Workshop MIAD 2009 Abstracts

Full Papers
Full Papers
Paper Nr: 3
Title:

BIOLOGICAL QUALIFICATION OF OOCYTE MATURITY WITH THE USE OF THE KARHUNEN – LOEVE TRANSFORM Computer-aided decision for selecting best oocytes before fertilization
Authors:

Charline André, Christian Pieralli, Bruno Wacogne, Christiane Joanne, Lionel Pazart, Christophe Roux and Charline André
Abstract: The estimation of the a priori maturity degree of oocytes before fertilizing is a critical step in In Vitro Fertilization (IVF). This kind of analysis is currently done by the practitioner with a visual microscope inspection and is therefore subjective. In this paper, we propose the use of an image processing called Karhunen-Loeve Transform (KLT). The KLT exploits the covariance matrix and its eigenvectors for obtaining a representation of information in several images. We show that the KLT permits the determination of the oocyte maturity by examining the KL images and more precisely the eigenvalues of each. The KLT could thus be a useful tool for IVF and we intend to develop a dedicated server for all French IVF centers as a computer aided decision for fertilization.
Download
Paper Nr: 5
Title:

Automated Segmentation and Clinical Information on Dementia Diagnosis
Authors:

Aura Conci, Alexandre Plastino, Andrea Souza, Debora Saade and Flavio Seixas
Abstract: This work intends to predict the clinical dementia rating (CDR) based on human brain volumetric segmentation measures from magnetic resonance (MR) images. These brain measures were extracted using an automated image segmentation method based on morphometry study and considering brain ana-tomical atlas. The prediction was achieved by Bayesian classifier. The classifier training was performed on 371 individuals from Open Access Series of Imaging Studies (OASIS) dataset. MR images and clinical information (including the Clinical Dementia Rating score) of each case are available on OASIS dataset. Experimentation results were assessed using true-positive rate. The final pur-pose of this work is to design a computer-aided diagnostic system that could be able to detect precociously neurodegenerative disorders, allowing early thera-peutic interventions.
Download
Paper Nr: 11
Title:

3D Segmentation for the study of Cell Cycle Progression in Live Drosophila Embryos
Authors:

Rambabu Chinta, Wee Choo Puah, Janos Kiston-Vizi and Martin Wasser
Abstract: We study the dynamics of cell division in live Drosophila embryos using fluorescent proteins and 3D time-lapse microscopy. Accurate segmentation of nuclei and mitotic chromosomes labeled by the live reporter histone-GFP is a prerequisite for subsequent tracking and quantitative object analysis. We propose an automated 3D segmentation method based on narrow band level sets that preserves the boundary of the cell nuclei and removes signals that are artifacts of live cell imaging. We introduce an improved 3D narrow band approach in the region shrinking and growing process for accurately segmenting the cell nuclei from background. The proposed method has been evaluated with the ground truth regarding the object level accuracy and segmentation quality. Both the object level accuracy and pixel accuracy of the proposed method are around 96% and 85% respectively. Our algorithm can robustly segment nuclei and chromosomes in different phase of the division cycle.
Download
Paper Nr: 12
Title:

Mammographic Density Classification Based on Local Histograms
Authors:

Rafael Llobet, Juan A. Solves, Francisco Ruiz-Perales and Juan C. Perez-Cortes
Abstract: In this work, the task of classifying mammograms according to breast density is studied using a local-histogram-based feature extraction method and a non-parametric classification scheme. Breast density estimation is important due to its association with a higher risk of cancer and an increased difficulty of diagnosis. 322 images from the Mammographic Image Analysis Society (MIAS) Database have been analyzed, and the density prediction accuracy of the method has been assessed. The obtained results show an agreement of 77.96% between automatic and expert radiologist manual classification.
Download
Paper Nr: 13
Title:

A Cost Efficient Approach for Automatic Non-Rigid Registration of Medical Images
Authors:

Walid Barhoumi and Sami Dhahbi
Abstract: A common approach for non-rigid medical image registration is the hierarchical image subdivision-based strategy. In this approach, images are progressively subdivided, locally registered, and elastically interpolated. Although this approach seems to be among the fastest approaches for non-rigid registration, computation time is still a real challenge. This work deals with this problem and proposes a new hierarchical strategy. To reduce computational complexity, we propose to combine in the same framework the hierarchical image subdivision-based strategy with a Gaussian pyramid. The hierarchical subdivision method ensures that the registration process deals with small and large deformations, whereas the use of Gaussian pyramid decreases the computation time enormously. The proposed framework is preliminary validated in the context of monomodal registration by matching breast mammograms andMRI brain images with simulated deformations. Registration quality is evaluated by using image differences, mean square error, peak signal to noise ratio and correlation coefficient. Complexity study and experimental results show that the proposed approach reduces considerably the computation cost meanwhile maintaining comparable accuracy.
Download
Paper Nr: 14
Title:

Performance Assessment of Patch-Based Bilateral Denoising
Authors:

