Elsevier

NeuroImage

Volume 31, Issue 3, 1 July 2006, Pages 1116-1128
NeuroImage

User-guided 3D active contour segmentation of anatomical structures: Significantly improved efficiency and reliability

https://doi.org/10.1016/j.neuroimage.2006.01.015Get rights and content

Abstract

Active contour segmentation and its robust implementation using level set methods are well-established theoretical approaches that have been studied thoroughly in the image analysis literature. Despite the existence of these powerful segmentation methods, the needs of clinical research continue to be fulfilled, to a large extent, using slice-by-slice manual tracing. To bridge the gap between methodological advances and clinical routine, we developed an open source application called ITK-SNAP, which is intended to make level set segmentation easily accessible to a wide range of users, including those with little or no mathematical expertise. This paper describes the methods and software engineering philosophy behind this new tool and provides the results of validation experiments performed in the context of an ongoing child autism neuroimaging study. The validation establishes SNAP intrarater and interrater reliability and overlap error statistics for the caudate nucleus and finds that SNAP is a highly reliable and efficient alternative to manual tracing. Analogous results for lateral ventricle segmentation are provided.

Introduction

Segmentation of anatomical structures in medical images is a fundamental task in neuroimaging research. Segmentation is used to measure the size and shape of brain structures, to guide spatial normalization of anatomy between individuals and to plan medical intervention. Segmentation serves as an essential element in a great number of morphometry studies that test various hypotheses about the pathology and pathophysiology of neurological disorders. The spectrum of available segmentation approaches is broad, ranging from manual outlining of structures in 2D cross-sections to cutting-edge methods that use deformable registration to find optimal correspondences between 3D images and a labeled atlas (Haller et al., 1997, Goldszal et al., 1998). Amid this spectrum lie semiautomatic approaches that combine the efficiency and repeatability of automatic segmentation with the sound judgement that can only come from human expertise. One class of semiautomatic methods formulates the problem of segmentation in terms of active contour evolution (Zhu and Yuille, 1996, Caselles et al., 1997, Sethian, 1999), where the human expert must specify the initial contour, balance the various forces which act upon it, as well as monitor the evolution.

Despite the fact that a large number of fully automatic and semiautomatic segmentation methods has been described in the literature, many brain research laboratories continue to use manual delineation as the technique of choice for image segmentation. Reluctance to embrace the fully automatic approach may be due to the concerns about its insufficient reliability in cases where the target anatomy may difference from the norm, as well as due to high computational demands of the approach based on image registration. However, the slow spread of semiautomatic segmentation may simply be due to the lack of readily available simple user interfaces. Semiautomatic methods require the user to specify various parameters, whose values tend to make sense only in the context of the method's mathematical formulation. We suspect that insufficient attention to developing tools that make parameter selection intuitive has prevented semiautomatic methods from replacing manual delineation as the tool of choice in the clinical research environment.

ITK-SNAP is a software application that brings active contour segmentation to the fingertips of clinical researchers. Our goals in developing this tool were (1) to focus specifically on the problem of segmenting anatomical structures, not allowing the kind of feature creep which would make the tool's learning curve prohibitively steep; (2) to construct a friendly and well-documented user interface that would break up the task of initialization and parameter selection into a series of intuitive steps; (3) to provide an integrated toolbox for manual postprocessing of segmentation results; and (4) to make the tool freely accessible and readily available through the open source mechanism. SNAP is a product of over 6 years of development in academic and corporate environments, and it is the largest end-user application bundled with the Insight Toolkit (ITK), a popular library of image analysis algorithms funded under the Visible Human Project by the U.S. National Library of Medicine (Ibanez et al., 2003). SNAP is available free of charge both as a stand-alone application that can be installed and executed quickly and as source code that can be used to derive new software.1

This paper provides a brief overview of the methods implemented in SNAP and describes the tool's core functionality. However, the paper's main focus is on the validation study, which we performed in order to demonstrate that SNAP is a viable alternative to manual segmentation. The validation was performed in the context of caudate nucleus segmentation in an ongoing child autism MRI study. Each caudate was segmented using both methods in multiple subjects by multiple highly trained raters and with multiple repetitions. The results of volume and overlap-based reliability analysis indicate that SNAP segmentation is very accurate, exceeding manual delineation in terms of efficiency and repeatability. We also demonstrate high reliability of SNAP in lateral ventricle segmentation.

