PT  - JOURNAL ARTICLE
AU  - Mathieu Hatt
AU  - Marie-Charlotte Desseroit
AU  - Baptiste Laurent
AU  - Dimitris Visvikis
AU  - Catherine Cheze Le Rest
TI  - <strong>Revisiting the identification of tumor sub-volumes predictive of residual uptake after (chemo)radiotherapy: comparison of segmentation methods on <sup>18</sup>F-FDG PET/CT images</strong>
DP  - 2018 May 01
TA  - Journal of Nuclear Medicine
PG  - 1750--1750
VI  - 59
IP  - supplement 1
4099  - http://jnm.snmjournals.org/content/59/supplement_1/1750.short
4100  - http://jnm.snmjournals.org/content/59/supplement_1/1750.full
SO  - J Nucl Med2018 May 01; 59
AB  - 1750Objectives: Several studies have suggested that PET high uptake pre-treatment sub-volumes have high overlaps with post-treament PET residual uptakes after (chemo)radiotherapy, which may lead to exploit these high uptake sub-volumes in radiotherapy planning. All these studies have used combinations of fixed thresholds (CFT) only. The aim of this study was to investigate the overlap between pre-treatment and post-treatment tumor PET uptakes, focusing on the image segmentation issues, using both CFT and automatic segmentation, as well as simulated images. Material and Methods: Simulated PET images with 3 different configurations (large, small or inexistent true overlap) allowed investigating the behaviour of each segmentation schemes, i.e. CFT between 30 and 90% of SUVmax or the fuzzy locally adaptive Bayesian algorithm (FLAB). Fifty-four patients with head and neck (H&N) or esophageal cancer treated with radiochemotherapy with both pre- and post-treatment PET/CT were retrospectively analyzed. PET scans were rigidly registered based on CT-CT registration. Tumour volumes were determined in both scans using each segmentation scheme. Four overlap metrics including Dice and overlap fraction were calculated. Results: The simulated study demonstrated that CFT can lead to biased (either large over- or under-estimation) overlap estimates, as well as accurate overlap metrics even through spatially inaccurate volumes. FLAB led to comparatively more consistently accurate estimations. In the clinical dataset, registration was feasible for esophageal but more difficult for H&N. Only 17 patients exhibited residual/relapse uptake smaller than the pre-treatment volume. Overlaps obtained with FLAB were consistently moderate for esophageal cases and low for H&N across the 4 metrics, whereas those obtained with CFT varied between low and high depending on thresholds values and metrics , which was consistent with the observations in the simulated cases. In both cases overlaps were highly variable across patients. Conclusions: Using an accurate and robust segmentation method, the overlaps between pre- and post-treatment volumes were moderate in esophageal, low in H&N, and highly variable across patients, which do not support optimization of radiotherapy planning based on pre-treatment PET/CT definition of a high-uptake sub-volume. In addition, only 35% of the initial cohort of patients could potentially benefit from such an optimization. Considering the comparison between segmentation schemes, we postulate that the use of combinations of arbitrary thresholds in previous studies may have led to overly optimistic overlap estimates.