MAP MRF joint segmentation and registration of medical images

Wyatt, P. B.; Noble, J. Alison

doi:10.1016/s1361-8415(03)00067-7

Cited by 87 publications

(52 citation statements)

References 17 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Voxel-based classification methods that have coupled registration and segmentation of misaligned spectral images include (Wyatt and Noble, 2002;Xiaohua et al, 2005 our purpose is to align an atlas to MR images and separate the images into anatomical structures, as also suggested by (Ashburner and Friston, 2005). The proposed non-rigid registration framework of Ashburner and Friston (2005) performs a local deformation of the atlas space under the assumption that the atlas is globally aligned to the image space.…”

Section: Introductionmentioning

confidence: 99%

“…This paper describes an integrated segmentation and registration approach related to voxel-based classification methods, which considers the anatomical structure associated with each voxel within a Bayesian framework (Wells et al, 1996;Van Leemput et al, 1999;Marroquin et al, 2003;Pohl et al, 2004a). These voxel-based classification methods explicitly model image artifacts in order to segment large data sets without manual intervention.Voxel-based classification methods that have coupled registration and segmentation of misaligned spectral images include (Wyatt and Noble, 2002;Xiaohua et al, 2005 our purpose is to align an atlas to MR images and separate the images into anatomical structures, as also suggested by (Ashburner and Friston, 2005). The proposed non-rigid registration framework of Ashburner and Friston (2005) performs a local deformation of the atlas space under the assumption that the atlas is globally aligned to the image space.…”

mentioning

confidence: 99%

See 1 more Smart Citation

A Bayesian model for joint segmentation and registration

et al. 2006

View full text Add to dashboard Cite

A statistical model is presented that combines the registration of an atlas with the segmentation of magnetic resonance images. We use an Expectation Maximization-based algorithm to find a solution within the model, which simultaneously estimates image artifacts, anatomical labelmaps, and a structure-dependent hierarchical mapping from the atlas to the image space. The algorithm produces segmentations for brain tissues as well as their substructures. We demonstrate the approach on a set of 22 magnetic resonance images. On this set of images, the new approach performs significantly better than similar methods which sequentially apply registration and segmentation. D 2005 Elsevier Inc. All rights reserved.Keywords: Registration; Segmentation; Subcortical segmentation; Bayesian modeling; Expectation -Maximization IntroductionTo better understand brain diseases, many neuroscience studies focus on the anatomical differences between control and diseased subjects. In order to find these differences, scientists often analyze medical images for brain structures which seem to be influenced by the disease. The analysis is frequently based on segmentations of the structures of interest that are mostly performed by human experts. However, this manual process is not only very expensive, but in addition, it increases risks related to inter-and intra-observer reliability (Kikinis et al., 1992). Neuroscientists are keenly interested in automatic methods, which often rely on prior information, to perform this task (Collins et al., 1999;Leventon et al., 2000;Marroquin et al., 2003;Fischl et al., 2004;Pohl et al., 2004a;Ashburner and Friston, 2005). With notable exceptions, these methods first register the prior information, i.e., an atlas, to the medical image and then segment the medical image into anatomical structures based on that aligned information. The goal of this work is to unify this process into a single Bayesian framework in order to overcome biases caused by commitment to the initial registration.When automatic segmentation methods are guided by prior information, they frequently are used to segment anatomical structures defined by weakly visible boundaries in medical images. For example, the intensity properties of the thalamus in T1-weighted magnetic resonance (MR) images are very similar to those of the neighboring white matter (Fig. 1). Algorithms cannot rely on the MR images alone in order to distinguish these two structures. However, the ventricles, the dark structures above the thalamus, are more easily identified. In order for the ventricles to guide the detection of the boundary between the thalamus and the white matter, automatic segmentation algorithms use spatial priors (Mazziotta et al., 1995;Thompson et al., 1996). These spatial priors capture the relationship between structures such as the fact that the ventricles are above the thalamus.As mentioned previously, most atlas-based algorithms perform registration and segmentation sequentially (Cocosco et al., 2003;Van Leemput et al., 1999;Fischl et al., 200...

