Respiratory-gated (4D) contrast-enhanced FDG PET-CT for radiotherapy planning of lower oesophageal carcinoma: feasibility and impact on planning target volume

Oesophageal cancer is the 8th commonest malignancy worldwide with approximately 456,000 cases diagnosed in 2012 [1]. Patients with locally advanced distal oesophageal cancer are increasingly treated with chemo-radiotherapy either in the neoadjuvant setting prior to definitive surgery or as stand-alone therapy [1]. Accurate target delineation, motion assessment and target localisation is a pre-requisite for high-precision radiotherapy treatment planning (RTP). Currently intravenous contrast-enhanced computed tomography (CT) in combination with endoscopic ultrasound are the standard techniques used for definition of gross tumour volume (GTV) prior to radiotherapy in oesophageal cancer [2]. A margin for microscopic extension is applied, the clinical target volume (CTV), and finally a margin for mechanical delivery uncertainties and internal organ motion relating to respiratory motion, results in a planning target volume (PTV). Contouring the GTV on conventional three-dimensional CT (3D CT) obtained during free-breathing may result in inaccurate representation of both tumour dimensions and mean tumour position relative to other organs. Four-dimensional CT (4D CT) performed with respiratory-gating, allows CT data acquired during the breathing cycle to be sub-divided into time-resolved 3D datasets (bins). The lower oesophagus moves significantly with breathing and 4D CT facilitates quantification of motion and allows patient-specific target volume delineation [3–5]. Use of 4D CT in RTP of oesophageal cancer is intended to ensure adequate coverage of the moving target volume within the radiation field and optimises normal tissue sparing compared to 3D CT [6]. 4D CT is now standard of care for RTP of lower third oesophageal cancer patients.

¹⁸Fluorine fluorodeoxyglucose (FDG) positron emission tomography – computed tomography (PET-CT) is firmly established in guiding optimal management of radically treatable oesophageal carcinoma [7, 8]. The role of FDG PET-CT in RTP of oesophageal carcinoma is less well developed. Preliminary studies evaluating the role of 3D PET-CT in RTP of oesophageal cancer have reported changes in target volume with potential impact on treatment planning [9–11]. Two recent studies by the same group evaluating 4D CT and 3D PET-CT in RTP of oesophageal cancer have reported that this combination of techniques impacted on target definition [12, 13]. However, motion artefacts with 3D PET-CT can reduce target contrast, overestimate lesion size and cause inaccurate assessment of standardized uptake value (SUV) [14]. Hypothetically motion management with respiratory gating to obtain a 4D PET-CT should provide additional information for RTP resulting in more consistent target definition [15]. There is a paucity of data in this clinical scenario with a single retrospective study having assessed dosimetric implications and another small prospective study evaluating the potential of 4D PET-CT for target volume delineation in oesophageal cancer, both showing promise for target volume definition [16, 17]. The impact of 4D PET-CT on PTV definition has not yet been reported in a prospective series to the best of our knowledge.

The purpose of this study was to assess the feasibility and potential impact on target delineation of contrast-enhanced 4D PET-CT acquired in the treatment planning position, for a prospective cohort of patients with lower third oesophageal cancer.

Respiratory motion affects imaging quality which is of direct relevance in RTP of potentially mobile tumours [21]. 4D CT guides generation of patient specific volumes accounting for physiological organ motion and is recommended in target definition protocols for lower oesophageal carcinoma [22]. Incorporating PET into the planning process has potential to improve the accuracy of target delineation but it is unclear if 4D PET-CT is superior to combining 3D PET with 4D–CT [12, 13, 15–17]. The aims of this pilot study were to assess the feasibility and impact of incorporating 4D PET-CT into the radiotherapy planning pathway of patients with lower oesophageal carcinoma.

This study has confirmed that incorporating 4D PET-CT into the RTP pathway of patients with lower oesophageal carcinoma is feasible. However, careful patient selection is important as it may not be possible in patients with an irregular breathing pattern. This prevented or significantly degraded 4D imaging in 4 patients in our study. Low-grade tumour uptake was also a limitation. The only other prospective study of 4D PET-CT in this setting reported 6/18 patients (33.3%) had imaging unsuitable for contouring for similar reasons [17]. Their study focused on optimal thresholding of PET data for GTV delineation and reported that a threshold setting of 20% standardised uptake value (SUV) or a fixed threshold of SUV 2.5 best correlated with tumour length, volume ratio and conformality index [17].

