Project: Segmentation Applied to Lesions delineated on PET images: Optimization of MEthods

In radiation therapy (RT), staging, treatment planning, monitoring and evaluation of response are traditionally based on computed tomography (CT) and magnetic resonance imaging (MRI). These radiological investigations have the significant advantage to show the anatomy with a high resolution, being also called anatomical imaging. In recent years, so called biological imaging methods which visualize metabolic pathways have been developed. These methods offer complementary imaging of various aspects of tumour biology, of high interest for the planning of local treatments like radiotherapy. To date, the most prominent biological imaging system in use is positron-emission tomography (PET), whose diagnostic properties have clinically been evaluated for years. In the context of radiotherapy planning, most authors have concentrated on FDG-PET, which shows the glucose metabolism that is increased in most solid and lymphatic malignancies. This tracer can therefore be broadly used for tumour depiction in the context of staging, treatment planning and monitoring. However, other tracers, like amino acids for the tumour depiction of brain tumours or F-MISO depicting hypoxic tumour sub-volumes and other tracers in the diagnostic pipeline, are also of high interest in radiation oncology._x000D__x000D_Determination of a volume from PET for RT-planning (gross tumour volume: GTV) is a critical step. For this purpose, various basic approaches were reported in the literature (Erdi et al. 1997, Knight et al. 1996, Pötzsch et al. 2006). The basic method, available at many systems is the visual contouring by the experienced physician, in analogy to the method used for CT-based contouring. However, despite the introduction of clinical contouring protocols, a significant inter-observer variation (IOV) reCOs (Pötzsch et al. 2006). While absolute or relative thresholding methods are integrated in many PET-evaluation consoles and RT-planning systems, it has unfortunately been shown, that these methods do often not lead to correct GTVs. Other methods use the image contrast for delineation, therefore better respecting the background activity or apply gradients or statistical models. These more advanced methods obviously lead to more plausible GTVs. However, the resulting volumes may still differ substantially compared to each other, especially in inhomogeneous tumours. As therefore the method for GTV contouring may have significant impact on the size of the GTV (Nestle et al. 2005), which directly relates to tumour coverage and normal tissue protection in radiotherapy, this is a critical fact._x000D__x000D_Therefore, the aim of the project is to implement new segmentation algorithms for GTV contouring, developed by the different academic Participants of the Consortium with the possibility to generate a consensus contour with respect to different delineation approaches. Theses new tools will be integrated in the AQUILAB medical image processing platform, with a specific care for safe and user-friendly interface. In addition, the AQUILAB platform, yet assigned to comparison of various segmentation algorithms in CT and MRI, will be used to compare visual and traditional thresholding PET segmentation methods with the new implemented ones._x000D__x000D_The resulting product of this project will target radiotherapy centres cooperating with nuclear medicine sites equipped with PET scanners. The number of PET and PET/CT scanners has been evaluated to about 450 in Europe for 2010 and about 1500 in the United States (HBS Consulting), while it reCOs very low in other parts of the world COly due the high cost of a PET scanner (2-3 million €). It has been predicted that the number of PET/CT scanners will increase in the next years, since their accessibility is still limited with less than 0.8 PET scanners per million populations._x000D__x000D_The Consortium is composed by 5 Participants from 3 European Member States. The Consortium is lead by AQUILAB (Lille - France), a SME providing quality assurance and evaluation software for medical imaging and radiotherapy. Then UMANIA (Bergamo - Italy) is another SME specialized in user interface design and ergonomics. Finally, three research and clinical centres (CHB in Rouen - France, RO Freiburg - Germany, U703 in Lille - France) are completing the Consortium. They have developed original delineation methods of PET positive tissues. They bring their expertise and knowledge in GTV segmentation in PET imaging and in the follow up of treatment response using PET images.

Acronym	SALOME (Reference Number: 5949)
Duration	01/03/2011 - 31/08/2013
Project Topic	The Project aims to implement new segmentation algorithms for GTV contouring on PET images, with the possibility to generate a consensus contour with respect to different delineation approaches. The user-interface will be carefully design by involving radiotherapy specialists in the process.
Project Results (after finalisation)	The project COly resulted in the development of a single prototype for contouring the GTV on PET images. Three segmentation algorithms are combined to generate an optimal consensus contour
Network	Eurostars
Call	Eurostars Cut-Off 5

Project partner

Number	Name	Role	Country
5	AQUILAB SAS	Coordinator	France
5	Centre Henri-Becquerel	Partner	France
5	Department for Radiation Oncology, University Hospital Freiburg	Partner	Germany
5	UMANIA S.r.l.	Partner	Italy
5	University of Law and Health Lille 2 - Inserm Unit U703 - Interventional Therapies Assisted by image and Simulation Laboratory	Partner	France