Yu, Lifeng

PurposeFiltering measured projections with a particular convolutional kernel is an essential step in analytic reconstruction of computed tomography (CT) images. A tradeoff between noise and spatial resolution exists for different choices of reconstruction kernel. In a clinical setting, this often requires producing multiple images reconstructed with different kernels for a single CT exam, which increases the burden of computation, networking, archival, and reading. We address this problem by training a deep convolutional neural network (CNN) to synthesize multiple input images into a single output image which exhibits low noise while also preserving features in images reconstructed with the sharpest kernels.MethodsA CNN architecture consisting of repeated blocks of residual units containing a total of 20 convolutional layers was used to combine features. The CNN inputs consisted of two images produced with different reconstruction kernels, one smooth and one sharp, which were stacked in the channel dimension. The network was trained using supervised learning with both full‐dose and simulated quarter‐dose abdominal CT images. After training, the performance was evaluated using a reserved set of full‐dose scans that were not used for network optimization. Noise reduction performance was measured by comparing root mean square (RMS) measurements in uniform regions. Spatial resolution was compared using line profiles of anatomic features.ResultsFor the regions tested, the synthetic images feature noise levels slightly below those of the smooth input images, while maintaining the resolution of anatomic details found in the sharp input images.ConclusionsA deep CNN can be used combine features from CT images reconstructed with different kernels to produce a single synthesized image series that exhibits both low noise and high spatial resolution. This approach has implications for improving image quality, reducing radiation dose, and simplifying the clinical workflow for CT imaging.

PurposeTo evaluate the quantitative accuracy and dose efficiency of simultaneous imaging of two contrast agents using dual‐energy computed tomography (DECT), two imaging tasks each representing one potential clinical application were investigated in a phantom study: biphasic liver imaging with iodine and gadolinium, and small bowel imaging with iodine and bismuth.MethodsTo separate and quantify mixtures of two contrast agents using a single DECT scan, mixed iodine and gadolinium samples were prepared with the contrast enhancement values corresponding to the late arterial (iodine) and the portal‐venous (gadolinium) phase for biphasic liver imaging. Mixed iodine and bismuth samples were prepared mimicking the arterial (iodine) and the enteric (bismuth) enhancement for small bowel imaging. For comparison to the reference condition of performing two single‐energy CT (SECT) scans, contrast samples were prepared separately to mimic separate scans in the arterial/venous phase and arterial/enteric enhancement. Samples were placed in a 35 cm wide water tank and scanned using a third‐generation dual‐source DECT scanner with three tube potential pairs: 80/Sn150, 90/Sn150, and 100/Sn150 kV, all with default dose partitioning between two x‐ray beams to acquire DECT data. The same scanner operated in a single‐energy mode acquired SECT data (120 kV). Total radiation dose (CTDIvol) was matched for the single‐scan DECT and the two‐scan SECT protocols. The DECT protocol was followed by a generic image‐based three‐material decomposition method to determine the material‐specific images, based on which concentrations of each basis material were quantified and noise levels were measured. To compare with the SECT images directly acquired with the SECT protocol, the concentration values in each contrast‐specific image were converted to CT numbers at 120 kV (i.e., virtual SECT (vSECT) images). The noise level and noise power spectra differences between the SECT and vSECT images were compared to evaluate the dose efficiency of the single‐scan DECT protocol. The impact of dose partitioning in the DECT protocol on quantitative dual‐contrast imaging performance was also studied.ResultsFor each imaging task, contrast materials were accurately quantified against the nominal concentrations using the DECT data with strong correlation (R2 ≥ 0.98 for both imaging tasks). Compared to the SECT protocol, the DECT protocol was not dose efficient. With the optimal x‐ray tube potential pair 80/Sn150 kV, the noise level in vSECT images increased by 401%/488% (arterial/portal‐venous) for the biphasic liver imaging task and by 10%/41% (arterial/enteric) for the small bowel imaging task compared to that in SECT images. The corresponding radiation dose increase is 2410%/3357% for the biphasic liver imaging task and 21%/99% for the small bowel imaging task, respectively, to achieve the same noise as that in SECT images. This could be improved by adjusting the dose partitioning in DECT.ConclusionsDECT can be used to simultaneously separate and quantify two contrast materials. However, compared to a two‐scan SECT protocol, much higher radiation dose is needed in a single‐scan DECT protocol to achieve the same image noise, especially for tasks involving the dual contrast of iodine and gadolinium.

