Tumor delineation in FDG-PET images using a new evidential clustering algorithm with spatial regularization and adaptive distance metric

Chunfeng Lian (1, 2) , Su Ruan (2) , Thierry Denoeux (1) , Hua Li , Pierre Vera (3, 2)

1 Heudiasyc - Heuristique et Diagnostic des Systèmes Complexes [Compiègne]
2 QuantIF-LITIS - Equipe Quantification en Imagerie Fonctionnelle
3 Service de médecine nucléaire [Rouen]

While accurate tumor delineation in FDG-PET is a vital task, noisy and blurring imaging system makes it a challenging work. In this paper, we propose to address this issue using the theory of belief functions, a powerful tool for modeling and reasoning with uncertain and/or imprecise information. An automatic segmentation method based on clustering is developed in 3-D, where, different from available methods, PET voxels are described not only by intensities but also com-plementally by features extracted from patches. Considering there are a large amount of features without consensus regarding the most informative ones, and some of them are even unreliable due to image quality, a specific procedure is adopted to adapt distance metric for properly representing clustering distortions and neighborhood similarities. A specific spatial regularization is also included in the clustering algorithm to effectively quantify local homogeneity. The proposed method has been evaluated by real-patient images, showing good performance .

Mots clés: PET Image Segmentation; Spatial Evi- dential c-Means; Adaptive Distance Metric; Feature Selection; Belief Functions

Domaines: Intelligence artificielle [cs.AI]; Traitement des images [eess.IV]; Imagerie médicale

Chunfeng Lian, Su Ruan, Thierry Denoeux, Hua Li, Pierre Vera. Tumor delineation in FDG-PET images using a new evidential clustering algorithm with spatial regularization and adaptive distance metric. 14th IEEE International Symposium on Biomedical Imaging (ISBI 2017), Apr 2017, Melbourne, Australia. pp.1177-1180, ⟨10.1109/ISBI.2017.7950726⟩. ⟨hal-02553198⟩