Yu, Lifeng

Abstract Photon counting detector (PCD) CT represents the newest advance in CT technology, with improved radiation dose efficiency, increased spatial resolution, inherent spectral imaging capabilities, and the ability to eliminate electronic noise. Its design fundamentally differs from conventional energy integrating detector CT because photons are directly converted to electrical signal in a single step. Rather than converting X-rays to visible light and having an output signal that is a summation of energies, PCD directly counts each photon and records its individual energy information. The current commercially available PCD-CT utilizes a dual-source CT geometry, which allows 66 ms cardiac temporal resolution and high-pitch (up to 3.2) scanning. This can greatly benefit pediatric patients by facilitating high quality fast scanning to allow sedation-free imaging. The energy-resolving nature of the utilized PCDs allows “always-on” dual-energy imaging capabilities, such as the creation of virtual monoenergetic, virtual non-contrast, virtual non-calcium, and other material-specific images. These features may be combined with high-resolution imaging, made possible by the decreased size of individual detector elements and the absence of interelement septa. This work reviews the foundational concepts associated with PCD-CT and presents examples to highlight the benefits of PCD-CT in the pediatric population.

We present an updated ultrafast dynamic computed tomography myelography technique that can be used for the localization of spontaneous spinal cerebrospinal fluid leaks in the setting of spontaneous intracranial hypotension. This has over twice the temporal resolution of previously described techniques at the same radiation dose output.

Background and Purpose Recent introduction of photon counting detector (PCD) computed tomography (CT) scanners into clinical practice further improve CT angiography (CTA) depiction of orbital arterial vasculature compared to conventional energy integrating detector (EID) CT scanners. PCD-CTA of the orbit can provide a detailed arterial roadmap of the orbit which can de diagnostic on its own or serve as a helpful planning adjunct for both diagnostic and therapeutic catheter-based angiography of the orbit. Methods For this review, EID and PCD-CT imaging was obtained in 28 volunteers. The volume CT dose index was closely matched. A dual-energy scanning protocol was used on EID-CT. An ultra-high-resolution (UHR) scan mode was used on PCD-CT. Images were reconstructed at 0.6 mm slice thickness using a closely matched medium-sharp standard resolution (SR) kernel. High-resolution (HR) images with the sharpest quantitative kernel were also reconstructed on PCD-CT at the thinnest slice thickness of 0.2 mm. A denoising algorithm was applied to the HR image series. Results The imaging description of the orbital vascular anatomy presented in this work was derived from these patients’ PCD-CTA images in combination with review of the literature. We found that orbital arterial anatomy is much better depicted with PCD-CTA, and this work can serve primarily as an imaging atlas of the normal orbital vascular anatomy. Conclusion With recent advances in technology, arterial anatomy of the orbit is much better depicted with PCD-CTA as opposed to EID-CTA. Current orbital PCD-CTA technology approaches the necessary resolution threshold for reliable evaluation of central retinal artery occlusion.

Objective Noise quantification is fundamental to computed tomography (CT) image quality assessment and protocol optimization. This study proposes a deep learning–based framework, Single-scan Image Local Variance EstimatoR (SILVER), for estimating the local noise level within each region of a CT image. The local noise level will be referred to as a pixel-wise noise map. Methods The SILVER architecture resembled a U-Net convolutional neural network with mean-square-error loss. To generate training data, 100 replicate scans were acquired of 3 anthropomorphic phantoms (chest, head, and pelvis) using a sequential scan mode; 120,000 phantom images were allocated into training, validation, and testing data sets. Pixel-wise noise maps were calculated for the phantom data by taking the per-pixel SD from the 100 replicate scans. For training, the convolutional neural network inputs consisted of phantom CT image patches, and the training targets consisted of the corresponding calculated pixel-wise noise maps. Following training, SILVER noise maps were evaluated using phantom and patient images. For evaluation on patient images, SILVER noise maps were compared with manual noise measurements at the heart, aorta, liver, spleen, and fat. Results When tested on phantom images, the SILVER noise map prediction closely matched the calculated noise map target (root mean square error <8 Hounsfield units). Within 10 patient examinations, SILVER noise map had an average percent error of 5% relative to manual region-of-interest measurements. Conclusion The SILVER framework enabled accurate pixel-wise noise level estimation directly from patient images. This method is widely accessible because it operates in the image domain and requires only phantom data for training.

