Abstract: In this study, we describe the preliminary results of a tool to assist the diagnosis for the characterization of pathological cases of epilepsy disease using cerebral tomoscintigraphy images. The tool is based on the use of multifractal modelling to detect the local changes of homogeneity.
Abstract: The radiotherapy planning procedure is achieved using images obtained from computed tomography (CT) or magnetic resonance imaging (MR). These images are taken before the treatment which is performed in several sessions over several weeks. At the beginning of each session, the patient has to be positioned on the treatment couch under the linear accelerator in the same position as during MR or CT imaging and planning, and the organs are assumed to be in the same place. Currently, the methods used for this repositioning are based on the external anatomy of the patient and assume that the internal structures do not move.
In this study, we present a new approach, suited to clinical practice, for the automatic repositioning of patients in prostate cancer radiotherapy. It is based on localisation by ultrasound images and optical stereolocalisation and on a matching with some images regenerated in the planning volume. The method exploits a statistical model of the prostate to automatically extract its contours.
The first tests in conditions of a radiotherapy session show that the method is able to obtain a patient setup with an accuracy of about 1.4 mm.
Abstract: The purpose of our study is to propose a practical and easy to use method for the calibration of a freehand ultrasound machine. We introduce a new spatial calibration of the probe based on a simple phantom. The method includes automatic features extraction by the Hough transform algorithm and a robust parameters optimization. Experiments demonstrate the reproducibility of the method. Many evaluation were conducted to evaluate the accuracy of the 3D measures realized from the image, we found errors less than 1.4 mm. The technique we describe is pragmatic way for a rapid and accurate calibration of a freehand ultrasound system. Technique that could be used for the intra-operative visualization procedures
Abstract: Purpose/Objective. Target volumes and organ at risk delineation is a time consuming task in radiotherapy planning. The development of automated segmentation tools still remains a difficult problem due to pelvic organs shape variability. In this paper 3D deformable model approach and seeded region growing algorithm for prostate and organ at risk automatic delineation on MR images are evaluated.
Methods and materials. Manual and automatic delineation were compared in 24 patients using a sagittal T2-w TSE and an axial T1-w 3D FFE or TSE sequences. For prostate automatic delineation, an organ model-based method was used. Prostate without seminal vesicles was delineated as CTV. For bladder and rectum automatic delineation, a seeded region growing method was used. Manual contouring was considered as the reference. The following parameters were measured: volume ratio (Vr) (automatic/manual), volume overlap (Vo) (ratio of the volume of intersection to the volume of union, optimal value=1), correctly delineated volume (Vc) (percent ratio of the volume of intersection to the manual defined volume, optimal value=100).
Results. For CTV, the Vr, Vo and Vc were 1.13 (±0.1), 0.78 (±0.05) and 94.75 (±3.3) respectively. For rectum, the Vr, Vo and Vc were 0.97 (±0.1), 0.78 (±0.06) and 86.52 (±5) respectively. Vr, Vo and Vc were 0.95 (±0.03), 0.88 (±0.03) and 91.29 (±3.1) for bladder respectively.
Conclusion. Our results show that organ model method is robustness and leads to reproducible prostate segmentation with minor interactive corrections. For bladder and rectum automatic delineation MRI soft tissue contrast allows to use region growing methods.
Abstract: This article discusses a method for the automatic segmentation of trans-abdominal ultrasound images of the prostate. Segmentation begins with the application of a filter to enhance the contours without modifying the image information. It combines adaptive morphological filtering and median filtering to detect the noise-containing regions and smooth them. A heuristic optimization algorithm searches for the contour initialized from a prostate model. The performance of the algorithm was tested by comparing the resulting contours with those obtained by manual segmentation. The average distance between the contours was 2.5 mm and the average coverage index was 93%.
Abstract: This paper describes a localization system composed of a stereovision system with two cameras and an ultrasound system. After calibration of 2 cameras, a mask of infra-red LEDs is mounted on the probe. A method for spatial and temporal calibration is presented and used to calibrate the ultrasound system. Position and orientation of each ultrasonic cross-section are precisely measured and a 3D localization is available from the US images with an accuracy of 0.4 mm. We present two clinical applications of the system which gave good results in each case.