In this study 4D CT and PET data acquisitions were acquired at the same attendance with the patient in the radiotherapy treatment position with immobilization to ensure inherent co-registration. Post-processing was used to generate an averaged 3D PET from the 4D PET scan. This methodology should largely obviate any potential error introduced by registration of separately acquired CT and PET data. To the best of our knowledge, this is the first study to report positional metric comparison of tumour volumes delineated on 4D CT and 4D PET-CT in lower oesophageal carcinoma. Concordance between the contoured volumes on different studies were high with no statistical significance reached between PTV measured on each combination for the 5 positional metrics analysed, although the small sample size limits the value of statistical analysis; overlap indices were greatest between 3D PET-CT and 4D PET-CT (median DICE similarity coefficient 0.88, median sensitivity index 0.89) and slightly lower when comparing 4D CT with 3D PET-CT (median DICE similarity coefficient 0.85, median sensitivity index 0.78) and 4D CT with 4D PET-CT (median DICE similarity coefficient 0.85, median sensitivity index 0.79). Median centre of gravity distance was smallest for PTV_3DPET4DCT to PTV_4DPETCT structures (3.52 mm) and greatest for PTV_4DCT to PT_4DPET4DCT structures (5.19 mm). The median mean distance to conformity was lowest (7.41 mm) for the PTV_3DPET4DCT to PTV_4DPETCT and highest for the PTV_4DPETCT to PTV_4DCT (10.68 mm). The observed differences between PTVs are potentially clinically important; for example, the median sensitivity index of 0.79 between 4D CT and 4D PET-CT implies 21% of the PTV_4DPETCT is not included within the PTV_4DCT. The use of 4D PET-CT might minimise the risk of geometric miss but this has not been confirmed and requires further evaluation.

A key unanswered question is whether 4D PET-CT provides incremental benefit compared with 4D CT and 3D PET-CT. Although positional metrics comparing PTVs generated using these methods suggested differences were small, the sensitivity index of 0.89 (implying that 11% of the PTV_4DPETCT was not included within PTV_3DPET4DCT) suggests that there might be a potential additional role for the use of 4D PET-CT but the cohort is too small to draw conclusions on this. A small retrospective study of 4 patients with lower oesophageal cancer reported that target volumes delineated on inhale and exhale phase 4D PET-CT deformably co-registered with a planning CT were up to 30% smaller in 3 of 4 patients on the exhale plan with potential for dose sparing to organs at risk [16]. Conversely, a pilot study of 4D PET-CT for delineation of ITV in lung tumours reported concordance between different contoured volumes assessed using Dice coefficient and 3D PET-CT was consistently found to underestimate ITV compared to 4D PET-CT [23].

An important issue regarding use of PET in tumour volume delineation is the methodology employed to define the functional volume of interest. Several different methods have been proposed varying by tumour site and range from simplistic manual delineation to fully automated adaptive thresholding techniques [24, 25]. To date there is no clear consensus on which method is best and some nervousness about automatic delineation techniques [26]. Expert manual delineation by an experienced clinical oncologist in collaboration with a Radiologist/Nuclear Medicine Physician is established best practice [27]. In the absence of a widely accepted method in this clinical scenario, we made a pragmatic decision to manually contour PET volumes collaboratively as a Radiotherapist/Radiologist pair. Images were displayed using preset window levels and/or colour scale as per a standardized protocol which has been used in routine clinical practice for many years at our institution. Comparison of different methods for tumour volume segmentation has been previously reported in this setting albeit without pathological validation [17]. As a consequence this aspect was not a focus in this study.

PTVs delineated using 4D CT are often smaller than with 3D CT and doses to organs at risk in oesophageal cancer (lungs, liver and heart) are reduced using 4D CT which may facilitate dose escalation without significant collateral damage [28]. In this study, average tumour length was smallest on 4D PET-CT (range 5.1–10.6 cm), slightly larger (difference not statistically significant) on 4D CT (range 5.4–10.9 cm) and larger still on 3D PET-CT (range 6.4–11.1 cm). Tumour volumes were also consistently smaller on 4D PET-CT (average PTV 480.6 cm³) compared to 4D CT (average PTV 528.4 cm³), although the difference between volumes was not statistically significant. 3D CT represents a random snapshot during the breathing cycle whereas 4D CT is a selected series of snapshots at points on the breathing cycle. PET is different; 3D PET is a time-averaged image (showing motion blur covering the full breathing cycle) and 4D PET is a time-averaged image covering a phase bin where the amount of motion blur depends on degree of motion within that bin. Differences between volumes in part reflects this but also that PET and CT are demonstrating different tumour characteristics which are unlikely to inherently have the same volume but should provide complimentary information. 4D PET-CT volumes may have been smallest due to the benefit of PET and CT information but with a reduction in the motion blurring of 3D PET.

The study has limitations particularly the small cohort size and absence of histological validation of tumour volumes which is frequently the case in feasibility studies. Despite these limitations this small study has proven the feasibility of the technique and confirmed that tumour volume delineation using 4D PET-CT results in PTV differences compared with either 4D CT alone or 4D CT combined with 3D PET. Our data add to a recent literature review of the potential utility of 4D PET-CT in GTV delineation of intra-thoracic tumours which concluded that this may be the best approach currently available but warrant further investigation in future prospective studies [29]. Further work to establish if dose distribution varies for different PTVs and to evaluate outcome in a larger patient cohort with or without 4D PET-CT as part of RTP should be considered.

Acquisition of a planning 4D PET-CT is feasible with careful patient selection. PTVs generated using 4D CT, 4D CT co-registered with 3D PET-CT, and 4D PET-CT were of similar volume; however, percentage overlap analysis demonstrates that approximately 20% of the PTV_3DPET4DCT and PTV_4DPETCT are not included in the PTV_4DCT, theoretically leading to under-coverage of the target volume and a potential geometric miss. The current study is too small to draw any conclusions about clinical benefit and further investigation is warranted.