Arnaud de Decker, M. Verleysen and John Lee
Abstract: In the field of medical image analysis, denoising is one of the most important preprocessing steps before medical analysis. The design of an efficient, robust, and computationally effective edge-preserving denoising algorithm is a widely studied, and yet unsolved problem. One of the most efficient edge-preserving denoising algorithms is the bilateral filter, which is an intuitive generalization of the local M-smoother. In this paper, we propose to modify both the bilateral filter and the local M-smoother to use patches of the image instead of single voxels in the denoising process. Using patches instead of single voxels in the filtering process is a way to adapt the filter to the textures, ramps, and edges of the image, and make the filter more discriminant. The filtering performances of the patch-based algorithms are evaluated on a benchmark and a CT phantom image and compared to the bilateral filter and local M-smoother.
Download
Paper Nr: 15
Title:

SEMI-AUTOMATIC MODELING OF BONES FOR REAL-TIME SURGERY SUPPORT
Authors:

Roger Cuypers and Benjamin Weyers
Abstract: A model-based reconstruction of human bones in the context of surgery support is an upcoming field of research in informatics and kinematics. Tools for planning surgeries and real-time support require appropri-ate mathematical models for rendering, interacting as well as for reconfiguration. Our conviction is that Su-perquadrics offer this powerful mathematical modeling capability. Image based data which result from MRI and X-ray examinations have to be extracted and gathered to 3D-point sets which are afterwards fitted by superquadrics-based models. The fitting process is complex and time consuming. To solve this problem and to provide real-time simulation for the field of surgery support, the knowledge of the expert user should be applied. This paper presents the concept and a prototypical implementation of an interactive system which involves the user into the fitting process to accelerate the calculation and to enhance the resulting model.
Download
Paper Nr: 18
Title:

SEMI-AUTOMATIC DETECTION OF CORTICALIS BORDERS IN TWO-DIMENSIONAL RADIOGRAPHIES TO IMPROVE PRE-OPERATIVE PLANNING
Authors:

Marc Schlimbach and Jürgen Wahrburg
Abstract: This paper suggests an algorithm to detect corticalis borders in a radiograph. It is robust against occurring interferences in radiographies, because it is a global working procedure. This segmentation is part of semiautomatic assistant functions, which are developed to support surgical interventions. The algorithm will be introduced related to the calculation of the femoral CCD-Angle. Further subjects of the presented proposal cover special aspects when using a two-dimensional radiography for pre-operative planning instead of CT or MRT scans.
Download
Paper Nr: 20
Title:

COMPUTER AIDED EVALUATION OF UPPER URINARY TRACT OBSTRUCTION
Authors:

Victor Neeman, V. Neeman and I. Leichter
Abstract: The purpose of this study is to examine a method for quantitative estimation of upper urinary tract clearance rate using conventional fluoroscopic images. To obtain quantitative information proportional to the amount of contrast media in the renal pelvis we used videodensitometric methods. The semi-quantitative densitometry included normalization procedure, logarithmic processing of the system response and non-specific density variations removing. The method was tested by analyzing 7 nephrostogram and 3 retrograde pyelography studies. The clearance rate was estimated by measuring the clearances curve in arbitrary units. Regression fitting of the clearances curve by an exponential decay yielded a correlation coefficient of 0.94±0.02. The integrated radio-density of the contrast media was found to decrease by 6±3% per minute, and the area of the contrast agent in the renal pelvis decreased by 5±2% per minute. The radio-density measurements during the first 10 minutes of the examination were sufficient to yield the overall exponential clearances curve. It was concluded that this method will enable to estimate quantitatively the degree of upper urinary tract obstruction by using only the initial phase of a routine urological modality.
Download
Paper Nr: 22
Title:

Empirical descriptors evaluation for mass malignity recognition
Authors:

Imene Cheikhrouhou, Nabil Derbel, Hichem Maaref, Khalifa Djemal, Dorra Sellami Masmoudi and Imene Cheikhrouhou
Abstract: In breast cancer field, radiologist and researchers aim to discriminate between masses due to benign breast diseases and tumors due to breast cancer. In general, benign masses have circumscribed contours, whereas, malignant tumors appear with spiculated and irregular boundaries. Recently, we proposed an original mass description based on three morphological mass descriptors, which are SPICULation (SPICUL), Contour Derivative Variation (CDV) and Skeleton End Points (SEP). In this paper, we detail an empirical mass evaluation based on these morphological descriptors which intend to distinguish between malignant and benign lesions. This evaluation is, first, assured by following descriptors evolution in two independent data sets: Alberta and MIAS. Secondly, for these two data sets, the Receiver Operating Characteristics (ROC) analysis is applied. A comparison between the classic use of Area and Perimeter descriptors only, and a combination with our three original evaluated descriptors is done. Obtained results prove that classification accuracy of the combination descriptors including: SPICUL, SEP and CDV outperforms that of the classic descriptors. Indeed, our original mass description provide the best Area under ROC Az = 0:986. Therefore, we affirm that our three original descriptors can serve as good shape descriptors for the benign-versus-malignant classification of breast masses in mammograms.
Download