show abstract

Section: Introductionmentioning

confidence: 99%

mentioning

confidence: 99%

A Bayesian model for joint segmentation and registration

et al. 2006

View full text Add to dashboard Cite

show abstract

“…Few results have been reported in mice (7,8). In our experience, user correction of errors arising from automated analysis requires that user interaction be intrinsic, intuitive, and minimal.…”

mentioning

confidence: 99%

Fast left ventricular mass and volume assessment in mice with three‐dimensional guide‐point modeling

Young

Barnes

Davison

et al. 2009

Magnetic Resonance Imaging

View full text Add to dashboard Cite

Purpose: To investigate the accuracy (vs. standard manual analysis) and precision (scan-rescan reproducibility) of three-dimensional guide-point modeling (GPM) for the assessment of left ventricular (LV) function in mice. Methods:Six male wildtype C57/Bl6 mice (weight 26.2 Ϯ 1.1 g) were scanned twice, 3 days apart. Each scan was performed twice, at 0.2 mm/pixel with one average and at 0.1 mm/pixel with two averages. The 24 studies were anonymized and analyzed in blinded fashion using GPM and standard manual slice summation. Results:The average error between GPM and standard analysis was 2.3 Ϯ 5.8 mg in mass, 1.7 Ϯ 3.2 L in enddiastolic volume, 2.3 Ϯ 3.1 L in end-systolic volume, Ϫ2.7 Ϯ 4.3% in ejection fraction, Ϫ0.6 Ϯ 3.3 L in stroke volume, and Ϫ0.31 Ϯ 1.56 ml ⅐ min Ϫ1 in cardiac output (mean difference Ϯ SD of differences, n ϭ 24). The average time taken was 8.0 Ϯ 2.5 minutes for 3D GPM and 48.5 Ϯ 8.9 minutes for standard analysis (n ϭ 24). Scan-rescan reproducibility results were similar to the standard analysis. No significant differences were found using linear mixed effects modeling in either accuracy or precision between scan resolutions or analysis method. Conclusion:3D GPM enables fast analysis of mouse LV function, with similar accuracy and reproducibility to standard analysis. An image resolution of 0.2 mm/pixel with one average is adequate for LV function studies.

show abstract

“…The word "automatic" is therefore not fully correct for most segmentation programs, as some user input, such as starting contours of the segmentation, is required to accomplish the segmentation task. Although strategies for automatic segmentation of the mouse heart have been reported (27), the performance of automatic segmentation procedures within a large population of mouse hearts has not been addressed so far. The purpose of this study was therefore to evaluate manual and automatic segmentation of short-axis CINE MR images.…”

mentioning

confidence: 99%

Evaluation of manual and automatic segmentation of the mouse heart from CINE MR images

Heijman

Aben²,

Penners³

et al. 2007

Magnetic Resonance Imaging

View full text Add to dashboard Cite

Purpose:To compare global functional parameters determined from a stack of cinematographic MR images of mouse heart by a manual segmentation and an automatic segmentation algorithm. Materials and Methods:The manual and automatic segmentation results of 22 mouse hearts were compared. The automatic segmentation was based on propagation of a minimum cost algorithm in polar space starting from manually drawn contours in one heart phase. Intra-and interobserver variability as well as validity of the automatic segmentation was determined. To test the reproducibility of the algorithm the variability was calculated from the intraand interobserver input. Results:The mean time of segmentation for one dataset was around 10 minutes and Ϸ2.5 hours for automatic and manual segmentation, respectively. There were no significant differences between the automatic and the manual segmentation except for the end systolic epicardial volume. The automatically derived volumes correlated well with the manually derived volumes (R 2 ϭ 0.90); left ventricular mass with and without papillary muscle showed a correlation R 2 of 0.74 and 0.76, respectively. The manual intraobserver variability was superior to the interobserver variability and the variability of the automatic segmentation, while the manual interobserver variability was comparable to the variability of the automatic segmentation. The automatic segmentation algorithm reduced the bias of the intra-and interobserver variability. Conclusion:We conclude that automatic segmentation of the mouse heart provides a fast and valid alternative to manual segmentation of the mouse heart.

show abstract

MAP MRF joint segmentation and registration of medical images

Cited by 87 publications

References 17 publications

A Bayesian model for joint segmentation and registration

A Bayesian model for joint segmentation and registration

Fast left ventricular mass and volume assessment in mice with three‐dimensional guide‐point modeling

Evaluation of manual and automatic segmentation of the mouse heart from CINE MR images

Contact Info

Product

Resources

About