Radiation dose reduction using deep learning-based image reconstruction for a low-dose chest computed tomography protocol: a phantom study.

Background: The aim of this study was to compare the dose reduction potential and image quality of deep learning-based image reconstruction (DLIR) with those of filtered back-projection (FBP) and iterative reconstruction (IR) and to determine the clinically usable dose of DLIR for low-dose chest computed tomography (LDCT) scans. Methods: Multi-slice computed tomography (CT) scans of a chest phantom were performed with various tube voltages and tube currents, and the images were reconstructed using seven methods to control the amount of noise reduction: FBP, three stages of IR, and three stages of DLIR. For subjective image analysis, four radiologists compared 48 image data sets with reference images and rated on a 5-point scale. For quantitative image analysis, the signal to noise ratio (SNR), contrast to noise ratio (CNR), nodule volume, and nodule diameter were measured. Results: In the subjective analysis, DLIR-Low (0.46 mGy), DLIR-Medium (0.31 mGy), and DLIR-High (0.18 mGy) images showed similar quality to the FBP (2.47 mGy) image. Under the same dose conditions, the SNR and CNR were higher with DLIR-High than with FBP and all the IR methods (all P<0.05). The nodule volume and size with DLIR-High were significantly closer to the real volume than with FBP and all the IR methods (all P<0.001). Conclusions: DLIR can improve the image quality of LDCT compared to FBP and IR. In addition, the appropriate effective dose for LDCT would be 0.24 mGy with DLIR-High.

Pseudo low-energy monochromatic imaging of head and neck cancers: Deep learning image reconstruction with dual-energy CT.

PURPOSE: Low-energy virtual monochromatic images (VMIs) derived from dual-energy computed tomography (DECT) systems improve lesion conspicuity of head and neck cancer over single-energy CT (SECT). However, DECT systems are installed in a limited number of facilities; thus, only a few facilities benefit from VMIs. In this work, we present a deep learning (DL) architecture suitable for generating pseudo low-energy VMIs of head and neck cancers for facilities that employ SECT imaging. METHODS: We retrospectively analyzed 115 patients with head and neck cancers who underwent contrast enhanced DECT. VMIs at 70 and 50 keV were used as the input and ground truth (GT), respectively. We divided them into two datasets: for DL (104 patients) and for inference with SECT (11 patients). We compared four DL architectures: U-Net, DenseNet-based, and two ResNet-based models. Pseudo VMIs at 50 keV (pVMI50keV) were compared with the GT in terms of the mean absolute error (MAE) of Hounsfield unit (HU) values, peak signal-to-noise ratio (PSNR), and structural similarity (SSIM). The HU values for tumors, vessels, parotid glands, muscle, fat, and bone were evaluated. pVMI50keV were generated from actual SECT images and the HU values were evaluated. RESULTS: U-Net produced the lowest MAE (13.32 ± 2.20 HU) and highest PSNR (47.03 ± 2.33 dB) and SSIM (0.9965 ± 0.0009), with statistically significant differences (P < 0.001). The HU evaluation showed good agreement between the GT and U-Net. U-Net produced the smallest absolute HU difference for the tumor, at < 5.0 HU. CONCLUSION: Quantitative comparisons of physical parameters demonstrated that the proposed U-Net could generate high accuracy pVMI50keV in a shorter time compared with the established DL architectures. Although further evaluation on diagnostic accuracy is required, our method can help obtain low-energy VMI from SECT images without DECT systems.

Evaluation of Stopping Power Ratio Calculation Using Dual-energy Computed Tomography With Fast Kilovoltage Switching for Treatment Planning of Particle Therapy.

