## A hierarchical statistical modeling approach for the unsupervised 3-D reconstruction of the scoliotic spine (2003)

Venue: | in Proc. 10th IEEE Int. Conf. Image Processing |

Citations: | 4 - 2 self |

### BibTeX

@INPROCEEDINGS{Benameur03ahierarchical,

author = {S. Benameur and M. Mignotte and S. Parent and H. Labelle and W. Skalli and J. A. De Guise},

title = {A hierarchical statistical modeling approach for the unsupervised 3-D reconstruction of the scoliotic spine},

booktitle = {in Proc. 10th IEEE Int. Conf. Image Processing},

year = {2003},

pages = {561--564}

}

### OpenURL

### Abstract

In this paper, we propose a new and accurate 3D reconstruction technique for the scoliotic spine from a pair planar and conventional radiographic images (postero-anterior and lateral). The proposed model uses a priori hierarchical global knowledge, both on the geometric structure of the whole spine and of each vertebra. More precisely, it relies on the specification of two 3D templates. The first, a rough geometric template on which rigid admissible deformations are defined, is used to ensure a crude registration of the whole spine. 3D reconstruction is then refined for each vertebra, by a template on which non-linear admissible global deformations are modeled, with statistical modal analysis of the pathological deformations observed on a representative scoliotic vertebra population. This unsupervised coarse-to-fine 3D reconstruction procedure is stated as a double energy function minimization problems efficiently solved with a stochastic optimization algorithm. The proposed method, tested on several pairs of biplanar radiographic images with scoliotic deformities, is comparable in terms of accuracy with the classical CT-scan technique while being unsupervised and requiring a lower amount of radiation for the patient. 1.

### Citations

3367 | A computational approach to edge detection
- Canny
- 1986
(Show Context)
Citation Context ...tential fields of each radiographic image. To compute the edge potential field Ψ associated with each radiographic view, we first use a Canny edge detector with the unsupervised technique proposed in =-=[2]-=-. Then, Ψ is defined as in [4]sby, Ψ(x, y) = � exp − � ξ 2 x + ξ 2 y τ � ���cos � γ(x, y) � � � �, where ξ = (ξx, ξy) is the displacement to the nearest edge point in the image, and τ is a smoothing f... |

2347 |
A method of registration of 3-d shapes
- Besl, McKay
- 1992
(Show Context)
Citation Context ...sition and orientation of the data; two data sets that differ only by a small translation or rotation may have substantially different octree-spline representations. The iterative closest point (ICP) =-=[22]-=- starts with two meshes and an initial estimation for their relative rigid-body transformation, and iteratively refines the transformation by repeatedly generating pairs of the closest points on the m... |

2218 | Marching Cubes: A high resolution 3D surface construction algorithm, Computer Graphics 21:4
- Lorensen, Cline
- 1987
(Show Context)
Citation Context ... these deformations, several 3-D reconstruction methods have been developed. Among these methods, the 3-D reconstruction methods of tomodensitometric imagery modalities (e.g., computerized tomography =-=[2]-=-, [3], or magnetic resonance [4], [5]) provide accurate 3-D information of the human anatomy or 4D information including time as one more degree of freedom. However, few hospitals can afford a magneti... |

1092 |
Active shape models – their training and application
- Cootes, Taylor, et al.
- 1995
(Show Context)
Citation Context .... In [23], Fleute and Lavallée propose a method allowing to reconstruct the 3-D geometry of the femur using a few orthogonal X-ray radiographic projections. They apply point distribution models (PDM) =-=[24]-=- to reconstruct a geometric prior model representing mean shapes and containing the typical deformation modes in statistical sense. Thereafter, they deform the model nonrigidly, according to the rays ... |

755 | Closed-from solution of absolute orientation using unit quaternions
- Horn
- 1987
(Show Context)
Citation Context ...plate, that is a set of points or landmarks1 where are the Cartesian coordinates of each surface point. Given a sample of shapes with the same number of points, we resort to the procedure proposed in =-=[35]-=- to align this training set. We then apply the PCA to reduce the dimension to . is defined in such a way that it describes a certain proportion of possible variations. This gives us nonlinear deformat... |

243 |
Constraints on deformable models: recovering 3d shape and nonrigid motion
- Terzopoulos, Witkin, et al.
- 1988
(Show Context)
Citation Context ...riori knowledge on the geometrical structure of the object to be reconstructed and, thus, expressed as the solution of a regularized inverse problem, have been proposed. Terzopoulos et al. propose in =-=[12]-=- a method allowing to recover the 3-D shape from the 2-D profiles of an object using, as geometric prior constraint, a deformable tube coupled to a deformable spine. The deformation is controlled by p... |

