights are computed independently for each atlas, without taking into account the fact that different atlases may produce similar label errors. To address this problem, we recently developed the joint label fusion technique and the corrective learning technique, which won the first place of the 2012 MICCAI Multi-Atlas Labeling Challenge and was one of the top performers in 2013 MICCAI Segmentation: Algorithms, Theory and Applications (SATA) challenge. To make our techniques more accessible to the scientific research community, we describe an Insight-Toolkit based open source implementation of our label fusion methods. Our implementation extends our methods to work with multi-modality imaging data and is more suitable for segmentation problems with multiple labels. We demonstrate the usage of our tools through applying them to the 2012 MICCAI Multi-Atlas Labeling Challenge brain image dataset and the 2013 SATA challenge canine leg image dataset. We report the best results on these two datasets so far.

Explicit B-spline regularization in diffeomorphic image registration.

Diffeomorphic mappings are central to image registration due largely to their topological properties and success in providing biologically plausible solutions to deformation and morphological estimation problems. Popular diffeomorphic image registration algorithms include those characterized by time-varying and constant velocity fields, and symmetrical considerations. Prior information in the form of regularization is used to enforce transform plausibility taking the form of physics-based constraints or through some approximation thereof, e.g., Gaussian smoothing of the vector fields [a la Thirion's Demons (Thirion, 1998)]. In the context of the original Demons' framework, the so-called directly manipulated free-form deformation (DMFFD) (Tustison et al., 2009) can be viewed as a smoothing alternative in which explicit regularization is achieved through fast B-spline approximation. This characterization can be used to provide B-spline "flavored" diffeomorphic image registration solutions with several advantages. Implementation is open source and available through the Insight Toolkit and our Advanced Normalization Tools (ANTs) repository. A thorough comparative evaluation with the well-known SyN algorithm (Avants et al., 2008), implemented within the same framework, and its B-spline analog is performed using open labeled brain data and open source evaluation tools.

The Design of SimpleITK.

SimpleITK is a new interface to the Insight Segmentation and Registration Toolkit (ITK) designed to facilitate rapid prototyping, education and scientific activities via high level programming languages. ITK is a templated C++ library of image processing algorithms and frameworks for biomedical and other applications, and it was designed to be generic, flexible and extensible. Initially, ITK provided a direct wrapping interface to languages such as Python and Tcl through the WrapITK system. Unlike WrapITK, which exposed ITK's complex templated interface, SimpleITK was designed to provide an easy to use and simplified interface to ITK's algorithms. It includes procedural methods, hides ITK's demand driven pipeline, and provides a template-less layer. Also SimpleITK provides practical conveniences such as binary distribution packages and overloaded operators. Our user-friendly design goals dictated a departure from the direct interface wrapping approach of WrapITK, toward a new facade class structure that only exposes the required functionality, hiding ITK's extensive template use. Internally SimpleITK utilizes a manual description of each filter with code-generation and advanced C++ meta-programming to provide the higher-level interface, bringing the capabilities of ITK to a wider audience. SimpleITK is licensed as open source software library under the Apache License Version 2.0 and more information about downloading it can be found at http://www.simpleitk.org.

Brain extraction using the watershed transform from markers.

Isolation of the brain from other tissue types in magnetic resonance (MR) images is an important step in many types of neuro-imaging research using both humans and animal subjects. The importance of brain extraction is well appreciated-numerous approaches have been published and the benefits of good extraction methods to subsequent processing are well known. We describe a tool-the marker based watershed scalper (MBWSS)-for isolating the brain in T1-weighted MR images built using filtering and segmentation components from the Insight Toolkit (ITK) framework. The key elements of MBWSS-the watershed transform from markers and aggressive filtering with large kernels-are techniques that have rarely been used in neuroimaging segmentation applications. MBWSS is able to reliably isolate the brain without expensive preprocessing steps, such as registration to an atlas, and is therefore useful as the first stage of processing pipelines. It is an informative example of the level of accuracy achievable without using priors in the form of atlases, shape models or libraries of examples. We validate the MBWSS using a publicly available dataset, a paediatric cohort, an adolescent cohort, intra-surgical scans and demonstrate flexibility of the approach by modifying the method to extract macaque brains.