BACKGROUND/AIM: This study evaluated the calculation accuracy of the stopping power ratio (SPR) using dual-energy computed tomography with fast kilovoltage switching (FKSCT) for particle therapy. MATERIALS AND METHODS: A tissue characterization phantom with various reference materials was scanned to obtain single-energy computed tomography (SECT) images and generate virtual monochromatic images at 77 keV (VMI77keV) and 140 keV (VMI140keV), water density (WD) images, and effective Z (Zeff) images. For SECT, VMI77keV and VMI140keV lookup tables were generated to convert the measured Hounsfield value into the theoretical SPR for a normal phantom size. Subsequently, the reference materials were scanned in small and large phantoms. The SPR was calculated using the lookup tables of SECT (SPRSECT) images, VMI77keV (SPR77keV), and VMI140keV (SPR140keV), and it was derived from the WD and Zeff (SPRWD). RESULTS: In the normal-sized phantom, the overall mean difference between SPRWD and theoretical SPR was -0.3%, and remained below 2% for most reference materials. For the large phantom, the overall mean absolute difference for SPR140keV (3.0%, p=0.006) and SPRWD (3.2%, p=0.002) for the reference materials was significantly lower than that for SPRSECT (5.9%). For the small phantom, a significant reduction in the mean difference in the SPR calculation was observed in SPR77keV (1.0%, p=0.001) and SPR140keV (1.1%, p=0.013) compared with SPRSECT (2.2%). CONCLUSION: VMI140keV generated using FKSCT significantly improves the estimation accuracy of SPR compared with SECT. Thus, FKSCT may be used to improve the dose calculation accuracy for treatment planning of particle therapy.

Technical Note: Evaluation of a 160-mm/256-row CT scanner for whole-heart quantitative myocardial perfusion imaging.

PURPOSE: The authors investigated the performance of a recently introduced 160-mm/256-row CT system for low dose quantitative myocardial perfusion (MP) imaging of the whole heart. This platform is equipped with a gantry capable of rotating at 280 ms per full cycle, a second generation of adaptive statistical iterative reconstruction (ASiR-V) to correct for image noise arising from low tube voltage potential/tube current dynamic scanning, and image reconstruction algorithms to tackle beam-hardening, cone-beam, and partial-scan effects. METHODS: Phantom studies were performed to investigate the effectiveness of image noise and artifact reduction with a GE Healthcare Revolution CT system for three acquisition protocols used in quantitative CT MP imaging: 100, 120, and 140 kVp/25 mAs. The heart chambers of an anthropomorphic chest phantom were filled with iodinated contrast solution at different concentrations (contrast levels) to simulate the circulation of contrast through the heart in quantitative CT MP imaging. To evaluate beam-hardening correction, the phantom was scanned at each contrast level to measure the changes in CT number (in Hounsfield unit or HU) in the water-filled region surrounding the heart chambers with respect to baseline. To evaluate cone-beam artifact correction, differences in mean water HU between the central and peripheral slices were compared. Partial-scan artifact correction was evaluated from the fluctuation of mean water HU in successive partial scans. To evaluate image noise reduction, a small hollow region adjacent to the heart chambers was filled with diluted contrast, and contrast-to-noise ratio in the region before and after noise correction with ASiR-V was compared. The quality of MP maps acquired with the CT system was also evaluated in porcine CT MP studies. Myocardial infarct was induced in a farm pig from a transient occlusion of the distal left anterior descending (LAD) artery with a catheter-based interventional procedure. MP maps were generated from the dynamic contrast-enhanced (DCE) heart images taken at baseline and three weeks after the ischemic insult. RESULTS: Their results showed that the phantom and animal images acquired with the CT platform were minimally affected by image noise and artifacts. For the beam-hardening phantom study, changes in water HU in the wall surrounding the heart chambers greatly reduced from >±30 to ≤ ± 5 HU at all kVp settings except one region at 100 kVp (7 HU). For the cone-beam phantom study, differences in mean water HU from the central slice were less than 5 HU at two peripheral slices with each 4 cm away from the central slice. These findings were reproducible in the pig DCE images at two peripheral slices that were 6 cm away from the central slice. For the partial-scan phantom study, standard deviations of the mean water HU in 10 successive partial scans were less than 5 HU at the central slice. Similar observations were made in the pig DCE images at two peripheral slices with each 6 cm away from the central slice. For the image noise phantom study, CNRs in the ASiR-V images were statistically higher (p < 0.05) than the non-ASiR-V images at all kVp settings. MP maps generated from the porcine DCE images were in excellent quality, with the ischemia in the LAD territory clearly seen in the three orthogonal views. CONCLUSIONS: The study demonstrates that this CT system can provide accurate and reproducible CT numbers during cardiac gated acquisitions across a wide axial field of view. This CT number fidelity will enable this imaging tool to assess contrast enhancement, potentially providing valuable added information beyond anatomic evaluation of coronary stenoses. Furthermore, their results collectively suggested that the 100 kVp/25 mAs protocol run on this CT system provides sufficient image accuracy at a low radiation dose (<3 mSv) for whole-heart quantitative CT MP imaging.