177 |
Object matching using deformable templates
- Jain, Zhong, et al.
- 1996
(Show Context)
Citation Context ...aphic image. To compute the edge potential field Ψ associated with each radiographic view, we first use a Canny edge detector with the unsupervised technique proposed in [2]. Then, Ψ is defined as in =-=[4]-=-sby, Ψ(x, y) = � exp − � ξ 2 x + ξ 2 y τ � ���cos � γ(x, y) � � � �, where ξ = (ξx, ξy) is the displacement to the nearest edge point in the image, and τ is a smoothing factor which controls the degre... |

112 |
Volumetric descriptions of objects from multiple views
- Martin, Aggarwal
- 1983
(Show Context)
Citation Context ...GINEERING, VOL. 52, NO. 12, DECEMBER 2005 literature [6]–[8]. Among the proposed algorithmic techniques, we can cite the 3-D reconstruction method from silhouettes, proposed by Martin and Aggarwal in =-=[9]-=-. Their method allows to reconstruct polygonal 3-D objects by back-projecting the silhouettes. A method for extracting 3-D geometry of bones from two orthogonal X-ray radiographic projections is also ... |

104 |
Genetic Algorithms+Data Structures=Evolution Programs
- Michalewicz
- 1996
(Show Context)
Citation Context ...rithm is also especially well suited to minimize complex (i.e., nonconvex) energy functions [40]. We compare the performance of this algorithm, in Section VII, with a classical genetic algorithm (GA) =-=[41]-=-. A. Exploration/Selection Algorithm The E/S algorithm belongs to the class of evolutionary algorithms. This class of algorithm has been successfully applied in diverse areas such as medical imaging [... |

57 |
Surface-based labeling of cortical anatomy using a deformable atlas
- Sandor, Leahy
- 1997
(Show Context)
Citation Context ...nstruction methods have been developed. Among these methods, the 3-D reconstruction methods of tomodensitometric imagery modalities (e.g., computerized tomography [2], [3], or magnetic resonance [4], =-=[5]-=-) provide accurate 3-D information of the human anatomy or 4D information including time as one more degree of freedom. However, few hospitals can afford a magnetic resonance system, it is time-consum... |

57 | Hierarchical active shape models, using the wavelet transform
- Davatzikos, Tao, et al.
- 2003
(Show Context)
Citation Context ...bra. Hierarchical statistical models have been applied in a fusion of multiresolution image data [26], segmentation [27], [28], restoration [29] and in active shape models using the wavelet transform =-=[30]-=-. Bernard et al. introduce in [28] a hierarchical strategy for segmenting cervical vertebrae. This hierarchical scheme is comprised of two levels; shape and appearance models. Appearance model describ... |

46 | A parametric deformable model to fit unstructured 3D data
- Bardinet, Cohen, et al.
- 1998
(Show Context)
Citation Context ...bject using, as geometric prior constraint, a deformable tube coupled to a deformable spine. The deformation is controlled by physically based internal and external forces. Bardinet et al. present in =-=[13]-=- a method that consists in matching a parametric deformable model to unstructured 3-D data. More precisely, a super-quadric model is fitted to a given point set, and the generated super-quadric model ... |

33 |
Generation of Point-Based 3D Statistical Shape Models for Anatomical Objects
- Lorenz, Krahnstoever
(Show Context)
Citation Context ...e deformations, several 3-D reconstruction methods have been developed. Among these methods, the 3-D reconstruction methods of tomodensitometric imagery modalities (e.g., computerized tomography [2], =-=[3]-=-, or magnetic resonance [4], [5]) provide accurate 3-D information of the human anatomy or 4D information including time as one more degree of freedom. However, few hospitals can afford a magnetic res... |

29 | A hierarchical statistical framework for the segmentation of deformable objects in image sequences
- Kervrann, Heitz
- 1994
(Show Context)
Citation Context ... a prioriknowledge on the geometric structure of both the whole spine and each vertebra. Hierarchical statistical models have been applied in a fusion of multiresolution image data [26], segmentation =-=[27]-=-, [28], restoration [29] and in active shape models using the wavelet transform [30]. Bernard et al. introduce in [28] a hierarchical strategy for segmenting cervical vertebrae. This hierarchical sche... |

