Mathieu Hatt

Impact of tumor size and tracer uptake heterogeneity in (18)F-FDG PET and CT non-small cell lung cancer tumor delineation.

The objectives of this study were to investigate the relationship between CT- and 18F-FDG PET–based tumor volumes in non–small cell lung cancer (NSCLC) and the impact of tumor size and uptake heterogeneity on various approaches to delineating uptake on PET images. Methods: Twenty-five NSCLC cancer patients with 18F-FDG PET/CT were considered. Seventeen underwent surgical resection of their tumor, and the maximum diameter was measured. Two observers manually delineated the tumors on the CT images and the tumor uptake on the corresponding PET images, using a fixed threshold at 50% of the maximum (T50), an adaptive threshold methodology, and the fuzzy locally adaptive Bayesian (FLAB) algorithm. Maximum diameters of the delineated volumes were compared with the histopathology reference when available. The volumes of the tumors were compared, and correlations between the anatomic volume and PET uptake heterogeneity and the differences between delineations were investigated. Results: All maximum diameters measured on PET and CT images significantly correlated with the histopathology reference (r > 0.89, P < 0.0001). Significant differences were observed among the approaches: CT delineation resulted in large overestimation (+32% ± 37%), whereas all delineations on PET images resulted in underestimation (from −15% ± 17% for T50 to −4% ± 8% for FLAB) except manual delineation (+8% ± 17%). Overall, CT volumes were significantly larger than PET volumes (55 ± 74 cm3 for CT vs. from 18 ± 25 to 47 ± 76 cm3 for PET). A significant correlation was found between anatomic tumor size and heterogeneity (larger lesions were more heterogeneous). Finally, the more heterogeneous the tumor uptake, the larger was the underestimation of PET volumes by threshold-based techniques. Conclusion: Volumes based on CT images were larger than those based on PET images. Tumor size and tracer uptake heterogeneity have an impact on threshold-based methods, which should not be used for the delineation of cases of large heterogeneous NSCLC, as these methods tend to largely underestimate the spatial extent of the functional tumor in such cases. For an accurate delineation of PET volumes in NSCLC, advanced image segmentation algorithms able to deal with tracer uptake heterogeneity should be preferred.

SU‐E‐J‐97: Predictive and Prognostic Clinical Values of 18F‐FDG PET Based Measurements in Locally Advanced Esophageal Cancer Are Not Improved by Partial Volume Effects Correction

Purpose: We have previously demonstrated that functional tumor volume (TV) and Total Lesion Glycolysis (TLG=TV×SUVmean) measured on pretreatment 18F‐FDG PET scans were significant predictors of response and prognostic factors of survival, whereas SUV measurements were not. The objective of this study was to investigate the impact of partial volume effects (PVE) correction (PVC) on this clinical value. Methods: 50 patients with esophageal cancer treated with concomitant radiochemotherapy were retrospectively analyzed. 18F‐FDG PET baseline scans were corrected for PVE with iterative deconvolution with wavelet‐based denoising. Tumors were subsequently delineated using the Fuzzy Locally Adaptive Bayesian (FLAB) algorithm on both original and corrected images. Maximum and peak SUV, TV, mean SUV, and TLG were extracted and compared. The value of each parameter (with or without PVC) was investigated using Kruskal‐Wallis tests regarding response and Kaplan‐Meier curves regarding survival. Results: Whereas PVC had a significant quantitative impact on the absolute values of each parameter (up to more than 100% for SUVmax), the respective clinical value was not significantly modified whether for overall survival or response to therapy. No significant improvement was observed after PVC for the already established significant predictive and prognostic value of TV and TLG. Similarly, the non significant predictive and prognostic value of the various SUV measurements was not improved by PVC and was even lowered in some cases. Conclusions: PVC did not modify significantly the previously established clinical value of tumor volume or TLG. In addition, the limited value of SUV measurements in this context may therefore not be explained by the lack of PVC since PVC did not improve their clinical value and in most cases it even lowered it.

SU-E-J-113: Impact of 4D PET and Motion Correction in the Delineation of Gross Tumor Volume for Radiotherapy Treatment Planning in Lung Cancer

Purpose: To quantitatively assess the impact of motion correction in 4D‐PET images on the accuracy of automatic lungtumour volume delineation for radiotherapytreatment planning and the resulting dosimetry modifications.Methods: : Simulated 18F‐FDG‐PET data using the NURBS‐based Cardiac‐Torso phantom and Geant4 Application for Tomography Emission were considered. Homogeneous and heterogeneous spherical cases were designed with two tumor‐to‐background (T:B) contrasts (4 and 10) and 3 motion amplitudes (0.5, 1.5 and 2.5cm). Two more realistic cases derived from real clinical PET/CT datasets were also generated. Data were corrected for respiratory motion using two Methods: reconstruction incorporating elastic transformations and super‐resolution. The tumours were segmented with the Fuzzy Locally Adaptive Bayesian algorithm on each respiratory phase with or without correction and on the motion average image. For heterogeneous cases global and boost volumes were delineated. The union of the volumes at each phase with or without motion correction, the average PET volume and the volume with internal margins added to one respiratory phase were compared to the union of the simulated volumes (ground‐truths). For each of these target volumes, IMRT planning was used to compare the different motion management approaches in terms of impact on dosimetry.Results: The smallest and largest volumes were obtained on the average PET and the one with internal margins respectively. The best compromise between sensitivity and positive predictive value were obtained with corrected PET, with similar results for both corrections. The best compromise between ground‐truth volume coverage and organs‐at‐risk sparing was achieved by the volumes with most accurate delineation. The volume with margins always covered the ground‐truth Conclusions: 4D‐PET with motion correction leads to more accurate delineation of lungtumours and boost volumes in the presence of respiratory motion. This also leads to improved margins for treatment planning.

SU‐E‐J‐53: Multi Observation PET Image Fusion for Patient Follow‐Up Quantitation in Oncology

Purpose: In Positron Emission Tomography(PET)imaging, an early therapeutic response can be assessed based on variations of semi‐quantitative parameters such as maximum standardized uptake values (SUV max) measured in PET scans carried out before and during the treatment. However, these measurements do not reflect tumor volume or radiotracer uptake distribution variations.Methods: The proposed approach is based on multi‐observation imagestatistical analysis for merging several PET acquisitions to assess tumour metabolic volume and uptake variations. The parameters defining the mixture distribution are estimated using the stochastic expectation maximization (SEM) method combined with adaptive spatial correlation estimation (ASEM). The proposed fusion process (ASEM) has been applied to simulated and clinical follow‐up PETimages of patients classified as partial responders (PR). We have compared the multi‐observation fusion with threshold‐based (TB) used in clinical practice for the assessment of the therapeutic response, applied independently to each follow‐up scans. Volume errors (VE) and quantitative variations of SUV measurements and tumour volume (dV) were considered on simulated and clinical datasets respectively.Results: : On simulated datasets, the adaptive threshold applied independently on both images led to significantly (p<0.05) higher errors than the ASEM fusion in the first follow‐up scan (VE ASEM = −1±7%, VE TB = 21±8%), and higher for the second follow‐up scan (VE ASEM = −22±19%, VE TB = 28±9%). This trend was enhanced with the clinical datasets, for which the adaptive threshold exhibited incoherent volume variations of +14±167% due to large overevaluation of tumor volumes in the second scan due to challenging conditions of contrast and noise. On the other hand, dV ASEM was −65±11% which is more appropriate since patients are PR Conclusions: The ASEM method demonstrated more accurate tumor volume evolution estimation than threshold‐based independent segmentations. Future work will consist in applying the method to clinical multi‐tracers datasets.