Dual-energy CT and its potential use for quantitative myocardial CT perfusion.

Application of quantitative myocardial CT perfusion (CTP) for the assessment of coronary artery disease may have a significant effect on patient care as the functional significance of a coronary stenosis can be evaluated through absolute measurement of the downstream myocardial perfusion (MP) both at rest and under exercise or pharmacologic stress. A main challenge of myocardial CTP is beam hardening (BH), arising from the polychromatic nature of x-rays used in CT scanning and the presence of highly attenuating contrast agent in the heart chambers during the CT acquisition. The BH effect induces significant nonuniform shifts in CT numbers which, if uncorrected, can lead to inaccurate assessment of MP. With the recent developments of dual-energy CT (DECT) scanning on clinical scanners, the BH effect on MP measurement could be reduced with the generation of monochromatic images relatively free of BH artifacts from the acquired dual-energy data. Here, we review the different techniques of acquiring dual-energy scans and generating monochromatic images, followed by discussion on the progress of developing a DECT technique with reduced radiation dose for quantitative myocardial CTP.

Prospectively ECG-triggered rapid kV-switching dual-energy CT for quantitative imaging of myocardial perfusion.

Dual-energy computed tomography (DECT) has recently been introduced for clinical use. One potential application of DECT is myocardial perfusion imaging through the significant reduction of beam-hardening artifacts by using monochromatic image reconstruction; analysis of these images can improve the accuracy of quantitative measurement of myocardial perfusion. Single-source DECT enabled by rapid switching between the low and high tube potentials (kV) can minimize misregistration of the high and low kV projection datasets from cardiac motion. We have recently implemented prospective electrocardiography-triggering capability in our rapid kV-switching computed tomography (CT) scanner to reduce the high effective dose from a quantitative CT myocardial perfusion imaging study with DECT. Our initial investigation suggests that prospectively electrocardiography-triggered rapid kV-switching DECT can eliminate beam hardening and provide a more reproducible myocardial perfusion measurement compared with the traditional single-energy CT protocol.

Improvement of in-stent lumen measurement accuracy with new high-definition CT in a phantom model: comparison with conventional 64-detector row CT.

The purpose of this study was to evaluate improvement of measurement accuracy of in-stent lumen using coronary stent phantoms on new high-definition CT (HDCT) compared with conventional 64 detector-row CT (MDCT). To estimate the spatial resolution, a high-resolution insert of CATPHAN (The Phantom Laboratory, NY, USA) was scanned by both HDCT (Discovery CT750 HD) and MDCT (LightSpeed VCT). Also, we developed six types of stent phantom, which have 2.5- and 3.0-mm-diameter with three different types of stents (Velocity: Johnson & Johnson, Driver: Medtronic, Multilink-Rx: Guidant). A 50% stenotic segment made of acrylic resin was built at the center inside the stent. Those coronary vessel phantoms were made of acrylic resin and filled with diluted Iodine (350 HU in 120 kVp), and each stent was fixed inside of those vessels. Those phantoms in water-filled tank were scanned on both HDCT and MDCT. The luminal diameter obtained using digital calipers at five different points and the mean luminal diameter (MLD) were calculated. The underestimate ratio (UR) and △UR was defined as follows: UR = [True diameter of stent-MLD]/True diameter of stent; △UR = [MLD at HDCT-MLD at MDCT]/True diameter of stent. The spatial resolution was estimated to be 0.71 mm on MDCT and 0.50 mm on HDCT. At the non-stenotic segments, the △URs were 11.6% (Velocity), 16.4% (Driver) and 7.2% (Multilink) for the 2.5-mm stents, and 14.0% (Velocity), 16.3% (Driver) and 13.3% (Multilink) for the 3.0-mm stents. At the stenotic segment, the △URs were 23.2% (Velocity), 8.0% (Driver) and 13.6% (Multilink) for the 2.5-mm stents, and 20.0% (Velocity), 14.7% (Driver) and 15.3% (Multilink) for the 3.0-mm stents. Superior spatial resolution of HDCT could be promising for more accurate measurement of in-stent diameter.