AbstractPurposeTo determine the suitability of a quality assurance (QA) program based on the American College of Radiology's (ACR) CT quality control (QC) manual to fully evaluate the unique capabilities of a clinical photon‐counting‐detector (PCD) CT system.MethodsA daily QA program was established to evaluate CT number accuracy and artifacts for both standard and ultra‐high‐resolution (UHR) scan modes. A complete system performance evaluation was conducted in accordance with the ACR CT QC manual by scanning the CT Accreditation Phantom with routine clinical protocols and reconstructing low‐energy‐threshold (T3D) and virtual monoenergetic images (VMIs) between 40 and 120 keV. Spatial resolution was evaluated by computing the modulation transfer function (MTF) for the UHR mode, and multi‐energy performance was evaluated by scanning a body phantom containing four iodine inserts with concentrations between 2 and 15 mg I/cc.ResultsThe daily QA program identified instances when the detector needed recalibration or replacement. CT number accuracy was impacted by image type: CT numbers at 70 keV VMI were within the acceptable range (defined for 120 kV). Other keV VMIs and the T3D reconstruction had at least one insert with CT number outside the acceptable range. The limiting resolution was nearly 40 lp/cm based on MTF measurements, which far exceeds the 12 lp/cm maximum capability of the ACR phantom. The CT numbers in the iodine inserts were accurate on all VMIs (3.8% average percentage error), while the iodine concentrations had an average root mean squared error of 0.3 mg I/cc.ConclusionProtocols and parameters must be properly selected on PCD‐CT to meet current accreditation requirements with the ACR CT phantom. Use of the 70 keV VMI allowed passing all tests prescribed in the ACR CT manual. Additional evaluations such an MTF measurement and multi‐energy phantom scans are also recommended to comprehensively evaluate PCD‐CT scanner performance.

AbstractBackgroundIt is challenging to assess the accuracy of volume CT Dose Index (CTDIvol) when the axial scan modes corresponding to a helical scan protocol are not available. An alternative approach was proposed to directly measure using helical acquisitions and relatively small differences (< 20%) from CTDIvol were observed.PurposeTo visually demonstrate the 3D dose distribution for both axial and helical CT acquisitions and quantitively compare and CTDIvol.Methods3D dose distribution within the standard CTDI phantoms (16 and 32 cm diameter) from a single CT projection, Dp(x,y,z) was first generated using Monte Carlo simulation (GEANT4) with 9×108 photons per combination of tube voltage (80–140 kV), collimation width (1–8 cm), and z‐axis location of the central ray of the x‐ray beam, with a spatial resolution of 1 mm3. These dose distributions from one single projection were analytically ensembled to simulate 3D dose volumes DA(x,y,z) and DH(x,y,z) for axial and helical scans, respectively, with different helical pitches (0.3–2) and scan lengths (100–150 mm). 2D planar dose distributions were obtained by integrating the inside 100 mm of the dose volumes. CTDIvol and were calculated using the planar dose data at corresponding pencil chamber locations and the percentage differences (PD) were reported.ResultsHigh spatial resolution 3D CT dose volumes were generated and visualized. PDs between and CTDIvol had strong dependency on scan length and peripheral chamber locations, with subtle dependency on collimation width and pitch. PDs were mostly within the range of ± 3% for a scan length of 150 mm with four peripheral chamber locations.ConclusionsWith a scan length covering the entire phantom length, directly measured from helical scans can serve as an alternative to CTDIvol only if all four peripheral locations were measured.

