Project

Back to overview

Adaptive atlas-based segmentation of head and neck lymph nodes in CT images for accurate radiotherapy planning

Applicant Thiran Jean-Philippe
Number 124797
Funding scheme Project funding (Div. I-III)
Research institution Laboratoire de traitement des signaux 5 EPFL - STI - IEL - LTS5
Institution of higher education EPF Lausanne - EPFL
Main discipline Information Technology
Start/End 01.04.2009 - 31.03.2012
Approved amount 155'631.00
Show all

Keywords (10)

atlas-based segmentation; computer-assisted image analysis; radiotherapy; treatment planning; Non-Rigid Registration; IMRT; Lymph Node; Regions; Head and Neck; Image Processing

Lay Summary (English)

Lead
Lay summary
Intensity modulated radiotherapy (IMRT) is the high precision technique to accurately deliver X-ray radiation treatment for different tumor locations. However, its implementation in routine clinical practices is facing significant obstacles, particularly in the treatment of lymph nodes of the Head and Neck (H&N) region. The major bottleneck is the inability to accurately and automatically delineate the lymph node regions. Lymph node regions are constructed volumes in the H&N region and they often do not have any visibly distinct boundaries in the images; rather, they are defined with respect to other visible landmark structures in the image and hence, posing challenges in their automated segmentation.Although our current segmentation scheme, developed in a previous SNF project, is more accurate than other methods, it can be noticed from quantitative evaluations, that it is still far from an ideal segmentation. The objective of this proposal is to further improve the robustness as well as accuracy of the H&N lymph node regions segmentation, and thereby minimize, if not able to completely eliminate, the corrections to be made by the Physician to the automated segmentation results.The main ideas that we propose here for improving the lymph node regions segmentation are as follows:o Development of a self-adapting framework that learns from the error-correction process and performs a more accurate segmentation when used next time.o Selection of patient specific subpopulation of atlases that best suit the patient's image to be segmented and construction of an optimal statistical atlas from the selected subpopulation.o Identification and labeling more structures in the atlas image that are close to the lymph nodes and have distinct boundaries. This will result in more accurate deformation field computation and thus more accurate lymph nodes segmentation.The proposed research ultimately envisages developing a clinically usable software system for the automated segmentation of target volumes for IMRT of the H&N region. This would be of great value for the radiation-oncology field in general and will represent a starting point of the widespread use of the IMRT technique.
Direct link to Lay Summary Last update: 21.02.2013

Responsible applicant and co-applicants

Employees

Publications

Publication
Brain tissue segmentation on diffusion weighted magnetic resonance data
Esteban-Sanz Oscar, Gorthi Subrahmanyam, Wollny Gert, Daducci Alessandro, Ledesma-Carbayo María Jesús, Santos Andrés, Thiran Jean-Philippe, Bach Cuadra Meritxell (2012), Brain tissue segmentation on diffusion weighted magnetic resonance data, in IEEE International Symposium on Biomedical Imaging (ISBI) , Barcelona, Spain, 2-5 May 2012.
Evaluation of Atlas Fusion Strategies for Segmentation of Head and Neck Lymph Nodes for Radiotherapy Planning
Gorthi Subrahmanyam, Bach Cuadra Meritxell, Schick Ulrike, Tercier Pierre-Alain, Allal Abdelkarim S., Thiran Jean-Philippe (2012), Evaluation of Atlas Fusion Strategies for Segmentation of Head and Neck Lymph Nodes for Radiotherapy Planning, in IEEE International Symposium on Biomedical Imaging (ISBI), Barcelona, Spain, May 2-5, 2012.
Active deformation fields: Dense deformation field estimation for atlas-based segmentation using the active contour framework
Gorthi S, Duay V, Bresson X, Cuadra MB, Castro FJS, Pollo C, Allal AS, Thiran JP (2011), Active deformation fields: Dense deformation field estimation for atlas-based segmentation using the active contour framework, in MEDICAL IMAGE ANALYSIS, 15(6), 787-800.
COMPARISON OF ENERGY MINIMIZATION METHODS FOR 3-D BRAIN TISSUE CLASSIFICATION
Gorthi S, Thiran JP, Cuadra MB (2011), COMPARISON OF ENERGY MINIMIZATION METHODS FOR 3-D BRAIN TISSUE CLASSIFICATION, in IEEE International Conference on Image Processing, Brussels, Belgium.
Comparison of Energy Minimization Methods for 3-D Brain Tissue Classification
Gorthi Subrahmanyam, Thiran Jean-Philippe, Bach Cuadra Meritxell (2011), Comparison of Energy Minimization Methods for 3-D Brain Tissue Classification, in IEEE International Conference on Image Processing, Brussels, Belgium, September 11-14, 2011.
Fusion of Multi-Atlas Segmentations with Spatial Distribution Modeling
Gorthi Subrahmanyam, Bach Cuadra Meritxell, Schick Ulrike, Tercier Pierre-Alain, Allal Abdelkarim S., Thiran Jean-Philippe (2011), Fusion of Multi-Atlas Segmentations with Spatial Distribution Modeling, in MICCAI Worskshop on Multi-Atlas Labeling and Statistical Fusion, Toronto, Canada, September, 18-22, 2011.
Multi-Atlas based Segmentation of Head and Neck CT Images using Active Contour Framework
Gorthi Subrahmanyam, Bach Cuadra Meritxell, Schick Ulrike, Tercier Pierre-Alain, Allal Abdelkarim S., Thiran Jean-Philippe (2010), Multi-Atlas based Segmentation of Head and Neck CT Images using Active Contour Framework, in MICCAI workshop on 3D Segmentation Challenge for Clinical Applications, Beijing, China, September 20-24, 2010.
Active Contour-Based Segmentation of Head and Neck with Adaptive Atlas Selection
Gorthi Subrahmanyam, Duay Valérie, Bach Cuadra Meritxell, Tercier Pierre-Alain, Allal Abdelkarim S., Thiran Jean-Philippe (2009), Active Contour-Based Segmentation of Head and Neck with Adaptive Atlas Selection, in MICCAI workshop on 3D Segmentation Challenge for Clinical Applications, London, September 20-24, 200.
Segmentation of Head and Neck Lymph Node Regions for Radiotherapy Planning Using Active Contour-Based Atlas Registration
Gorthi S, Duay V, Houhou N, Cuadra MB, Schick U, Becker M, Allal AS, Thiran JP (2009), Segmentation of Head and Neck Lymph Node Regions for Radiotherapy Planning Using Active Contour-Based Atlas Registration, in IEEE JOURNAL OF SELECTED TOPICS IN SIGNAL PROCESSING, 3(1), 135-147.