Quantitative myocardial perfusion imaging using rapid kVp switch dual-energy CT: preliminary experience.

BACKGROUND: Quantitative myocardial CT perfusion (CTP) is susceptible to beam-hardening (BH) artifact from conventional single-energy (kVp) CT (SECT) scanning, which can mimic perfusion deficits. OBJECTIVE: We evaluated the minimization of BH artifact with dual-energy (kVp) CT (DECT) generated monochromatic CT images to improve perfusion estimates. METHODS: We investigated the performance of DECT with a scanner capable of rapid kVp switching with respect to (1) BH artifact in a myocardium phantom model comparing SECT with image-based DECT and projection-based DECT, (2) optimal imaging parameters for measuring iodine concentration at high contrast-to-noise ratio in a tissue characterization phantom model, and (3) the feasibility of a dynamic time-resolved scan protocol with the projection-based DECT technique to measure myocardial perfusion in normal (nonischemic) porcine. RESULTS: In a myocardium phantom model, projection-based DECT 70 keV was better able to minimize the difference in the attenuation of the myocardium (19.9 HU) between having and not having contrast in the heart chambers in comparison to SECT using 80 kVp (30.4 HU) or 140 kVp ( 23.3 HU) and image-based DECT 70 keV (27.5 HU). Further, projection-based DECT 70 keV achieved the highest contrast-to-noise ratio (3.0), which exceeded that from imaged-based DECT 70 keV (2.0), 140 kVp SECT (1.3), and 80 kVp SECT (2.9). In 5 normal pigs, projection-based DECT at 70 keV provided a more uniform perfusion estimate than SECT. CONCLUSION: By effectively reducing BH artifact, projection-based DECT may permit improved quantitative myocardial CTP compared with the conventional SECT technique.

Detection of a coronary artery vessel wall: performance of 0.3 mm fine-cell detector computed tomography--a phantom study.

The purpose of this study was to evaluate whether experimental fine-cell detector computed tomography with a 0.3125 mm cell (0.3 mm cell CT) can improve the detection of coronary vessel walls compared with conventional 64-slice computed tomography with a 0.625 mm cell (0.6 mm cell CT). A coronary vessel wall phantom was scanned using 0.6 mm cell CT and 0.3 mm cell CT. The data for 0.3 mm cell CT were obtained using four protocols: a radiation dose equal, double, triple or quadruple that were used in the 0.6 mm cell CT protocol. The detectable size of the vessel wall was assessed based on the first and second derivative functions, and the minimum measurable values were compared using a paired t-test. As a result, the minimum detectable wall thickness of 0.6 mm cell CT (1.5 mm) was significantly larger than that of 0.3 mm cell CT performed using the triple- and quadruple-dose protocols (0.9 mm) and the double-dose protocol (1.1 mm). The difference in the minimum detectable vessel wall thickness measured using 0.6 mm cell CT (1.5 ± 0.1 mm) and 0.3 mm cell CT (0.9 ± 0.1 mm, 1.1 ± 0.2 mm) was significant (p < 0.01). We concluded that 0.3 mm cell CT improved the detection of coronary vessel walls when a more than double-dose protocol was used compared with 0.6 mm cell CT.

[Radiation dose reduction using a non-linear image filter in MDCT].

The development of MDCT enabled various high-quality 3D imaging and optimized scan timing with contrast injection in a multi-phase dynamic study. Since radiation dose tends to increase to yield such high-quality images, we have to make an effort to reduce radiation dose. A non-linear image filter (Neuro 3D filter: N3D filter) has been developed in order to improve image noise. The purpose of this study was to evaluate the physical performance and effectiveness of this non-linear image filter using phantoms, and show how we can reduce radiation dose in clinical use of this filter. This N3D filter reduced radiation dose by about 50%, with minimum deterioration of spatial reduction in phantom and clinical studies. This filter shows great potential for clinical application.