21 | Evolutionary Computation in Medicine: An Overview
- Peńa-Reyes, Sipper
- 2000
(Show Context)
Citation Context ...rgy functions [40]. This success has initiated the development of various evolutionary algorithm variants and stimulated the theoretical research about convergence properties of these algorithms (see =-=[43]-=- for a good review of evolutionary algorithm in medical imagery). The E/S algorithm can be summarized as follows (more details are given in [39]). Let be a finite discrete subset of the Cartesian prod... |

20 |
Applied Statistics, Principles and Examples
- Cox, Snell
- 1981
(Show Context)
Citation Context ...ypotheses of the statistical significance of these local deformations relatively to the global deformations (with a level of significance ), the Student’s -test, namely, has to be lower then (i.e., ) =-=[48]-=-. Table VII, thus, shows that local deformations give a significant improvement in the accuracy of reconstruction except in the case of lumbar vertebrae (probably due to the low size of the validation... |

19 |
Model extraction from magnetic resonance volume data using the deformable pyramid
- Lotjonen, Reissman, et al.
- 1999
(Show Context)
Citation Context ... reconstruction methods have been developed. Among these methods, the 3-D reconstruction methods of tomodensitometric imagery modalities (e.g., computerized tomography [2], [3], or magnetic resonance =-=[4]-=-, [5]) provide accurate 3-D information of the human anatomy or 4D information including time as one more degree of freedom. However, few hospitals can afford a magnetic resonance system, it is time-c... |

19 |
Nonrigid 3-D/2-D Registration of images Using Statistical Models
- Fleute, Lavallée
- 1999
(Show Context)
Citation Context ...of each vertebra in a radiographic environment. Methods using statistical a priori knowledge of the geometric shapes of the objects of interest lead to better constrain the reconstruction problem. In =-=[23]-=-, Fleute and Lavallée propose a method allowing to reconstruct the 3-D geometry of the femur using a few orthogonal X-ray radiographic projections. They apply point distribution models (PDM) [24] to r... |

19 | Hierarchical statistical models for the fusion of multiresolution image data - Laferté, Heitz, et al. - 1995 |

19 |
Image denoising based on scale-space mixture modeling of wavelet coefficients
- Moulin, Liu
(Show Context)
Citation Context ...e geometric structure of both the whole spine and each vertebra. Hierarchical statistical models have been applied in a fusion of multiresolution image data [26], segmentation [27], [28], restoration =-=[29]-=- and in active shape models using the wavelet transform [30]. Bernard et al. introduce in [28] a hierarchical strategy for segmenting cervical vertebrae. This hierarchical scheme is comprised of two l... |

15 |
Towards automated image understanding
- Grenander, Keenan
- 1989
(Show Context)
Citation Context ...ution for the scale factor, but this a priori assumption would have required, as a preliminary, estimate the parameters of this Normal law. (1) (2) (3) and Keenan on stochastic pattern representation =-=[36]-=-, a statistical local deformations process is now applied to the control points or “landmarks” which approximate the geometrical shape of each vertebra. These local deformations are modeled as local r... |

13 |
J.C.: Endocardial boundary estimation and tracking in echocardiographic images using deformable templates and markov random fields. Pattern Analysis and Applications 4
- Mignotte, Meunier, et al.
- 2001
(Show Context)
Citation Context ...]. A. Exploration/Selection Algorithm The E/S algorithm belongs to the class of evolutionary algorithms. This class of algorithm has been successfully applied in diverse areas such as medical imaging =-=[42]-=- and localization of shapes [40] to minimize complex energy functions [40]. This success has initiated the development of various evolutionary algorithm variants and stimulated the theoretical researc... |

11 |
3D/2D registration and segmentation of scoliotic vertebrae using statistical models
- Benameur, Mignotte, et al.
(Show Context)
Citation Context ...es in statistical sense. Thereafter, they deform the model nonrigidly, according to the rays back-projected from the contour points of the projections taken from a patient. Benameur et al. propose in =-=[25]-=- a 3-D statistical reconstruction method, for each individual vertebra, using a pair of planar and conventional (postero-anterior with normal incidence and lateral) radiographic images for a patient i... |

8 |
Thoracic Human Vertebrae - Quantitative Three-Dimensional Anatomy Spine Vol 16, No 8,pg888
- Panjabi, Takata, et al.
- 1991
(Show Context)
Citation Context ... ≤ α l ≤ α l−1 + ∆α l , where α is the rotation vector (ensuring the rotation in x, y or z axes). ∆tx, ∆ty, ∆tz, ∆k and ∆α are given by statistical knowledge on the scoliotic deformation of the spine =-=[7]-=-[8]. A global configuration of the deformable spine model is thus described by 9 rigid transformation parameters for each cubic template associated with each vertebral level. x l +1 l l −1 y z (a) (b)... |