Scientific events

Active participation

Title Type of contribution Title of article or contribution Date Place Persons involved
MICCAI Worskshop on Multi-Atlas Labeling and Statistical Fusion 18.09.2011 Toronto, Canada
IEEE International Conference on Image Processing 11.09.2011 Brussels, Belgium
MICCAI workshop on 3D Segmentation Challenge for Clinical Applications 20.09.2010 Beijing, China
MICCAI workshop on 3D Segmentation Challenge for Clinical Applications 20.09.2009 London, UK


Associated projects

Number Title Start Funding scheme
144334 MRF-based Fusion Framework for the Automated Parcellation of Human Cerebral Cortex 01.10.2012 Project funding (Div. I-III)

Abstract

Intensity modulated radiotherapy (IMRT) is the high precision technique to accurately deliver X-ray radiation treatment for different tumor locations. However, its implementation in routine clinical practices is facing significant obstacles, particularly in the treatment of lymph nodes of the Head and Neck (H&N) region. The major bottleneck is the inability to accurately and automatically delineate the lymph node regions. Lymph node regions are constructed volumes in the H&N region and they often do not have any visibly distinct boundaries in the images; rather, they are defined with respect to other visible landmark structures in the image and hence, posing challenges in their automated segmentation.During our previous SNF project (number 3252B0-107873), we proposed a novel atlas-based joint registration and segmentation model for segmenting the H&N lymph node regions. Unlike previous approaches in the literature, we proposed to segment the lymph node regions without directly including them in the registration process; instead, they are segmented using the dense deformation field computed from the registration of selected structures with distinct boundaries and surrounding the lymph node regions. The lymph node segmentation results from the proposed method are evaluated on a dataset of 10 patients, and the results are compared with two other state of the art methods: “Radial basis function algorithm” and “Demons algorithm”. The comparison is done using various statistical and geometrical metrics: sensitivity, dice similarity coefficient and Hausdorff distance. The proposed method is found to perform a more accurate segmentation than the other methods. Although our current segmentation scheme is more accurate than other methods, it can be noticed from the quantitative evaluation presented later, that it is still far from an ideal segmentation. The objective of this proposal is to further improve the robustness as well as accuracy of the H&N lymph node regions segmentation, and thereby minimize, if not able to completely eliminate, the corrections to be made by the Physician to the automated segmentation results. The main ideas that we propose here for improving the lymph node regions segmentation are as follows:- Development of a self-adapting framework that learns from the error-correction process and performs a more accurate segmentation when used next time.- Selection of patient specific subpopulation of atlases that best suit the patient’s image to be segmented and construction of an optimal statistical atlas from the selected subpopulation.- Identification and labeling more structures in the atlas image that are close to the lymph nodes and have distinct boundaries. This will result in more accurate deformation field computation and thus more accurate lymph nodes segmentation.The proposed research ultimately envisages developing a clinically usable software system for the automated segmentation of target volumes for IMRT of the H&N region. This would be of great value for the radiation-oncology field in general and will represent a starting point of the widespread use of the IMRT technique.This is a proposal for a 3-year project with one PhD position; this is for the continuation of the SNF project number 3252B0-107873.
-