1.1

Clinical PCCT

A phantom (Fig. 1) containing solutions of various contrast elements was prepared using a stock solution of 10 mg/mL elemental standard to prepare subsequent dilutions (Iodine, I; Gadolinium, Gd; Hafnium, Hf; Tantalum, Ta; Bismuth, Bi; Calcium, Ca). Dilutions were prepared in 2% nitric acid. For imaging studies, solutions were filled in 50 mL centrifuge tubes in concentrations of 10, 5, 2.5, 1.25, and 0.625 mg/mL. Our experiments used an FDA-approved clinical PCCT system (Siemens NAEOTOM Alpha) installed at Duke. The scans were performed at 120 kVp with a spiral pitch of 0.8 and used two fixed energy thresholds at 20 and 65 keV. The dose was 16 mGy. The phantom and mouse images were reconstructed with a dual-energy (DE) protocol using a Qr40 kernel and a voxel size of 0.29x0.29x0.4 mm, which yielded 70 keV virtual monoenergetic CT images (VMI) and two binned images (low energy, 20-65 keV; high energy, 65-120 keV). The reconstructed matrix size was 512x512. The reconstructed data was also decomposed on the scanner to create I maps overlaid on virtual non contrast (VNC) images. The binned images allowed alternative material decompositions.

Figure 1:

The elements phantom: (A) a photo, (B) CT image corresponding to VMI at 70 keV and showing measurement ROIs, and (C) the iodine decomposition map (red) overlaid on top of the VNC image. Note that the I map also shows positive enhancement for Gd, Bi, and Ca; however, the Hf and Ta vials are clearly separated from I.

A second ex vivo mouse scan was performed to achieve higher spatial resolution (0.15x0.15x0.2 mm, kernel Qr 98u). The radiation dose was increased to 64 mGy to reduce noise. The scanner reconstructed 70 keV VMI images, but does not currently reconstruct separate bin images or perform material decomposition for this high resolution scanning mode. For both the high and low resolution clinical images we performed post-reconstruction denoising using our rank-sparse kernel regression algorithm[6].

1.2

PC Micro-CT

Our PC micro-CT system contains two imaging chains based on using both an energy-integrating detector and a PCD [3]. Each chain uses a G-297 X-ray tube (Varian Medical Systems, Palo Alto, CA) with a 0.3 mm focal spot size powered by an Epsilon high-frequency X-ray generator (EMD Technologies, Quebec, Canada). For this work, we have used only the imaging chain with a SANTIS 1604 CdTe-based PCD developed by Dectris, Ltd. (www.dectris.com). This detector has 150-μm pixel size and four energy thresholds. The source-to-detector distances were 831 and 680 mm, giving a magnification of approximately 1.2. To extend the PCD field-of-view along the z-axis and to reduce ring artifacts in our reconstructions, the subject was placed on a translational stage, and scans were performed using a helical trajectory. Our selection of kVp (120) and energy thresholds (20, 34, 50 and 65 keV) was based on the clinical scanner settings. The first and the last threshold (20 and 65 keV) were selected to match the clinical thresholds, while the other two were intended to match I (33.2 keV) and Gd (50 keV) K-edges. We used 200 ms exposures and 1 mA for each projection and 900 projections at each threshold. Based on ionization chamber measurements, the absorbed radiation dose was ∼100 mGy. The projection data were air normalized, and the gaps between the tiles of the PCD were filled via interpolation.

Image Reconstruction: We performed an iterative reconstruction using the split Bregman method with the add-residual-back strategy [7] and rank-sparse kernel regression regularization (RSKR,[6]), solving the following optimization problem: and solved the following optimization problem:

This algebraic reconstruction problem solves for the vectorized, reconstructed data, the columns of X, for each energy threshold simultaneously (indexed by e). The reconstruction for each threshold minimizes the reprojection error (R, system projection matrix) relative to the log-transformed projection data acquired at each threshold (the columns of Y). To reduce noise in the reconstructed results, this data fidelity term is minimized subject to the bilateral total variation (BTV) measured within and between energies via RSKR.

1.3

Material Decomposition

Extending the approach of Alvarez and Macovski [8], we performed post-reconstruction spectral decomposition both on clinical PCCT and PC micro-CT data. On the clinical DE PCCT data, we studied the following basis material decomposition: a) I and Ca, b) I and Gd and c) I and Ta. Using the PC micro-CT data we have performed a) I and Gd decomposition using 20 and 65 keV threshold images (unsubtracted) and b) photoelectric effect (PE), Compton scattering (CS), I and Gd decomposition using all 4 thresholds images. Vials of known concentrations of each material were used to construct a sensitivity matrix of enhancement per unit concentration (HU/mg/mL). Spectral decomposition was then performed by matrix inversion, followed by sub-space projection to prevent negative concentrations. After decomposition, the material maps were color coded and combined in ImageJ for visualization.

1.4

Figures of merit

We have assessed our noise performance by measuring the standard deviation in water vials. The sensitivity of DE CT to particular contrast materials can be quantified by the DE index computed as follows:

where XE1, XE2 are the measured CT numbers normalized by concentration for each element at the two energies.

Another figure of merit for material decomposition used here was the condition number of the sensitivity matrix (after unit normalization of each material’s sensitivity vector). The condition number is the ratio of the largest and smallest singular values of a matrix, and it quantifies the potential for error amplification when performing matrix inversion (decomposition). A better spectral decomposition would correspond to a lower condition number.

1.5

Animal Experiments

For the tumor imaging study, we used a carcinogen-induced and genetically engineered primary model of soft tissue sarcoma developed in p53^fl/fl mice. Primary sarcoma lesions were generated in the hind limb by intramuscular delivery of adenovirus expressing Cre recombinase [9], followed by injection of 300 μg 3-methylcholanthrene (Sigma). Tumors resembling human undifferentiated pleomorphic sarcoma developed approximately 8-12 weeks after injection. Imaging studies were initiated when tumors were palpable (>100 mg). Liposomal-based contrast agents containing iodine (Lip-I) and gadolinium (Lip-Gd) were fabricated similar to methods described previously[10]. Four mice with sarcoma tumors were intravenously injected with Lip-Gd contrast agent (0.32 mg Gd/kg body weight). Three days later (Day 3), the same animals were injected with Lip-I (1.32 mg I/kg body weight) and shortly after they were euthanized. The mice were next immersed in vials with formalin and scanned with both clinical PCCT and PC micro-CT.