Photon-counting CT offers the potential to provide new diagnostic information. In this study, we sought to determine the interplay between material identification errors, image quality indicators, and radiation doses using photon-counting spectral CT, and to determine whether this relationship is replicated in spectral CT scans of mice. Custom-built Perspex phantoms were used to measure signal-to-noise ratio and spatial resolution, and to measure radiation dose using thermoluminescent dosimeters. A multi-contrast calibration phantom containing inserts with different concentrations of gadolinium (1, 2, 4 and 8 mg/mL), hydroxyapatite rods (0, 54.3, 104.3, 211.7, 402.3 and 808.5 mg/mL) along with water and lipid was used to assess material identification errors. Image acquisition was performed using the MARS photon-counting scanner with four energy channels (30-45, 45-60, 60-78 and 78-118 keV) at four different tube currents (24, 34, 44 and 55 mA). As increased tube current showed no significant effect on material characterisation, small animal dosimetry was performed with 24 mA tube current using two non-contrast mice and one mouse injected with gadolinium. Results demonstrated that a tube current increase from 24 to 54 mA improved the signal-to-noise ratio and spatial resolution by <10%, gadolinium identification by 160%. Imaging results of the mice showed no obvious artefacts, and the mean absorbed dose measured for the three mice was 27.3±2.4 mGy. The results suggest that the energy resolving capability of photon-counting CT maintains diagnostically relevant image quality with high levels of material discrimination at reduced radiation dose.
Assessment of disease burden and drug efficacy is achieved preclinically using high resolution micro computed tomography (CT). However, micro-CT is not applicable to clinical human imaging due to operating at high dose. In addition, the technology differences between micro-CT and standard clinical CT prevent direct translation of preclinical applications. The current proof-of-concept study presents spectral photon-counting CT as a clinically translatable, molecular imaging tool by assessing contrast uptake in an ex-vivo mouse model of pulmonary tuberculosis (TB). Iodine, a common contrast used in clinical CT imaging, was introduced into a murine model of TB. The excised mouse lungs were imaged using a standard micro-CT subsystem (SuperArgus) and the contrast enhanced TB lesions quantified. The same lungs were imaged using a spectral photoncounting CT system (MARS small-bore scanner). Iodine and soft tissues (water and lipid) were materially separated, and iodine uptake quantified. The volume of the TB infection quantified by spectral CT and micro-CT was found to be 2.96 mm 3 and 2.83 mm 3 , respectively. This proof-of-concept study showed that spectral photon-counting CT could be used as a predictive preclinical imaging tool for the purpose of facilitating drug discovery and development. Also, as this imaging modality is available for human trials, all applications are translatable to human imaging. In conclusion, spectral photon-counting CT could accelerate a deeper understanding of infectious lung diseases using targeted pharmaceuticals and intrinsic markers, and ultimately improve the efficacy of therapies by measuring drug delivery and response to treatment in animal models and later in humans.
15 October 2020 in IEEE Transactions on Radiation and Plasma Medical Sciences
In this article, we propose a slice-based interactive segmentation of spectral CT datasets using a bag of features method. The data are acquired from a MARS scanner that divides up the X-ray spectrum into multiple energy bins for imaging. In literature, most existing segmentation methods are limited to performing a specific task or tied to a particular imaging modality. Therefore, when applying generalized methods to MARS datasets, the additional energy information acquired from the scanner cannot be sufficiently utilized. We describe a new approach that circumvents this problem by effectively aggregating the data from multiple channels. Our method solves a classification problem to get the solution for segmentation. Starting with a set of labeled pixels, we partition the data using superpixels. Then, a set of local descriptors, extracted from each superpixel, are encoded into a codebook and pooled together to create a global superpixel-level descriptor (bag of features representation). We propose to use the vector of locally aggregated descriptors as our encoding/pooling strategy, as it is efficient to compute and leads to good results with simple linear classifiers. A linear support vector machine is then used to classify the superpixels into different labels. The proposed method was evaluated on multiple MARS datasets. Experimental results show that our method achieved an average of more than 10% increase in the accuracy over other state-of-the-art methods.