%0 Journal Article %A Pierrick Gouel %A Paul Desbordes %A Frederic Di Fiore %A Pierre Vera %A Romain Modzelewski %T Predictive value of Intra tumor heterogeneity on baseline 18F-FDG PET images in esophageal cancer. %D 2015 %J Journal of Nuclear Medicine %P 1403-1403 %V 56 %N supplement 3 %X 1403 Objectives The aim of this study is to assess the early predictive value of 18F-FDG uptake heterogeneity characterized by textural features from pre-therapeutic PET in esophageal cancer treated by chemo/radio therapy.Methods 59 patients presenting oesophageal cancer were included in this study. All exams were performed before any treatment with identical parameters. Metabolic volumes (MV) were defined with 6 semi automatic delineation methods (40% and 50% threshold of the SUV max, adaptive, Nestle, Fitting, Black) and one manual segmentation made by a senior physician. 45 indexes were calculated for each delineation: 6 standard metabolic indexes, 36 textural features derived from first order,(12) co-occurrence, (11) uniform length matrix, Grigsby coefficient and 3 shape index. Each index was computed in 32 and 64 gray levels. Response to treatment was assessed after 1month by physiological examination.Results ANOVA analysis showed no significant difference between MV determined by a 40 %, adaptive, Nestle ,Fitting and manual segmentation. 50% threshold of the SUVmax and Black segmentation showed significant difference compared to manual segmentation. MV is not predictive of an early response except for Black segmentation method (Mann and Whitney, p<0,03). The variability of index according to the segmentation methods used, in particular for indexes derived from co-occurrence matrix. The relative resampling in 32 and 64 grey levels did not modify the measure of heterogeneity indexes. For every segmentation methodology, intra tumor heterogeneity feature given by the SGLGE parameter coming from run length matrix was predictive to an early response at 1 month.Conclusions There is no variability of the heterogeneity indexes resampled in 32 or 64 grey levels. We did not highlight intra tumor heterogeneity indexes derived from the SUV or derived from texture analysis capable of predicting the early response to the treatment at 1 month in oesophageal cancer. %U