CT and MRI have reasonable sensitivity but low specificity in identifying recurrent gynaecologic disease [1, 2]. Consequently, significant observer variation has been noted in contouring the GTVCT. 18F-FDG PET/CT plays an increasingly important role in the staging and management of gynaecologic cancer including RT treatment planning [3, 4]. 18F-FDG PET/CT has demonstrated a high sensitivity and accuracy of more than 90% with average specificity in locally advanced or recurrent gynaecologic pelvic carcinoma. Furthermore 18F-FDG PET/CT can help to distinguish between tumor recurrence and post-therapy changes [4, 8]. Kidd et al. have shown that cervical cancer patients treated with 18F-FDG PET/CT-guided IMRT had improved survival and decreased treatment related toxicity compared with patients treated with non-IMRT radiotherapy .
This delineation study evaluated inter- and intra-observer variability of CT-based and 18F-FDG PET/CT-based target volume delineation in local recurrent or postsurgical residual gynaecological cancer. The results were compared with an automated PET segmented technique using adaptive thresholding technique. In other cancer sites such as head and neck and lung, 18F-FDG PET/CT was reported to decrease inter- and intra-observer variability in tumor contouring . Our results suggest that GTV delineation using 18F-FDG PET/CT could be superior to CT alone in this group of patients. GTVPET/CT was significantly smaller than the GTVCT with a trend for reduced inter- and intra-observer variability using PET/CT. The inter-observer agreement was moderate for the GTVCT and substantial for the GTVPET/CT. The inter-observer reliability was lower than the intra-observer reliability. This is in agreement with observations made by other authors . It has been considered that the observers tend to agree more with themselves rather than with each other. Inter- and intra-observer variability has been mostly investigated in lung cancer and the increased observer reliability on 18F-FDG PET/CT in our study is in line with these findings . Only one study by our group evaluated the inter-observer variability in PET/CT-based target volume delineation in the pelvis . A trend of reduced inter-observer variability has been observed in the delineation of the intraprostatic recurrence lesion using 18F-choline PET/CT. In gynaecologic cancer no inter- or intra-observer variability in PET-based GTV-delineation has been evaluated until now.
Our study demonstrate that the size of GTVPET/CT was significantly smaller than the GTVCT with the implementation of a coregistered 18F-FDG PET/CT. When the GTVSBR volumes were analyzed and compared with manual delineated target volume, it was observed that the GTVSBR was significantly smaller than the median GTVCT and GTVPET/CT. This was also manifested in the overlap analysis, where the overlap fraction increased from OFCT1 to OFPET/CT1 and OFSBR. Overall, the comparison of GTVs delineated in primary and recurrent cancer did not result in any significant differences.
The strength of our study includes the use of contrast enhanced CT scans for GTCT and GTVPET/CT determination and that the exams were performed on a dedicated PET/CT scanner for virtual simulation and fused with a recent MRI. Nevertheless the inter- and intra-observer variability was relatively high with both imaging modalities, highlighting the difficulty to determine the target volumes in this group of patients. An automated segmentation of the target volume using the adaptive thresholding technique could eventually help to reduce inter- and intra-observer variability. One potential limitation of our study is that the observers were at liberty to adjust the window, level and contrast setting of the images. This could have increased the inter- and intra-observer variability. However, all observers were experienced in PET/CT-based target volume delineation and were helped by a nuclear medicine physician. Another drawback of this study is the lack of comparison of the PET/CT results with pathologic findings after surgery.
The delineation of target volumes and organ at risk is a very critical step in high-precision RT treatment planning. Good image quality and reliable delineation protocol are important for accurate target volume delineation. One of the challenges of PET/CT-guided target volume delineation is the accurate segmentation of noisy and low resolution functional PET images. This is in particular true in recurrent or residual gynaecological cancer where vascular and urinary activity hampers target volume delineation. The result is a relatively high inter- and intra-observer variability. Various PET image segmentation techniques for target volume delineation were developed and evaluated to overcome this drawback . Among them, manual contouring by visual examination is the most commonly used method. The determination of an appropriate window and level for viewing the PET images is highly operator-dependent and is subject to high variability between operators .
An improved concordance in target volume delineation using PET/CT implies a greater accuracy and can help to determine a more appropriate treatment plan. In our study, the inter-observer variability coefficient prevailed is still relatively high. Variability negatively impacts the quality of treatments delivered to cancer patients. Alternatively an automated segmented target volume could be considered. There is consensus in the need for highly objective and automatic segmentation methods, and various groups have observed that semi- or fully-automated delineation techniques reduce inter-observer variability and improve reproducibility . The adaptive thresholding technique is one of the most widely used segmentation techniques for target volume determination in clinical setting. However, knowledge of the true target volume in relation the GTVSBR in gynaecologic tumors is needed for validation purposes. PET-based target volume delineation in gynaecologic tumors is actually not recommended outside clinical studies. It has to be emphasized that both patients with recurrent or postsurgical residual gynaecologic cancer are challenging cohorts for reliable target volume delineation and thus it is more likely that high inter- and intra-observer variability will result. In the absence of a more accurate information on the target volume position in gynaecologic cancer, a composite of GTVCT and GTVPET/CT can be recommended to optimize the GTV definition.