Statistical Analysis of Tubular Structures in PET Imaging Data of Human Sarcoma

We consider the problem of representing the spatial metabolic activity of sarcomas, an infrequent form of cancer affecting soft tissue, cartilage and bone. The typical treatment protocol for sarcoma tumors is surgical resection, generally preceded by chemotherapy in advanced cases. The median survival time for patients with sarcoma is about five years after surgery. Sarcoma patients are usually followed using Positron Emission Tomography (PET) scanning with fluoro-deoxyglucose (FDG). Such imaging data are typically analysed in terms of maximum standardized uptake value (SUVmax) in the tumor region of interest (ROI) [4]. However, advanced spatial models for sarcoma tumors and novel quantitative descriptors of FDG-PET imaging data may enhance prognostic understanding and, ultimately, patient treatment. We previously proposed alternative measures of spatial heterogeneity of the FDG uptake pattern of sarcoma tumors [4, 8, 7]. This set of measures consisted in measuring the deviation of the uptake pattern from a simple ideal ellipsoidal form. Here we investigate a model that offers a refinement of this elliptical representation. Fitting this representation allows the derivation of novel tumor spatial features to further describe localized variations in the uptake pattern. In this model, the uptake intensity is analyzed from the central core to the boundary of the tumor. This approach yields model-derived measures of tumor core non-linearity, tumor phase of development and lack of fit of the uptake model. In particular, the phase of development measured at the tumor core is found to be significant in terms of prognosis.