The remainder of the paper is organized as follows. A short overview of automatic image segmentation, as well as some popular medical imaging tools that support it, is given in Section 2. A brief summary of active contour segmentation and level set methods appears in Section 3.1. Section 3.2 highlights the main features of SNAP's user interface and software architecture. Validation in the context of caudate and ventricle segmentation is presented in Section 4. Finally, Section 5 discusses the challenges of developing open-source image processing software, notes the limitations of SNAP segmentation, and brings up the need for complimentary tools, which we plan to develop in the future.

Section snippets

Previous work

In many clinical laboratories, biomedical image segmentation involves having a trained expert delineate the boundaries of anatomical structures in consecutive slices of 3D images. Although this approach puts the expert in full control of the segmentation, it is time consuming as well as error-prone. In the absence of feedback in 3D, contours traced in subsequent slices may become mismatched, resulting in unnatural jagged edges that pose a difficulty to applications such as shape analysis.

Active contour evolution

SNAP implements two well-known 3D active contour segmentation methods: Geodesic Active Contours by Caselles et al., 1993, Caselles et al., 1997 and Region Competition by Zhu and Yuille (1996). In both methods, the evolving estimate of the structure of interest is represented by one or more contours. An evolving contour is a closed surface C(u, v; t) parameterized by variables u, v and by the time variable t. The contour evolves according to the following partial differential equation (PDE):tC(

Results

The new SNAP tool, with its combination of user-guided 3D active contour segmentation and postprocessing via manual tracing in orthogonal slices or using the 3D cut-plane tool, is increasingly replacing conventional 2D slice editing for a variety of image segmentation tasks. SNAP is used in several large neuroimaging studies at UNC Chapel Hill, Duke University, and the University of Pennsylvania. Segmentation either uses the soft threshold option for the definition of foreground and background,

Discussion

The caudate segmentation validation, which compares the SNAP tool to manual segmentation by highly trained raters, demonstrates the excellent reliability of the tool for efficient three-dimensional segmentation. While the volume-based reliability analysis shows a similar range of intramethod reliability for both segmentation approaches, overlap analysis reveals that SNAP segmentation exhibits significantly improved repeatability. SNAP cut the segmentation time by a factor of three and also

Conclusion

ITK-SNAP is an open source medical image processing application that fulfills a specific and pressing need of biomedical imaging research by providing a combination of manual and semiautomatic tools for extracting structures in 3D image data of different modalities and from different anatomical regions. Designed to maximize user efficiency and to provide a smooth learning curve, the user interface is focused entirely on segmentation, parameter selection is simplified using live feedback, and

Acknowledgments

The integration of the SNAP tool with ITK was performed by Cognitica Corporation under NIH/NLM PO 467-MZ-202446-1. The validation study is supported by the NIH/NIBIB P01 EB002779, NIH Conte Center MH064065, and UNC Neurodevelopmental Disorders Research Center, Developmental Neuroimaging Core. The MRI images of infants and expert manual segmentations are funded by NIH RO1 MH61696 and NIMH MH 64580 (PI: Joseph Piven). Manual segmentations for the caudate study were done by Michael Graves and Todd

References (36)

  • T. Cootes et al.

    Active appearance models

  • R.H. Davies et al.

    A minimum description length approach to statistical shape modeling

    IEEE Trans. Med. Imag.

    (2002)
  • D. Gering et al.

    An integrated visualization system for surgical planning and guidance using image fusion and an open MR

    J. Magn. Reson. Imaging

    (2001)
  • A. Goldszal et al.

    An image processing system for qualitative and quantitative volumetric analysis of brain images

    J. Comput. Assist. Tomogr.

    (1998)
  • K. Gurleyik et al.

    Quantification of errors in volume measurements of the caudate nucleus using magnetic resonance imaging

    J. Magn. Reson. Imaging

    (2002)
  • J. Haller et al.

    Three-dimensional hippocampal MR morphometry by high-dimensional transformation of a neuroanatomic atlas

    Radiology

    (1997)
  • L. Ibanez et al.

    The ITK Software Guide. Kitware, Inc.

    (2003)
  • S. Joshi et al.

    Landmark matching via large deformation diffeomorphisms

    IEEE Trans. Image Process.

    (2000)
  • Cited by (6341)

    View all citing articles on Scopus
    View full text