Development and performance evaluation of an experimental fine pitch detector multislice CT scanner.

The authors have developed an experimental fine pitch detector multislice CT scanner with an ultrasmall focal spot x-ray tube and a high-density matrix detector through current CT technology. The latitudinal size of the x-ray tube focal spot was 0.4 mm. The detector dimension was 1824 channels (azimuthal direction) x 32 rows (longitudinal direction) at row width of 0.3125 mm, in which a thinner reflected separator surrounds each detector cell coupled with a large active area photodiode. They were mounted on a commercial 64-slice CT scanner gantry while the scan field of view (50 cm) and gantry rotation speed (0.35 s) can be maintained. The experimental CT scanner demonstrated the spatial resolution of 0.21-0.22 mm (23.8-22.7 lp/cm) with the acrylic slit phantom and in-plane 50%-MTF 9.0 lp/cm and 10%-MTF 22.0 lp/cm. In the longitudinal direction, it demonstrated the spatial resolution of 0.24 mm with the high-resolution insert of the CATPHAN phantom and 0.34 mm as the full width at half maximum of the slice sensitivity profile. In low-contrast detectability, 3 mm at 0.3% was visualized at the CTDI(vol) of 47.2 mGy. Two types of 2.75 mm diameter vessel phantoms with in-stent stenosis at 25%, 50%, and 75% stair steps were scanned, and the reconstructed images can clearly resolve the stenosis at each case. The experimental CT scanner provides high-resolution imaging while maintaining low-contrast detectability, demonstrating the potentiality for clinical applications demanding high spatial resolution, such as imaging of inner ear, lung, and bone, or low-contrast detectability, such as imaging of coronary artery.

[Exposed dose comparison between coronary computed tomographic angiography and coronary angiography: basic examination by phantom].

The new type of coronary angiography(CAG)that uses 40 mm volumetric computed tomography(VCT)has great potential for cardiac disease. However, it is still necessary to be cognizant of exposure dose. We measured doses of CAG by both VCT and cardiovascular X-ray using a body phantom within 170 glass dosimeters. VCT protocols were 120 kV, 570 mA, and 0.35 sec/rot with and without the dose-reduction features(small cardiac X-ray beam filter and ECG mA modulation). The cardiovascular X-ray protocol was Auto(65 - 77 kV)kV, Auto(41 - 46 mA)mA, 5 secx11 shots+11 min fluoroscopy(minimum protocol for screening). VCT with and without the dose-reduction features has the same dose distribution, however, the dose-reduction features reduced the amount of dose by about 40-50%. For VCT with those features, measured dose was about 70 mGy in the cardiac area and 60 mGy at the skin of the back, whereas those of cardiovascular X-ray were 10 mGy and 30 mGy. We measured detailed dose distributions and variations in the phantom, and we also demonstrated the possibility of VCT's dose-reduction features. The CT dose was still higher than that of cardiovascular X-ray, however, there were advantages of CT scanning, for instance, information about calcification, soft plaque, and 3D visualization. We think it is important to use both systems with an understanding of their advantages and limitations.

[Evaluation of ultra-low-dose thoracic multi-detector-row CT using different reconstruction kernels and new reconstruction algorithms].

CT has great advantages in detecting early-stage small lung cancer and is becoming common in lung cancer screening. Multi-detector-row CT (MDCT) can provide thin-slice images with low radiation exposure. In this study, ultra-low-dose (5 mAs: 10 mAs, 0.5 sec/rot) thoracic MDCT images were evaluated. We describe the differences in image quality and quantity between the different reconstruction kernels. We also propose a new reconstruction algorithm (ultra-low-dose reconstruction algorithm: ULR) for ultra-low-dose thoracic CT, to reduce noise and streak artifacts. We are convinced of the usefulness and possibility of ultra-low-dose thoracic MDCT with ULR algorithms for lung cancer screening.