Search

International audience; Purpose: To automatically learn directional relative positions (DRP) between mediastinal lymph node stations and anatomical organs. Those spatial relationships are used to semiautomatically segment the stations in thoracic CT images. Methods: Fuzzy maps of DRP were automatically extracted by a learning procedure from a database composed of images with stations and anatomical structures manually segmented by consensus between experts. Spatial relationships common to all patients were retained. The segmentation of a new image used an initial rough delineation of anatomical organs and applied the DRP operators. The algorithm was tested with a leave-one-out approach on a database of 5 patients with 10 lymph stations and 30 anatomical structures each. Results were compared to expert delineations with dice similarity coefficient (DSC) and bidirectional local distance (BLD). Results: The overall mean DSC was 66% and the mean BLD was 1.7 mm. Best matches were obtained from stations S3P or S4R while lower matches were obtained for stations 1R and 1L. On average, more than 30 spatial relationships were automatically extracted for each station. Conclusions: This feasibility study suggests that mediastinal lymph node stations could be satisfactory segmented from thoracic CT using automatically extracted positional relationships with anatomical organs. This approach requires the anatomical structures to be initially roughly delineated. A similar approach could be applied to other sites where spatial relationships exists between anatomical structures. The complete database of the five reference cases is made publicly available.

International audience

Purpose. - This work evaluated the interobserver variability in cone beam computed tomography (CBCT) registration for prostate cancers treated with intensity-modulated radiotherapy. Material and methods. - Twelve technologists realized 286 CBCT/CT registrations (bone registration followed by prostate to prostate registration). The registration results were compared to those obtained by two radiation oncologists (reference). Each technologist reported the shifts calculated by the software in all three axes. A statistical analysis allowed us to calculate the minimum threshold under which 95% of the observers found similar values. A variance analysis followed by the post hoc test were used to find differences in interobserver registration variability and determine whether any individual users performed registrations which differed significantly from those of the other users. Results. - The registration differences compared to the reference in the three directions in terms of 95th percentile are: 2.1 mm left-right, 3.5 mm target-gun, 7.3 mm anterior-posterior. In the posterior direction, 4% of the observers have found differences superior to 8 mm, margin used in routine without the use of a daily CBCT. The variance test revealed a P-value

International audience; Purpose/Objective: Since September 2012, 14 patients with locally advanced NSCLC were included in a phase 2 randomized clinical trial that compares ITV and mid-position conformational planning strategies (50 patients planned). We provide here a preliminary dosimetric analysis in terms of target volumes variation and normal tissue exposure.

International audience; Deformable image registration is of considerable interest to the radiation therapy community. It is considered a key enabling technology for applications such as adaptive radiotherapy, contour propagation in 4D CT images, atlas-based segmentation, breathing motion modeling, dose accumulation and many more. Deformable registration is slowly finding its way to commercial software packages. However, validation of the estimated deformation fields remains challenging because no ground truth is available. In general, validation of deformable registration relies on manually generated data, often in the form of hand-clicked landmarks or handdrawn contours. Obtaining such data is a laborious and time-consuming task. Strictly speaking, the resulting validation should be considered as partial, due to the high number of degrees of freedom of the system. In addition, the ill-posed nature of the problem implies the procedure needs to be repeated when dealing with images of different quality, nature or modality. Open-source and open-data concepts constitute an approach where source code and data are provided for free usage, within the limits of a defined agreement, and with a strong concern on users' feedback and participation. It has proved to be an efficient strategy to accelerate scientific advances in a difficult field, based on distributed and collaborative effort, in particular by enabling evaluation and comparison. Many initiatives exist for the application of validating deformable registration, several of which have had a considerable impact in the field. In this presentation, we will describe some open-source and open-data tools that have been proposed and used for validation of deformable image registration. We will review some freely available image processing toolkits (e.g. ITK, elastix), publicly accessible databases of annotated images (e.g. Dir-Lab, POPI), visualization freeware (e.g. VTK, Osirix, VV) and implementation contests or challenges (EMPIRE10). Advantages and limitations of each will also be discussed.