In spite of recent advances in treatment, lung cancer continues to be the leading cause of cancer-related deaths worldwide [1
]. Imaging plays a pivotal role in the diagnosis, staging, therapy planning, response assessment and management of these patients. The primary imaging modality is contrast-enhanced computed tomography (CE-CT), as it provides excellent morphological assessment of the primary tumor, involvement of mediastinal lymph nodes and most sites of distant metastases. However, conventional CE-CT does not provide any functional information about tumor metabolism, perfusion or physiology [2
]. Therefore, in NSCLC, it is often combined with 18-Fluorodeoxyglucose positron emission tomography (18
F-FDG PET/CT) in order to provide information about glucose metabolism for staging [4
]. Other imaging techniques, such as dynamic-contrast-enhanced CT (DCE-CT) and dual energy CT (DE-CT), can be used to quantify perfusion parameters and iodine concentration, respectively [5
]. Although these imaging modalities represent different aspects of tumor physiology, previous studies have suggested a positive correlation between perfusion parameters, FDG uptake and iodine concentration [10
There are some inconsistencies in the current literature on this topic. Miles et al. showed an inconsistent relationship between the standardized perfusion value and standardized uptake value (SUV) dependent on tumor size. This study was comparable to the results of Tateishi et al., which showed a correlation between peak attenuation, relative flow and FDG uptake. Contrarily, van Elmpt et al. found no direct correlation between DCE-CT parameters and FDG uptake [10
Based on the correlation between iodine uptake and perfusion parameters, some studies have hypothesized that a correlation between DE-CT parameters and FDG uptake is plausible [15
]. Similar to studies on DCE-CT parameters, there are discrepancies in the current literature. The studies of Schmid-Bindert et al. and Baxa et al. found a strong correlation between SUVmax
and DE-CT parameters in patients with non-small-cell lung cancer (NSCLC), whereas the studies of Kupik et al. and Aoki et al. found no correlation regardless of tumor histology, but a moderate negative correlation for small primary tumors (<3 cm) [11
In all previous studies, either a small region of interest (ROI) or the entire tumor volume was assessed. As NSCLC often exhibits high intra- and intertumoral heterogeneity, assessment of a small region of the tumor is a limitation [11
The aim of the present study was to investigate the correlation between 18F-FDG PET/CT parameters and DE-CT-derived iodine concentrations for the total volume of primary lung tumors. In addition, the same correlation was assessed at the level of sub-volumes of the tumor.
2. Materials and Methods
2.1. Design and Patients
This study was retrospective, following the STROBE guidelines [19
], including all patients with NSCLC who had undergone both DE-CT and 18
F-FDG PET/CT between June 2016 and July 2017 at Aarhus University Hospital. In Denmark, it is standard for patients with NSCLC that can potentially undergo curative treatment to receive both a contrast-enhanced CT and an 18
F-FDG PET/CT. Inclusion criteria: histologically verified NSCLC, written informed consent for the use of image and patient data and a period of less than 60 days between DE-CT and 18
F-FDG PET/CT. Furthermore, to ensure a minimal difference of the tumors between the DE-CT and 18
F-FDG PET/CT, both examinations should have been performed prior to the initiation of treatment.
2.2. Contrast-Enhanced DE-CT Imaging Protocol
CT imaging was performed with a multiple-row dual-layer detector CT scanner (Philips IQon CT; Philips Healthcare, Best, The Netherlands). CT acquisition parameters were 64 × 0.625 mm collimation, kV 120–140, mAs/slice 150–250, rotation time 0.75 s, reconstruction thickness 2 mm, increment 1 mm, pitch 1.078, FOV 35 cm and matrix 512 × 512. Computed tomography (CT) examinations included the chest and upper abdomen. Iodixanol 320 mg/mL (Visipaque® 320; GE Healthcare, Boston, MA, USA) was injected intravenously at weight-adjusted doses of 1.7 mL/kg body weight to adjust for differences in distribution volume, and with an injection rate of 4 mL/s. A bolus tracking technique with an ROI in the descending aorta at the level of carina was used to adjust for differences in cardiac output. CT was performed during inspiration breath hold after a delay of 15 s for the chest and upper abdomen (late arterial phase), and 50 s for the upper abdomen (portovenous phase) after a threshold of 150 HU was obtained according to a standardized protocol at our institution. The arterial phase was chosen for assessment, as it covered the entire chest. From the scan, we obtained a contrast-enhanced CT scan reconstruction (CE-CT), and an iodine concentration (IC) map, providing the IC in mg per ml in each voxel.
