SADM 2012 Abstracts


Full Papers
Paper Nr: 2
Title:

SEGMENTATION OF VESSEL GEOMETRIES FROM MEDICAL IMAGES USING GPF DEFORMABLE MODEL

Authors:

Si Yong Yeo, Xianghua Xie, Igor Sazanov and Perumal Nithiarasu

Abstract: We present a method for the reconstruction of vascular geometries from medical images. Image denoising is performed using vessel enhancing diffusion, which can smooth out image noise and enhance vessel structures. The Canny edge detection technique which produces object edges with single pixel width is used for accurate detection of the lumen boundaries. The image gradients are then used to compute the geometric potential field which gives a global representation of the geometric configuration. The deformable model uses a regional constraint to suppress calcified regions for accurate segmentation of the vessel geometries. The proposed framework show high accuracy when applied to the segmentation of the carotid arteries from CT images.

Paper Nr: 4
Title:

ROBUST DEFORMABLE MODEL FOR SEGMENTING THE LEFT VENTRICLE IN 3D VOLUMES OF ULTRASOUND DATA

Authors:

Carlos Santiago, Jorge S. Marques and Jacinto Nascimento

Abstract: The segmentation of the left ventricle (LV) in echocardiographic data has proven itself a useful methodology to assess cardiac function and to detect abnormalities. Traditionally, cardiologists segment the LV border at the end-systolic and end-diastolic phases to determine the ejection fraction. However, the manual segmentation of the LV is a tedious and time demanding task, which means automated segmentation systems can provide a powerful tool to improve workflow in a clinical setup. This paper proposes a robust 3D segmentation system consisting of a deformable model that uses a probabilistic data association filter (PDAF) to robustly detect the LV border. Results show that the algorithm performs well in both synthetic and real data, without significantly compromising its performance. The obtained LV segmentations are compared with the manual segmentations performed by an expert, yielding an average distance of 4 pixel between points from both segmentations.

Paper Nr: 5
Title:

ALGORITHM TO MAINTAIN LINEAR ELEMENT IN 3D LEVEL SET TOPOLOGY OPTIMIZATION

Authors:

Christopher J. Brampton, Alicia H. Kim and James L. Cunningham

Abstract: In level set topology optimization the boundary of the structure is defined by level set function values stored at the nodes of a regular gird of simple bilinear elements. By changing the level set function values according to optimization sensitivities the boundary of the structure is moved to create an optimal structure. However it is possible for the boundary to cut an element more than once; violating the linear element assumptions resulting in insufficient nodal information for the optimization sensitivity calculations. To resolve this the local boundary of the structure is moved so that each element is only cut once. In 2D where a square element mesh is used an element cut twice times is altered by moving one of the boundaries within the element to intercept the node closest to it removing the extra cut from the element. In 3D where a voxel mesh is used the process of moving the boundary within an element is more complicated due to the greater number of boundary cuts possible and the effect that it can have on neighbouring elements. An algorithm is developed which allows the boundary within a 3D element to be moved with these considerations taken into account.

Paper Nr: 7
Title:

FACIAL EXPRESSION RECOGNITION USING LOG-EUCLIDEAN STATISTICAL SHAPE MODELS

Authors:

Bartlomiej W. Papiez, Bogdan J. Matuszewski, Lik-Kwan Shark and Wei Quan

Abstract: This paper presents a new method for facial expression modelling and recognition based on diffeomorphic image registration parameterised via stationary velocity fields in Log-Euclidean framework. The validation and comparison are done using different statistical shape models (SSM) built using the Point Distribution Model (PDM), velocity fields, and deformation fields. The obtained results show that the facial expression representation based on stationary velocity field can be successfully utilised in facial expression recognition, and this parameterisation produces higher recognition rate than the facial expression representation based on deformation fields.

Short Papers
Paper Nr: 6
Title:

A NEW PARALLELIZABLE DEFORMATION METHOD - Automatic Comparision between Foot and Last

Authors:

J. M. Buades, M. Gonzalez-Hidalgo, Francisco J. Perales, S. Ramis-Guarinos, A. Oliver and V. Blanch

Abstract: This document shows the research project developed by the Computer Graphics and Vision Group in collaboration with Inescop. The main objective of this project is to define a new methodology to design shoes exactly adapted to the shape foot. The system proposed is based on a biomechanical anatomical structure of the foot and using a deformable shape. Using with several joints adapted the different foots to different shoe last evaluating the error between the surfaces. The system is automatic driven in the selection of significant foot points. We consider several anthropometrical parts of the foot in order to apply the deformations of the shape with different axis. The proposed system is implemented in software. An associated Application Programming Interface (API) is also developed for commercial use of the system. Also the results presented evaluate the error between deformations.

Paper Nr: 8
Title:

A STATISTICAL APPROACH TO BUILD 3D PROTOTYPES FROM A 3D ANTHROPOMETRIC SURVEY OF THE SPANISH FEMALE POPULATION

Authors:

M. V. Ibañez, A. Simó, J. Domingo, E. Durá, G. Ayala, S. Alemany, G. Vinué and C. Solves

Abstract: Fitting cloth is a problem for both the customer and the apparel industry, but analysis of anthropometric data can be useful to define better sizing systems. In 2006, the Spanish Ministry of Health coordinated a study to obtain 3D anthropometric data of the Spanish women. Our aim in this work is to develop a statistical methodology to define prototypes based on the 3D clouds of points obtained from 3D scans of a great number of women and apply it to the 3D anthropometric survey of the Spanish female population. To build the prototypes, 3D images will be built, and after registration, homologous 2D sections will be averaged, and a 3D ”mean” shape will be reconstructed from them.