PurposeThe aim of this study was to demonstrate that a low helical pitch causes increased photon starvation artifacts at ultra‐low‐dose CT.MethodsA cylindrical water phantom with a diameter of 30 cm was scanned on two different generation CT scanners: a 64‐slice scanner (Sensation 64, Siemens Healthcare) and a 192‐slice scanner (Somatom Force, Siemens Healthcare) at multiple effective mAs levels (mAs/pitch = 200, 100, 50, 25, and 12). The corresponding CTDIvol values were 4.1, 2.0, 1.0, 0.5 mGy, on the 64‐slice scanner and 3.8, 1.9, 1.0, 0.5 mGy on the 192‐slice scanner, for the selected effective mAs values. For each dose setting, the scan was repeated at four helical pitches: 1.2, 0.9, 0.6, and the lowest achievable pitch on each scanner. The tube current was automatically adjusted by the scanner so that the effective mAs, and thus CTDIvol, were kept the same for different pitches. All CT data sets were reconstructed with a slice thickness of 3mm and a medium smooth kernel. Images acquired at the same dose level but different helical pitches were visually inspected to assess photon starvation artifacts and noise levels.ResultsAt the same radiation dose, image noise increased with the decreasing helical pitch. The increase was more severe on the old‐generation 64‐slice scanner. Photon starvation artifacts were evident at 200 effective mAs on the 64‐slice scanner at 80 kV. On the 192‐slice scanner there was no visible photon starvation artifacts at both 200 and 50 effective mAs (CTDIvol = 4.1 mGy and 1.0 mGy, respectively); nor was there a visible impact from the lower helical pitch. Only when the dose was lowered to be extremely low (~0.26 mGy, achievable at 70 kV), did photon starvation artifacts become evident.ConclusionsA low helical pitch may increase image noise and photon starvation artifacts compared to a higher pitch for the same dose level, particularly at ultra‐low dose CT.

PurposeThis work aims to develop a new framework of image quality assessment using deep learning‐based model observer (DL‐MO) and to validate it in a low‐contrast lesion detection task that involves CT images with patient anatomical background.MethodsThe DL‐MO was developed using the transfer learning strategy to incorporate a pretrained deep convolutional neural network (CNN), a partial least square regression discriminant analysis (PLS‐DA) model and an internal noise component. The CNN was previously trained to achieve the state‐of‐the‐art classification accuracy over a natural image database. The earlier layers of the CNN were used as a deep feature extractor, with the assumption that similarity exists between the CNN and the human visual system. The PLSR model was used to further engineer the deep feature for the lesion detection task in CT images. The internal noise component was incorporated to model the inefficiency and variability of human observer (HO) performance, and to generate the ultimate DL‐MO test statistics. Seven abdominal CT exams were retrospectively collected from the same type of CT scanners. To compare DL‐MO with HO, 12 experimental conditions with varying lesion size, lesion contrast, radiation dose, and reconstruction types were generated, each condition with 154 trials. CT images of a real liver metastatic lesion were numerically modified to generate lesion models with four lesion sizes (5, 7, 9, and 11 mm) and three contrast levels (15, 20, and 25 HU). The lesions were inserted into patient liver images using a projection‐based method. A validated noise insertion tool was used to synthesize CT exams with 50% and 25% of routine radiation dose level. CT images were reconstructed using the weighted filtered back projection algorithm and an iterative reconstruction algorithm. Four medical physicists performed a two‐alternative forced choice (2AFC) detection task (with multislice scrolling viewing mode) on patient images across the 12 experimental conditions. DL‐MO was operated on the same datasets. Statistical analyses were performed to evaluate the correlation and agreement between DL‐MO and HO.ResultsA statistically significant positive correlation was observed between DL‐MO and HO for the 2AFC low‐contrast detection task that involves patient liver background. The corresponding Pearson product moment correlation coefficient was 0.986 [95% confidence interval (0.950, 0.996)]. Bland–Altman agreement analysis did not indicate statistically significant differences.ConclusionsThe proposed DL‐MO is highly correlated with HO in a low‐contrast detection task that involves realistic patient liver background. This study demonstrated the potential of the proposed DL‐MO to assess image quality directly based on patient images in realistic, clinically relevant CT tasks.

AbstractObjectivesBoth projection and dual‐energy (DE)‐based methods have been used for metal artifact reduction (MAR) in CT. The two methods can also be combined. The purpose of this work was to evaluate these three MAR methods using phantom experiments for five types of metal implants.Materials and MethodsFive phantoms representing spine, dental, hip, shoulder, and knee were constructed with metal implants. These phantoms were scanned using both single‐energy (SE) and DE protocols with matched radiation output. The SE data were processed using a projection‐based MAR (iMAR, Siemens) algorithm, while the DE data were processed to generate virtual monochromatic images at high keV (Mono+, Siemens). In addition, the DE images after iMAR were used to generate Mono+ images (DE iMAR Mono+). Artifacts were quantitatively evaluated using CT numbers at different regions of interest. Iodine contrast‐to‐noise ratio (CNR) was evaluated in the spine phantom. Three musculoskeletal radiologists and two neuro‐radiologists independently ranked the artifact reduction.ResultsThe DE Mono+ at high keV resulted in reduced artifacts but also lower iodine CNR. The iMAR method alone caused missing tissue artifacts in dental phantom. DE iMAR Mono+ caused wrong CT numbers in close proximity to the metal prostheses in knee and hip phantoms. All musculoskeletal radiologists ranked SE iMAR > DE iMAR Mono+ > DE Mono+ for knee and hip, while DE iMAR Mono+ > SE iMAR > DE Mono+ for shoulder. Both neuro‐radiologists ranked DE iMAR Mono+ > DE Mono+ > SE iMAR for spine and DE Mono+ > DE iMAR Mono+ > SE iMAR for dental.ConclusionsThe SE iMAR was the best choice for the hip and knee prostheses, while DE Mono+ at high keV was best for dental implants and DE iMAR Mono+ was best for spine and shoulder prostheses. Artifacts were also introduced by MAR algorithms.