8 |
Assessment of 3D reconstruction and high resolution geometrical modeling of the human skeletal trunk from 2D radiographic images
- Delorme, Petit, et al.
- 2003
(Show Context)
Citation Context ... radiographic image) of the vertebrae to produce 3-D coordinates. Adding the kriging technique using the 3-D reconstructed points as control points yields good visualization of the vertebral geometry =-=[18]-=-. There is a need for more landmarks to obtain a better description of all the vertebrae. In the same way, a nonstereo corresponding points technique is also described by Mitton et al. in [19] to impr... |

7 |
Unsupervised localization of shapes using statistical models
- Destrempes, Mignotte
- 2002
(Show Context)
Citation Context ...s not depend on the function to minimize, and convergence is asymptotically ensured [6]. Let us add that this optimization algorithm is also especially well suited to minimize complex energy function =-=[3]-=-. E/S Algorithm E(.) A real-valued l-variable function, defined on F , to be minimized F A finite discrete subset of the Cartesian product Π l j=1[mj, Mj] n The size of the population r A real number ... |

7 |
Computer-aided simulation for bone surgery
- Caponetti, Fanelli
- 1993
(Show Context)
Citation Context ...polygonal 3-D objects by back-projecting the silhouettes. A method for extracting 3-D geometry of bones from two orthogonal X-ray radiographic projections is also proposed by Caponetti and Fanelli in =-=[10]-=-. The initial estimation of the 3-D bone structure, produced by back-projecting profile points, is then refined by using a B-spline interpolation. In [11], Benjamin generates surface information using... |

7 | Global optimization with exploration/selection algorithms and simulated annealing
- François
- 2000
(Show Context)
Citation Context ...tion of the deformation parameters and consequently to the performance of the optimization procedure. In our application, we use the stochastic optimization algorithm recently proposed by François in =-=[39]-=-. We use this algorithm because the adjustment of all internal parameters does not depend on the function to be minimized. Moreover, the convergence is asymptotically ensured [39]. Let us add that thi... |

6 |
Rational design for closerange photogrammetry. University of Illinois at Urbana-Champaign. Doctoral dissertation
- Marzan
- 1975
(Show Context)
Citation Context ...ple graphic interface. User interaction is limited to simply placing this point on the two radiographic images. The 3D coordinates of this landmark were obtained by direct linear transformation (DLT) =-=[5]-=-. This crude registration estimates the rough position t = (tx, ty, tz), scale k and orientation α of each vertebra. • This optimization problem is then used with the template presented in Subsection ... |

6 |
Measurements of the threedimensional shape of the rib cage
- Dansereau, Stokes
- 1988
(Show Context)
Citation Context ...upervising the 3-D reconstruction technique. In this sense, a 3-D multiplanar radiography reconstruction method using a direct linear transformation (DLT) [16] is presented by Dansereau and Stokes in =-=[17]-=-. They rely on manual stereo-digitization of six anatomical landmarks (on each X-ray radiographic image) of the vertebrae to produce 3-D coordinates. Adding the kriging technique using the 3-D reconst... |

5 | 3d biplanar reconstruction of scoliotic vertebrae using statistical models
- Benameur, Mignotte, et al.
- 2001
(Show Context)
Citation Context ...n a lying position, which is incompatible with many diagnostic protocols evaluating scoliosis. To circumvent this difficulty, an original 3D biplanar statistical reconstruction method was proposed in =-=[1]-=- which uses only a pair of planar and conventional (postero-anterior (IP A) and lateral (ILAT )) radiographic images (and consequently ensures a lo∗ École de technologie supérieure, Montréal, Canada E... |

5 |
3D reconstruction method from biplanar radiography using nonstereocorresponding points and elastic deformable meshes
- Mitton, Landry, et al.
- 2000
(Show Context)
Citation Context ...eometry [18]. There is a need for more landmarks to obtain a better description of all the vertebrae. In the same way, a nonstereo corresponding points technique is also described by Mitton et al. in =-=[19]-=- to improve the accuracy of the 3-D reconstruction by identifying more anatomical landmarks on the X-ray projections. This method is then based on deformation of an elastic deformable mesh that respec... |

