A new method for volume segmentation in PET: Use of possibility theory

F-18-fluorodeoxyglucose positron emission tomography ((18)FDG PET) is a nuclear medicine imaging technique of growing interest in oncology. In context of external radiotherapy, optimization of radiotherapy planning as well as objective patient monitoring require accurate tumor volume measurement. In this paper, a new, nearly automatic and operator-independent segmentation method that does not require prior calibration is presented. This method is based on possibility theory and, using membership degrees, it accounts for the uncertainty and inaccuracy inherent in the image. The results obtained on phantom data (percent error = 18,4% (mean) +/- 9.7 % (standard deviation)) and on simulated data (classification error = 23.6% (mean) +/- 18.6% (standard deviation) for SUV equal to 4.5) show the efficiency and the robustness of the method. (C) 2011 Published by Elsevier Masson SAS.