2.3. 18F-FDG PET/CT Imaging Protocol
Whole-body 18F-FDG PET/CT was performed with an integrated PET/CT scanner (Siemens Biograph w. 64-slice CT scanner; Siemens Healthcare, Forchheim, Germany). Participants were instructed to fast for 6 h prior to the examination. Approximately 400 MBq FDG was injected intravenously. 18F-FDG PET/CT scans were performed after a delay of 60 min. The FDG PET acquisition was performed with respiratory gating on inspiration, and the images were corrected for scatter and iteratively reconstructed. The nonenhanced CT acquisition parameters were: 120 kVp, 50 mAs, CTDIvol 3 mGy, 24 × 1.2 mm collimation. Reconstruction parameters: 3.0 mm section thickness, and 1.5 mm increment. The CT scan was acquired with inspiration breath hold.
2.4. Delineation of Lesions and Image Assessment
An expert in thoracic radiology with 12 years’ experience performed all delineations of the primary tumors. The radiologist was blinded to patient identifiers, clinical data and tissue sampling. Delineations were performed semi-automatically using Intellispace v-11, Tumor Tracking (Philips Healthcare, Best, The Netherlands). An ROI including the total circumference of the primary lung tumor was delineated on the CE-CT. Visible blood vessels within the tumor were excluded from the ROI. The ROIs were saved as RT structure sets and transferred for further analysis.
2.5. Data Processing
Image registration: In-house developed software was used to register the imaging data. First, the CT scan from the PET was rigidly registered to the CE-CT scan. Correlation ratio was used as metric, only taking the tumor region plus a 1 cm margin into account. Both translations and rotations were used. After visual inspection of the CT registration, the transformation matrix was also applied to the PET scan. Again, visual inspection was performed, and manual adaptation of the registration was allowed to correct for any residual mis-matches which could have been caused by differences in the inspiration breath hold in the different imaging modalities.
Data analysis: The imaging data were cropped to the tumor volume plus a 1 cm margin, and all scans were resampled to a 1 mm uniform grid, and exported in nifty format. PET scans were first converted from raw counts to standard uptake values (SUVs). Data analysis was performed in python 3.7 (Python Software Foundation). For each tumor, the following parameters from the PET were calculated: SUVmax, SUVmean, metabolic tumor volume (MTV = SUV > 2.5), MTV50% (volume at the threshold of 50% of SUVmax), MTV42%, MTV30%, MTV20%, MTV10%; total lesion glycolysis (TLG = MTV × SUVmean) for all the different MTV threshold levels (TLGSUV2.5, TLG50%, TLG42%, TLG30%, TLG20%, TLG10%) and SUVpeak (highest average SUV for a 1 cm3 spherical ROI within the tumor).
Based on the IC, we calculated the following parameters from the tumor region: Iodinemax, Iodinemean, total iodine content (TIC = tumor volume × Iodinemean), Iodinepeak (highest average iodine concentration for a 1 cm3 spherical ROI within the tumor similar to what was used for SUVpeak), Iodinemax@SUVpeakROI (maximum iodine concentration within the ROI used for the SUVpeak).
Finally, we calculated the tumor volume based on the voxels included in the tumor ROI, and the distance between the centroid of the SUVpeak ROI, and the centroid of the Iodinepeak ROI.
2.6. Statistical Analysis
All PET parameters were correlated with the iodine parameters using Pearson’s correlation coefficient. QQ plots were used to test for normality. Parameters related to volume were not normally distributed, and therefore, a log10 transform of the data was used before calculating the correlation coefficient. Excel 2019 (Microsoft Corporation, Redmond, Washington, DC, USA) was used for the calculations, and p-values < 0.05 were deemed significant.