5 |
Matching of medical images for computed and robot assisted surgery
- Lavallee, Brunie, et al.
- 1991
(Show Context)
Citation Context ...n contained in the two X-ray radiographic projections (e.g., the contours of each vertebra) or the statistical knowledge of the possible deformation of the object to be reconstructed. Lavallée et al. =-=[21]-=- describe an algorithm which minimizes the 3-D distances between the rays (corresponding to the points on the contour) and the closest point on the surface of the object. A 3-D distance map is precomp... |

4 |
Three-dimensional (3-D) reconstruction of the spine from a single X-ray image and prior vertebra models
- Novosad, Cheriet, et al.
(Show Context)
Citation Context ...y studied in the statistical framework, and few references exist in the computer vision and medical imagerys2042 IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 52, NO. 12, DECEMBER 2005 literature =-=[6]-=-–[8]. Among the proposed algorithmic techniques, we can cite the 3-D reconstruction method from silhouettes, proposed by Martin and Aggarwal in [9]. Their method allows to reconstruct polygonal 3-D ob... |

4 | Three-dimensional reconstruction of the left ventricle from two angiographic views - Medina, Garreau, et al. - 1997 |

4 |
3D reconstruction of the femoral bone using two X-ray images from orthogonal views
- Nikkhade-Dehkori, Bro-Nielsen, et al.
- 1996
(Show Context)
Citation Context ...rmable model to unstructured 3-D data. More precisely, a super-quadric model is fitted to a given point set, and the generated super-quadric model is deformed locally using free-form deformations. In =-=[14]-=-, Nikkhade et al. present 3-D reconstruction method of femurs from two orthogonal X-ray radiographic projections. They separate the femur into three subparts; each of them assumed to be round. They fi... |

4 |
Elastic-model driven analysis of several views of a deformable cylindrical object
- Kita
- 1996
(Show Context)
Citation Context .... They separate the femur into three subparts; each of them assumed to be round. They fit cubic parametric surface patches to the subparts and then assemble them to a complete model. Kita develops in =-=[15]-=- a method allowing to analyze X-ray radiographic projections of the stomach using a deformable 3-D model. Their a priori model is a tube which is first initialized using only one projection. Afterward... |

3 |
Global optimization with exploration/selection algorithms and simulated annealing", The Annals of Applied Probability
- Olivier
- 2002
(Show Context)
Citation Context ...ruction step. In our application, due to the complexity and number of variables involved in the energy function expressed by Eq. (4), we use the stochastic optimization algorithm recently proposed in =-=[6]-=- (see Algo. 1.). We use this algorithm because the adjustment of all internal parameters does not depend on the function to minimize, and convergence is asymptotically ensured [6]. Let us add that thi... |

3 |
De Guise. “Morphometric analysis of anatomic scoliotic specimens
- Parent, Labelle, et al.
- 2002
(Show Context)
Citation Context ...replace θi by ˆ θ 3. k ←k + 1 and p=k −1/D until a stopping criterion is met; Algorithm 1: E/S optimization algorithm 5. EXPERIMENTAL RESULTS In our application, we use the training base described in =-=[9]-=-. The first m eigenmodes {(λ1, φ1), . . . , (λm, φm)} are chosen to cover at least 90% of the population’s variability. For the experiments, we have chosen β = 0.02 for the weighting factor penalizing... |

3 |
Three-dimensional (3D) detailed reconstruction of human vertebrae from low-dose digital stereoradiography. The European Journal of Orthopaedic Surgery & Traumatology
- Bras, Laporte
- 2003
(Show Context)
Citation Context ...udied in the statistical framework, and few references exist in the computer vision and medical imagerys2042 IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 52, NO. 12, DECEMBER 2005 literature [6]–=-=[8]-=-. Among the proposed algorithmic techniques, we can cite the 3-D reconstruction method from silhouettes, proposed by Martin and Aggarwal in [9]. Their method allows to reconstruct polygonal 3-D object... |

3 |
Incorporation of patient displacement into a trunk reconstruction technique
- Marcil, Dansereau, et al.
- 1995
(Show Context)
Citation Context ...reau et al., Mitton et al. and Delorme et al. are limited due to the inherent inaccuracy produced in identifying anatomical landmarks (leading to reconstruction errors). Marcil et al. have studied in =-=[20]-=- an effect of the displacement of the patient between the X-ray exposures in the 3-D stereoradiographic of scoliotic spines and rib cages. Besides, they do not exploit all the information contained in... |

