Dual-Layer Spectral CT Imaging of Upper Aerodigestive Tract Cancer: Analysis of Spectral Imaging Parameters and Impact on Tumor Staging

CT Scanning Technique

All patients underwent a head and neck protocol using the same DLSCT scanner (IQon Spectral CT). Images were acquired after injection of 50 mL of Optiray 350 (Mallinckrodt; 74.0 g/100 mL) at a rate of 1.5 mL/s, followed by a further 50 mL of Optiray 350 injected at a rate of 2 mL/s after a delay of 100 seconds. Patients were scanned 160 seconds after the start of the first contrast bolus. The scanned volume extended from the aortic arch to the top of the ear with a reconstructed FOV of 250 mm, 64 × 0.625 mm beam collimation, 0.27-second rotation time, and 1.234:1 helical pitch. IQon Spectral CT uses conventional CT acquisitions with a dual-layered scintillation detector, comprising an upper layer of an yttrium-based garnet low-density scintillator for detection of lower energy photons and a lower layer of a gadolinium oxysulphide high-density scintillator for detection of higher energy photons. The design of the thickness of the 2 scintillators is optimized for energy separation and an optimal SNR. While all photons are represented superficially, lower energy photons are filtered out with increased depth. The difference in detected photon energy is secondary to augmentation by tissues of different densities. Combining the raw data from each detector layer, IntelliSpace Portal software (Philips Healthcare) reconstructs CT images as conventional CT and additional spectral data overlay maps for each parameter.

Specifically, 2 spectral parameters, iodine density with and without color overlay and Z-effective color overlay, were analyzed in this study. The Z-effective value is the effective atomic number of the tissue, representing the combination of atomic numbers of the elements that comprise the tissue.⁶ The Z-effective value is measured by the ratio between 2 monoenergetic density coefficients, according to the Joshi Method, and is calibrated on the DLSCT system by fitting measured data to expected values from the National Institute of Standards and Technology data base.^6,8 The Z-effective value is dimensionless and has no unit of measure. The calculation of the ratio between density coefficients at 2 energies eliminates the dependency on material density. Non-iodine-enhanced soft tissue is expected to have an effective atomic number of about 7.5, whereas with the presence of iodine, the effective atomic number is increased, accounting for the iodine molecules within the soft tissue. The iodine density result is derived by projecting each voxel from the Compton scattering/photoelectric base to the water/iodine base and quantifying the magnitude of the iodine projection.⁶ The water/iodine base is defined using the National Institute of Standards and Technology data base, in which water and iodine densities are projected to the Compton scattering and photoelectric base functions. Non-iodine-enhanced soft tissue is expected to have zero or close to zero iodine content in contrast to iodine-enhanced tissue. Iodine density is measured by milligram/milliliter.

Postprocessing and Image Analysis

For each patient, qualitative assessment and quantitative spectral analysis were performed using the IntelliSpace Portal software (Philips advanced visualization platform, Version 9.0) for the iodine density, iodine density color overlay, and Z-effective images. Each study was initially read independently by 2 readers, C.C.-T.H. (a fellowship-trained neuroradiologist with 5 years of experience) and L.D. (a fellowship-trained radiologist and nuclear medicine specialist with 10 years of experience), with final placement of ROIs performed together through a consensus reading session. Normal structures were evaluated with 3 nonoverlapping circular ROIs. A total of 10 mucosa and 3 muscle subsites were evaluated, each with 3 ROIs, corresponding to a total of 39 ROIs per control subject. The average ROI area evaluated per structure was 53.2 mm² (range, 25.4–67.8 mm²). For HNSCC, each tumor was evaluated, with a total of 9 ROIs placed on at least 3 separate sections. The tumor ROIs were placed in the homogeneous-appearing enhancing part of the lesion, avoiding areas of cystic change/necrosis or visible artifacts. The minimum individual ROI diameter used was 3 mm (sampled ROI of 7.1 mm²), and the maximum diameter used was 5 mm (sampled ROI of 19.6 mm²) (Fig 1). Mean (SD) iodine density (milligrams/milliliter) and Z-effective values were measured within the ROIs. The average value of the 3 ROIs per area was recorded. Normative data for mucosa and muscles were obtained at 6 different levels in the head and neck on the axial plane to enable sampling of the mucosa subsites, including the floor of the mouth, buccal mucosa, oral tongue, palatine tonsil, soft palate, tongue base, aryepiglottic fold, vocal fold, post-cricoid hypopharynx, and cervical esophagus. Measured neck muscles from cranial to caudal included the following: longus coli, genioglossus, and the sternocleidomastoid muscle. For consistency, the right side of the neck was evaluated, except for the genioglossus muscles, for which because of their relatively small size, the largest available structure of both sides was evaluated. The time taken to interpret and perform the spectral analysis in healthy controls and subjects with HNSCC was recorded in minutes.

• Download figure
• Open in new tab
• Download powerpoint

FIG 1.

Conventional (A), iodine density (B), and Z-effective (C) images illustrating an ROI placed over a right tongue base squamous cell carcinoma. ROI size is 4.89 mm in diameter (ED), with an average area (Av) of 18.76 mm². ROI measurements were obtained over 3 image slices (not shown), with the average iodine and Z-effective values calculated. A right level IIa cystic nodal metastasis is also present (not labeled).

Local Tumor Staging

Two readers (C.C.-T.H. and L.D.) blinded to the clinical history reviewed the DLSCT studies independently; if any discrepancy occurred, the final staging was decided by consensus. Two reading sessions occurred a week apart. In the first reading session, only the conventional CT images (120 kV[peak]) were available. In the second reading session, the conventional CT images with the addition of the iodine density, color overlay iodine density, and Z-effective spectral images were available. The local tumor staging data were compared with the surgical tumor stage determined from the multidisciplinary team meeting after histopathologic confirmation. The outcome of radiologic assessment compared with surgical staging was recorded as no change in T-staging, upstaging, or downstaging.

Statistical Analysis

All statistical analyses were computed using STATA (Version 12.1; StataCorp). Mucosa, muscle, and tumor tissue iodine density and Z-effective values were reported as mean and SD. Comparison of iodine density and Z-effective values between HNSCC and healthy muscle and mucosa was performed using an unpaired 2-tailed Student t test. The absolute contrast difference between HNSCC and healthy mucosa iodine and Z-effective images was calculated because both the iodine density and Z-effective images are noiseless with the SD approaching 0. Receiver operating Characteristic (ROC) curves quantified the performance of iodine density and Z-effective values to discriminate tumor from normal mucosal tissue. ROC areas under the curve were presented as an overall assessment of the diagnostic value of each technique and were compared. An optimal cut-point for each was determined (Youden Index: maximum sensitivity + specificity–1) that maximized both sensitivity and specificity. Using these cut-points for the Z-effective score and iodine density, we compared tumor identification with the predefined histopathologic tumor status. The κ statistic was used to assess the level of agreement between each spectral technique and known tumor status. The number and details of nonagreement were also noted. Finally, descriptive statistical analysis of the local tumor staging of conventional CT images only versus conventional CT with the addition of each spectral data technique was compared with the criterion standard surgical staging/histopathology. Time taken to interpret and perform the spectral analysis was recorded. Mean (SD) and range taken for each reader to interpret the spectral data were calculated.