In this retrospective study, we found a moderate correlation between SUVmax and Iodinemax as well as a strong positive correlation between the MTV and total iodine content when considering the whole tumor. In addition, we also found a weak-to-moderate positive correlation between the tumor size and SUVmax and between the tumor size and Iodinemax. Furthermore, when correlating the SUVpeak with the Iodinemax the correlation was dependent on what area the latter was measured in. When considering the entire tumor volume, a moderate-to-strong positive correlation was found, while when only a 1 cm3 sphere surrounding the SUVpeak was considered, the association changed to a moderate negative correlation. Finally, we found a moderate positive correlation of the distance between the centroid of SUVpeak and iodinepeak with the tumor volume.
Previous studies investigating the correlation between FDG uptake and iodine concentration all have a common denominator. The methodology in tumor delineation either included the entire tumor volume or a small ROI without registration between the PET scan and the IC images. In the study of Schmid-Bindert et al., the iodine parameters were measured by placement of three ROIs within the tumor; no specifics regarding the placement of these are described, and the mean and standard deviation of the measurements were subsequently recoded. The 18
F-FDG PET-related measurements were based on an ROI encircling the entire tumor, including a 1 cm rim. Even though both data sets were available when the delineation was performed, it was performed by individual radiologists without a previous image registration between DE-CT and 18F-FDG PET/CT. In the study of Baxa et al., two radiologists performed the delineation of both the DE-CT and the 18
F-FDG PET/CT using a semiautomatic approach encircling the entire lesion. However, similar to the study of Schmid-Bindert et al., no registration between the two modalities was performed. This may explain the discrepancies between the previous studies and our study. Baxa et al. and Schmid-Bindert et al. found a correlation between metabolic parameters and iodine-related parameters, which is in contrast to the results of Kupik et al. and Masahiko et al., who found no correlation. In addition, Kupik et al. found a moderate negative correlation for tumors <3 cm [11
]. In this current study, we accounted for intratumoral heterogeneity by first performing an image registration between the DE-CT and the 18
F-FDG PET/CT followed by a correlation of maximum iodine concentration with SUVpeak
both within the entire tumor and in a sphere around the SUVpeak
. The results show that the maximum iodine concentration was located outside of the metabolic hotspot within the tumor, especially when the tumors became larger (Figure 4
). This is further substantiated by the results showing an increase in the distance between SUVpeak
and iodine peak locations in larger tumors (Figure 5
). It is apparent that the metabolic activity and iodine quantification were located in different areas of the tumor, underlining that the physiological processes behind the parameters are not the same.
F-FDG is a measurement for the metabolic activity. This has previously been correlated with high proliferation measured by the KI-67 index, suggesting the areas containing the SUVpeak
are rapidly growing areas within the tumor [20
]. As tumor cells are dependent on fast formation of new vessels by angiogenesis to supply the nutrients and oxygen needed to continue growth, the areas with a high proliferation will usually have poorly formed vessels with an inadequate basement membrane, causing them to leak [21
]. Folkman et al. described how this increases the interstitial pressure within the tumor, and, combined with a relative absence of lymphatic vessels, it can cause vascular compression to some extent, prohibiting iodinated contrast from reaching the vessels [22
]. In our study, we only had a late-arterial-phase DE-CT scan, which mainly represents blood flow and blood volume. The late-arterial-phase iodine concentration data likely show subsegments of tumor with a high degree of unobstructed normal vasculature that allow iodinated contrast to enter the area. This may be a potential explanation for the negative correlation between the SUV and iodine measurements within the SUVpeak
The primary limitation of this study was the small study population with only two squamous cell carcinoma patients; thus, a subgroup analysis on tumor type would not be meaningful. Secondly, there was a lack of histological and metabolomic correlation with our findings. A registration of the histological cross sections and metabolism of the tumor with high FDG and high iodine values could potentially explain the mechanisms behind our findings based on growth patterns, vessel density or vessel types and measurements of proliferation. Another limitation was the registration accuracy between the PET images and the SCE-CT quantitative images. As they were obtained at different points in time, and all had different protocols regarding breathing during the acquisition, we were not able to obtain a registration accuracy to the level where a voxel-by-voxel correlation could be performed. Furthermore, due to the retrospective design, there was a temporal difference between DE-CT and PET. Finally, this study was retrospective and a single-center study performed only on one platform for obtaining the iodine quantifications.