3 |
3D biomedical modeling: merging image processing and computer aided design
- JA, Martel
- 1988
(Show Context)
Citation Context ...h a CT-scan device. Each CT slice of 1 mm thickness was taken with a resolution of 512 512 pixels and 12 bits per pixel. The CT scan reconstruction was made by piling the CT slices of the vertebrae 5 =-=[44]-=- using SliceOmatic software. This software allows one to automatically segment the CT-scan slices and then to correct the automatic segmentation manually to distinguish the different objects in the in... |

2 |
Imaging in scoliosis: what, why and how
- Cassar-Pullicino, Eisenstein
- 2002
(Show Context)
Citation Context ...rmations. The classical evaluation obtained by the spinal projections on the two-dimensional (2-D) radiographic planar images does not give a full and accurate interpretation of scoliotic deformities =-=[1]-=-. To analyze the 3-D characteristics of these deformations, several 3-D reconstruction methods have been developed. Among these methods, the 3-D reconstruction methods of tomodensitometric imagery mod... |

2 |
Object-based 3-D X-ray imaging
- Benjamin
- 1995
(Show Context)
Citation Context ... is also proposed by Caponetti and Fanelli in [10]. The initial estimation of the 3-D bone structure, produced by back-projecting profile points, is then refined by using a B-spline interpolation. In =-=[11]-=-, Benjamin generates surface information using common points, present in different projections, and the tracks between the points, extracted from about ten projection images. Due to the ill-posed natu... |

2 |
Segmenting articulated structures by hierarchical statistical modeling of shape, appearance, and topology
- Bernard, Lika, et al.
- 2001
(Show Context)
Citation Context ...oriknowledge on the geometric structure of both the whole spine and each vertebra. Hierarchical statistical models have been applied in a fusion of multiresolution image data [26], segmentation [27], =-=[28]-=-, restoration [29] and in active shape models using the wavelet transform [30]. Bernard et al. introduce in [28] a hierarchical strategy for segmenting cervical vertebrae. This hierarchical scheme is ... |

2 |
Morphometric analysis of anatomic scoliotic specimens
- Parent, Labelle, et al.
- 2002
(Show Context)
Citation Context ...hese data were obtained by digitization of the anatomical points on anatomical specimens selected in the Hamann Todd osteology collection, Cleveland, OH, and Smithsonian Institution in Washington, DC =-=[46]-=- 8 Fastrack (POLHEMUS, A Rockwell Collins Company) is the name of the electromagnetic device used to digitize each vertebra by means of a pointer. The accuracy of this device is evaluated at 0.2 mm [4... |

1 |
Quantitative three-dimensional anatomy of lumbar spine
- Semaan, Skalli, et al.
- 2001
(Show Context)
Citation Context ...ndmarks on each vertebra specimen using a pointer of the electromagnetic device, thus creating a set of approximately 200 points depending on the level measured with regard to its particular geometry =-=[34]-=- (see Section VII-A). Different points were acquired in a specific order and recorded in this sequence. All vertebrae of same vertebral level have the same number of points, and the 3-D coordinates of... |

1 |
Morphometric evaluations of personalized 3D reconstructions and geometric models of the human spine,” Med
- Aubin, Dansereau, et al.
- 1997
(Show Context)
Citation Context ...ically segment the CT-scan slices and then to correct the automatic segmentation manually to distinguish the different objects in the initial image. The accuracy of this technique is evaluated at 1mm =-=[45]-=- and it allows a CT scan reconstruction of the vertebrae containing up to 7000 points. Twelve landmarks (extremities of pedicles, extremities of the diameter of the spinal canal) were identified on ea... |

1 |
Three-dimensional measurement of a scoliotic spine,” Three-Dimensional Analysis of Spinal Deformities (IOS
- Borgne, Skalli, et al.
- 1995
(Show Context)
Citation Context ...6] 8 Fastrack (POLHEMUS, A Rockwell Collins Company) is the name of the electromagnetic device used to digitize each vertebra by means of a pointer. The accuracy of this device is evaluated at 0.2 mm =-=[47]-=-. The 3-D coordinates of the pointer were recorded in a specific reference system. Let us recall that the digitizing protocol consisted of measuring specific anatomical landmarks on each vertebra, thu... |

1 | 4 – 3D reconstruction of vertebra segment of scoliotic spine. (a) Postero-anterior, (b) Lateral image (512x256 pixels), (c) and (d) Visualization of reconstruction vertebra segment (L1/L2/L3/L4/L5) from the sagittal and coronal views. 6. CONCLUSION We hav - FIG |