3D Active Shape Models Using Gradient Descent Optimization of Description Length

Heimann, Tobias; Wolf, Ivo; Williams, T.G.; Meinzer, Hans-Peter

doi:10.1007/11505730_47

Cited by 77 publications

(74 citation statements)

References 23 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…The 3D MDL introduced by Heimann et al [96] was implemented and extended with some small improvements, as described in this chapter. If we look at the short result section, we see that the algorithm decreases the description length of the model, but has some problems:…”

Section: Discussionmentioning

confidence: 99%

“…We hebben verschillende methoden getest om deze punten te vinden. De eerste methode is minimum description length (MDL) [86] [96]. Deze methode vindt corresponderende punten door de punten zo te verplaatsen dat de variatie tussen de punten zo klein mogelijk wordt, en de beschrijving door de ASM zo compact mogelijk.…”

Section: Discussionunclassified

“…The first MDL algorithm was introduced by Davies et al [86]. In this chapter, we present an robust MDL method for 1D, 2D and 3D corresponding point optimization, which extends on the work of Heimann et al [96].…”

Section: Introductionmentioning

confidence: 94%

“…Therefore we want to calculate the derivative of the singular values of the SVD using the approach of Heimann et al [96]:…”

Section: Minimizing Description Lengthmentioning

confidence: 99%

“…Because the longitude is undefined on the poles, and wraps around at the meridian. Heimann et al [96], uses truncated Gaussian kernels which do not cover the poles of the sphere. Random rotation matrices are used to frequently change the location of the poles, to still optimize the whole object during the MDL optimization.…”

Section: Optimizing On a Spherementioning

confidence: 99%

See 4 more Smart Citations

Segmentation of the Mandibular Canal in Cone-Beam CT Data

Kroon¹

View full text Add to dashboard Cite

Section: Discussionmentioning

confidence: 99%

Section: Discussionunclassified

Section: Introductionmentioning

confidence: 94%

“…Therefore we want to calculate the derivative of the singular values of the SVD using the approach of Heimann et al [96]:…”

Section: Minimizing Description Lengthmentioning

confidence: 99%

Section: Optimizing On a Spherementioning

confidence: 99%

See 3 more Smart Citations

Segmentation of the Mandibular Canal in Cone-Beam CT Data

Kroon¹

View full text Add to dashboard Cite

Segmentation of human skull in MRI using statistical shape information from CT data

Wang

Shi

Chu

et al. 2009

Magnetic Resonance Imaging

View full text Add to dashboard Cite

Purpose: To automatically segment the skull from the MRI data using a model-based three-dimensional segmentation scheme. Materials and Methods:This study exploited the statistical anatomy extracted from the CT data of a group of subjects by means of constructing an active shape model of the skull surfaces. To construct a reliable shape model, a novel approach was proposed to optimize the automatic landmarking on the coupled surfaces (i.e., the skull vault) by minimizing the description length that incorporated local thickness information. This model was then used to locate the skull shape in MRI of a different group of patients. Results:Compared with performing landmarking separately on the coupled surfaces, the proposed landmarking method constructed models that had better generalization ability and specificity. The segmentation accuracies were measured by the Dice coefficient and the set difference, and compared with the method based on mathematical morphology operations. Conclusion:The proposed approach using the active shape model based on the statistical skull anatomy presented in the head CT data contributes to more reliable segmentation of the skull from MRI data. BECAUSE OF THE weak signal of bony structures in MRI and the high complexity of the skull topology, automatic segmentation of the skull with good reliability in MRI still remains a challenging problem. Whereas the segmentation of skull in MRI is crucial in research related to analyzing the skull morphology, such as the construction of realistic models of the head (1) and the quantification of bone growth abnormality (2). In such tasks, the motivation of using the MRI instead of computed tomography (CT) is to protect subjects from ionizing radiation and to acquire clear imaging of soft tissues at the same time for multiple analysis purposes.Being able to characterize shapes with natural variability, the active shape model (ASM) introduced by Cootes et al (3) found a variety of applications in medical image segmentation and analysis. Typically, when applied to medical image segmentation problems, the active shape model is built from a group of training shapes, which share the same modality as the images to be segmented (4). To define skull boundaries in MRI, the CT data are worthwhile to be referred to for anatomical information. The first reason is that the active shape model is unlikely to be acquired from the MRI data per se, as automatic segmentation of skull in MRI is the main problem to be solved while annotating the skull landmarks manually to establish the training set is infeasible. Second, brain atlases with skull information only represent the anatomy of a single subject, which is not statistically meaningful. Third, the bony structures are very clear in the CT data, and they can be easily segmented using simple thresholding.Thresholding and region growing are often applied to segment scalps, skulls, and brains (5). Dogdas et al (6) used a combination of thresholding and mathematical morphological operations to find the boundaries of sc...

show abstract

Open-Curve Shape Correspondence Without Endpoint Correspondence

Richardson

Wang

2006

Medical Image Computing and Computer-Assisted Intervention – MICCAI 2006

View full text Add to dashboard Cite

Abstract. Shape correspondence is the foundation for accurate statistical shape analysis; this is usually accomplished by identifying a set of sparsely sampled and well-corresponded landmark points across a population of shape instances. However, most available shape correspondence methods can only effectively deal with complete-shape correspondence, where a one-to-one mapping is assumed between any two shape instances. In this paper, we present a novel algorithm to correspond 2D open-curve partial-shape instances where one shape instance may only be mapped to part of the other, i.e., the endpoints of these open-curve shape instances are not presumably corresponded. In this algorithm, some initially identified landmarks, including the ones at or near the endpoints of the shape instances, are refined by allowing them to slide freely along the shape contour to minimize the shape-correspondence error. To avoid being trapped into local optima, we develop a simple method to construct a better initialization of the landmarks and introduce some additional constraints to the landmark sliding. We evaluate the proposed algorithm on 32 femur shape instances in comparison to some current methods.

show abstract

3D Active Shape Models Using Gradient Descent Optimization of Description Length

Cited by 77 publications

References 23 publications

Segmentation of the Mandibular Canal in Cone-Beam CT Data

Segmentation of the Mandibular Canal in Cone-Beam CT Data

Segmentation of human skull in MRI using statistical shape information from CT data

Open-Curve Shape Correspondence Without Endpoint Correspondence

Contact Info

Product

Resources

About