AbstractBackgroundPhoton‐counting‐detector CT (PCD‐CT) enables the production of virtual monoenergetic images (VMIs) at a high spatial resolution (HR) via simultaneous acquisition of multi‐energy data. However, noise levels in these HR VMIs are markedly increased.PurposeTo develop a deep learning technique that utilizes a lower noise VMI as prior information to reduce image noise in HR, PCD‐CT coronary CT angiography (CTA).MethodsCoronary CTA exams of 10 patients were acquired using PCD‐CT (NAEOTOM Alpha, Siemens Healthineers). A prior‐information‐enabled neural network (Pie‐Net) was developed, treating one lower‐noise VMI (e.g., 70 keV) as a prior input and one noisy VMI (e.g., 50 keV or 100 keV) as another. For data preprocessing, noisy VMIs were reconstructed by filtered back‐projection (FBP) and iterative reconstruction (IR), which were then subtracted to generate “noise‐only” images. Spatial decoupling was applied to the noise‐only images to mitigate overfitting and improve randomization. Thicker slice averaging was used for the IR and prior images. The final training inputs for the convolutional neural network (CNN) inside the Pie‐Net consisted of thicker‐slice signal images with the reinsertion of spatially decoupled noise‐only images and the thicker‐slice prior images. The CNN training labels consisted of the corresponding thicker‐slice label images without noise insertion. Pie‐Net's performance was evaluated in terms of image noise, spatial detail preservation, and quantitative accuracy, and compared to a U‐net‐based method that did not include prior information.ResultsPie‐Net provided strong noise reduction, by 95 ± 1% relative to FBP and by 60 ± 8% relative to IR. For HR VMIs at different keV (e.g., 50 keV or 100 keV), Pie‐Net maintained spatial and spectral fidelity. The inclusion of prior information from the PCD‐CT data in the spectral domain was able to improve a robust deep learning‐based denoising performance compared to the U‐net‐based method, which caused some loss of spatial detail and introduced some artifacts.ConclusionThe proposed Pie‐Net achieved substantial noise reduction while preserving HR VMI's spatial and spectral properties.

AbstractBackgroundDeep artificial neural networks such as convolutional neural networks (CNNs) have been shown to be effective models for reducing noise in CT images while preserving anatomic details. A practical bottleneck for developing CNN‐based denoising models is the procurement of training data consisting of paired examples of high‐noise and low‐noise CT images. Obtaining these paired data are not practical in a clinical setting where the raw projection data is not available. This work outlines a technique to optimize CNN denoising models using methods that are available in a routine clinical setting.PurposeTo demonstrate a phantom‐based training framework for CNN noise reduction that can be efficiently implemented on any CT scanner.MethodsThe phantom‐based training framework uses supervised learning in which training data are synthesized using an image‐based noise insertion technique. Ten patient image series were used for training and validation (9:1) and noise‐only images obtained from anthropomorphic phantom scans. Phantom noise‐only images were superimposed on patient images to imitate low‐dose CT images for use in training. A modified U‐Net architecture was used with mean‐squared‐error and feature reconstruction loss. The training framework was tested for clinically indicated whole‐body‐low‐dose CT images, as well as routine abdomen‐pelvis exams for which projection data was unavailable. Performance was assessed based on root‐mean‐square error, structural similarity, line profiles, and visual assessment.ResultsWhen the CNN was tested on five reserved quarter‐dose whole‐body‐low‐dose CT images, noise was reduced by 75%, root‐mean‐square‐error reduced by 34%, and structural similarity increased by 60%. Visual analysis and line profiles indicated that the method significantly reduced noise while maintaining spatial resolution of anatomic features.ConclusionThe proposed phantom‐based training framework demonstrated strong noise reduction while preserving spatial detail. Because this method is based within the image domain, it can be easily implemented without access to projection data.