PET segmentation challenge using a data management and processing infrastructure

Brief

Automated segmentation of PET images for the delineation of tumor volumes has been the focus of intense research efforts for the last few years [1]. There has also been a few limited efforts to compare several methods on common datasets [2], but in the majority of cases, each method has been evaluated on different image sets according to different evaluation criteria, and a comparison of currently available methods based on literature analysis only is thus challenging, if not impossible. A MICCAI challenge is an interesting opportunity to compare numerous existing methods implemented by their original authors (thus avoiding issues associated with re-implementation of methods by others [3]) on a common set of images. This challenge will be organized and funded by France Life Imaging, a national French infrastructure dedicated to medical imaging sciences, and co-sponsored by the taskgroup 211 “Classification, Advantages and Limitations of the Auto-Segmentation Approaches for PET” of the American Association of Physicists in Medicine (AAPM)[1], following in particular the recommendations it has set regarding benchmarking efforts (choice of image datasets and evaluation strategies) [4], [5].

 

References

[1] B. Foster, U. Bagci, A. Mansoor, Z. Xu, and D. J. Mollura, “A review on segmentation of positron emission tomography images,” Comput. Biol. Med., vol. 50, pp. 76–96, Jul. 2014.

[2] T. Shepherd, M. Teras, R. R. Beichel, R. Boellaard, M. Bruynooghe, V. Dicken, M. J. Gooding, P. J. Julyan, J. A. Lee, and S. Lefèvre, “Comparative study with new accuracy metrics for target volume contouring in PET image guided radiation therapy,” IEEE Trans. Med. Imaging, vol. 31, no. 11, pp. 2006–2024, Nov. 2012.

[3] M. Hatt and D. Visvikis, “Regarding ‘Segmentation of heterogeneous or small FDG PET positive tissue based on a 3D-locally adaptive random walk algorithm’ By DP. Onoma et al,” Comput. Med. Imaging Graph. Off. J. Comput. Med. Imaging Soc., vol. 46 Pt 3, pp. 300–301, Dec. 2015.

[4] T. Shepherd, B. Berthon, P. Galavis, E. Spezi, A. Apte, J. Lee, D. Visvikis, M. Hatt, E. de Bernardi, S. Das, I. El Naqa, U. Nestle, C. Schmidtlein, H. Zaidi, and A. Kirov, “Design of a benchmark platform for evaluating PET-based contouring accuracy in oncology applications,” Eur. J. Nucl. Med. Mol. Imaging, vol. 39, pp. S264–S264, Oct. 2012.

[5] M. Hatt, J. Lee, C. R. Schmidt, I. El Naqa, C. Caldwell, E. De Bernardi, W. Lu, S. Das, X. Geets, V. Gregoire, R. Jeraj, M. MacManus, O. Mawlawi, U. Nestle, A. Pugachev, H. Schöder, T. Schepherd, E. Spezi, D. Visvikis, H. Zaidi, and A. Kirov, “Classification and evaluation strategies of auto-segmentation approaches for PET. Report of the American Association of Physicists in Medicine Task group 211,” Medical Physics, p. under revision, 2016.

 

Objectives and relevance to MICCAI

The goal of this challenge is to provide and maintain a comparative study of a range of algorithms on a common database, running on a common computational infrastructure. It also provides a snapshot of currently popular methods for PET image segmentation in tumor volume delineation. In addition, it will provide to the community a sizeable amount of training and test images containing rigorous ground-truth as well as being representative of challenging cases for evaluation. In addition to a common shared and open-access database, this challenge will also provide a computing solution, accessible through web services, that will allow an objective computational benchmarking of the different solutions.

 

[1]http://aapm.org/org/structure/default.asp?committee_code=TG211

 

Contact

In case of any questions, please contact: challenges-iam@inria.fr