Physics Contribution
Comparison of Five Segmentation Tools for 18F-Fluoro-Deoxy-Glucose–Positron Emission Tomography–Based Target Volume Definition in Head and Neck Cancer

https://doi.org/10.1016/j.ijrobp.2007.07.2333Get rights and content

Purpose

Target-volume delineation for radiation treatment to the head and neck area traditionally is based on physical examination, computed tomography (CT), and magnetic resonance imaging. Additional molecular imaging with 18F-fluoro-deoxy-glucose (FDG)–positron emission tomography (PET) may improve definition of the gross tumor volume (GTV). In this study, five methods for tumor delineation on FDG-PET are compared with CT-based delineation.

Methods and Materials

Seventy-eight patients with Stages II–IV squamous cell carcinoma of the head and neck area underwent coregistered CT and FDG-PET. The primary tumor was delineated on CT, and five PET-based GTVs were obtained: visual interpretation, applying an isocontour of a standardized uptake value of 2.5, using a fixed threshold of 40% and 50% of the maximum signal intensity, and applying an adaptive threshold based on the signal-to-background ratio. Absolute GTV volumes were compared, and overlap analyses were performed.

Results

The GTV method of applying an isocontour of a standardized uptake value of 2.5 failed to provide successful delineation in 45% of cases. For the other PET delineation methods, volume and shape of the GTV were influenced heavily by the choice of segmentation tool. On average, all threshold-based PET-GTVs were smaller than on CT. Nevertheless, PET frequently detected significant tumor extension outside the GTV delineated on CT (15–34% of PET volume).

Conclusions

The choice of segmentation tool for target-volume definition of head and neck cancer based on FDG-PET images is not trivial because it influences both volume and shape of the resulting GTV. With adequate delineation, PET may add significantly to CT- and physical examination–based GTV definition.

Introduction

Progress in radiation oncology enables delivery of radiation treatment with increasing geometric precision. This requires reevaluation of target-volume delineation, which traditionally is based on physical examination, computed tomography (CT), and magnetic resonance imaging (MRI). In recent years, new methods were introduced for visualization of tumor tissue. In addition to anatomic data supplied by CT and MRI, “functional” and “molecular” imaging techniques, such as positron emission tomography (PET), single-photon emission CT, and magnetic resonance spectroscopy, allow visualization of biologic characteristics with several potential advances. The primary tumor may be identified more accurately, with consequences for the size and shape of the gross tumor volume (GTV). Tumor characteristics relevant for radiation sensitivity can be visualized (e.g., hypoxia), which may assist in the selection of patients for customized treatments (1). Also, intratumoral heterogeneity of these characteristics may be identified, providing an opportunity for “dose painting” (2). Finally, when imaging modalities become more accurate, interobserver and intraobserver variations in tumor delineation will decrease, resulting in improved standard of care.

Metabolic information, provided by imaging 18F-fluoro-deoxy-glucose (FDG) with PET, was incorporated into target-volume delineation by many groups (3). Tumor localizations can be identified and localized with high sensitivity because of the high-contrast resolution of PET. However, application of FDG-PET data for target-volume delineation is not straightforward, as identification of tumor boundaries on PET suffers from a relative low spatial resolution and a “blurry” appearance of lesions. Furthermore, FDG-PET usually is interpreted qualitatively in diagnostic nuclear medicine, whereas in radiation oncology a more quantitative approach is required for tumor contouring (4). Currently, various methods for FDG-PET–based target-volume definition are in use. Visual interpretation is the most commonly used method 5, 6, 7, 8, 9, 10, 11, 12. However, this method is susceptible to the window-level settings of the images and is highly operator dependent. Therefore, other more objective methods were explored. Examples are isocontouring based on either a standardized uptake value (SUV) of 2.5 around the tumor 10, 13, 14, a fixed threshold of the maximum signal intensity 15, 16, 17, 18, 19, 20, 21, or a threshold that is adaptive to the signal-to-background ratio (SBR) (22). The utility of these methods for tumor delineation in the head and neck area currently is unknown.

The choice of method for tumor delineation on FDG-PET may influence GTV determination, with consequences for the outcome of radiation therapy. The aim of this study is to compare different methods for tumor delineation with FDG-PET relative to CT-based delineation for radiation therapy planning in patients with head and neck cancer.

Section snippets

Patients

Seventy-eight patients (59 men, 19 women; median age, 61 years; range, 43–86 years) with Stages II–IV squamous cell carcinoma of the head and neck area eligible for primary curative radiotherapy were prospectively enrolled from June 2003 to July 2006. Tumor characteristics are listed in Table 1. The study was approved by the Ethics Committee of the Radboud University Nijmegen Medical Centre, and all patients provided informed consent.

Image acquisition

Before treatment, a CT scan and FDG-PET scan were acquired in

Results

Seventy-eight patients were included in this study. Of these, 77 data sets were available for analysis; 1 patient was excluded because the primary tumor, a T2N2cM0 oropharyngeal carcinoma, was not visualized by using FDG-PET.

The GTVVIS could be generated for all 77 patients. The GTVSBR segmentation tool resulted in unsuccessful volume definition in 2 patients. This was observed in 4 patients for both GTV40% and GTV50%, 2 of whom also had an unsatisfactory GTVSBR. The GTVSUV determination was

Discussion

In this study, we compared five segmentation tools for FDG-PET–based target-volume definition in a large cohort of patients with head and neck cancer. There were three important observations. First, the GTVSUV method using a fixed threshold of 2.5 failed to provide successful delineation in a large number of cases. Second, the volume and shape of the GTV on PET largely depended on the segmentation tool used. Third, PET frequently detected extension of tumor tissue outside the GTVCT regardless

Conclusions

This study shows that FDG-PET may have important consequences for GTV definition, but the choice of a segmentation tool for target-volume definition of head and neck cancer based on PET images is not trivial. The absolute PET volume is dependent on the segmentation method used. Delineation using an SUV of 2.5 is insufficient and the other evaluated methods show inconsistencies. The SBR method seems preferable because it uses a threshold adapted to the SBR of an individual patient and does not

Acknowledgments

The authors thank John Lee Eng. Ph.D., for helpful collaboration in conducting calibrations of the adaptive segmentation tool developed at Université St Luc, Brussels, Belgium.

References (34)

Cited by (184)

View all citing articles on Scopus

Supported in part by E.C. F.P.6 funding (BIOCARE, L.S.H.C.-C.T.-2004-505785).

Conflict of interest: none.

View full text