Learned Full-Sampling Reconstruction

Objective. X-ray computed tomography employing sparse projection views has emerged as a contemporary technique to mitigate radiation dose. However, due to the inadequate number of projection views, an analytic reconstruction method utilizing filtered backprojection results in severe streaking artifacts. Recently, deep learning (DL) strategies employing image-domain networks have demonstrated remarkable performance in eliminating the streaking artifact caused by analytic reconstruction methods with sparse projection views. Nevertheless, it is difficult to clarify the theoretical justification for applying DL to sparse view computed tomography (CT) reconstruction, and it has been understood as restoration by removing image artifacts, not reconstruction. Approach. By leveraging the theory of deep convolutional framelets (DCF) and the hierarchical decomposition of measurement, this research reveals the constraints of conventional image and projection-domain DL methodologies, subsequently, the research proposes a novel dual-domain DL framework utilizing hierarchical decomposed measurements. Specifically, the research elucidates how the performance of the projection-domain network can be enhanced through a low-rank property of DCF and a bowtie support of hierarchical decomposed measurement in the Fourier domain. Main results. This study demonstrated performance improvement of the proposed framework based on the low-rank property, resulting in superior reconstruction performance compared to conventional analytic and DL methods. Significance. By providing a theoretically justified DL approach for sparse-view CT reconstruction, this study not only offers a superior alternative to existing methods but also opens new avenues for research in medical imaging. It highlights the potential of dual-domain DL frameworks to achieve high-quality reconstructions with lower radiation doses, thereby advancing the field towards safer and more efficient diagnostic techniques. The code is available at https://github.com/hanyoseob/HDD-DL-for-SVCT.

Deep-learning-based image reconstruction and noise reduction methods (DLIR) have been increasingly deployed in clinical CT. Accurate image quality assessment of these methods is challenging as the performance measured using physical phantoms may not represent the true performance of DLIR in patients since DLIR is trained mostly on patient images. In this work, we aim to develop a patient-data-based virtual imaging trial framework and, as a first application, use it to measure the spatial resolution properties of a DLIR method. The patient-data-based virtual imaging trial framework consists of five steps: (1) insertion of lesions into projection domain data using the acquisition geometry of the patient exam to simulate different lesion characteristics; (2) insertion of noise into projection domain data using a realistic photon statistical model of the CT system to simulate different dose levels; (3) creation of DLIR-processed images from projection or image data; (4) creation of ensembles of DLIR-processed patient images from a large number of noise and lesion realizations; and (5) evaluation of image quality using ensemble DLIR images. This framework was applied to measure the spatial resolution of a ResNet based deep convolutional neural network (DCNN) trained on patient images. Lesions in a cylindrical shape and different contrast levels (-500, -100, -50, -20, -10 HU) were inserted to the lower right lobe of the liver in a patient case. Multiple dose levels were simulated (50%, 25%, 12.5%). Each lesion and dose condition had 600 noise realizations. Multiple reconstruction and denoising methods were used on all the noise realizations, including the original filtered-backprojection (FBP), iterative reconstruction (IR), and the DCNN method with three different strength setting (DCNN-weak, DCNN-medium, and DCNN-strong). Mean lesion signal was calculated by performing ensemble averaging of all the noise realizations for each lesion and dose condition and then subtracting the lesion-present images from the lesion absent images. Modulation transfer functions (MTFs) both in-plane and along the z-axis were calculated based on the mean lesion signals. The standard deviations of MTFs at each condition were estimated with bootstrapping: randomly sampling (with replacement) all the DLIR/FBP/IR images from the ensemble data (600 samples) at each condition. The impact of varying lesion contrast, dose levels, and denoising strengths were evaluated. Statistical analysis with paired t-test was used to compare the z-axis and in-plane spatial resolution of five algorithms for five different contrasts and three dose levels. The in-plane and z-axis spatial resolution degradation of DCNN becomes more severe as the contrast or radiation dose decreased, or DCNN denoising strength increased. In comparison with FBP, a 59.5% and 4.1% reduction of in-plane and z-axis MTF (in terms of spatial frequencies at 50% MTF), respectively, was observed at low contrast (-10 HU) for DCNN with the highest denoising strength at 25% routine dose level. When the dose level reduces from 50% to 12.5% of routine dose, the in-plane and z-axis MTFs reduces from 92.1% to 76.3%, and from 98.9% to 95.5%, respectively, at contrast of -100 HU, using FBP as the reference. For most conditions of contrasts and dose levels, significant differences were found among the five algorithms, with the following relationship in both in-plane and cross-plane spatial resolution: FBP>DCNN-Weak>IR>DCNN-Medium>DCNN-Strong. The spatial resolution difference among algorithms decreases at higher contrast or dose levels. A patient-data-based virtual imaging trial framework was developed and applied to measuring the spatial resolution properties of a DCNN noise reduction method at different contrast and dose levels using real patient data. As with other non-linear image reconstruction and post-processing techniques, the evaluated DCNN method degraded the in-plane and z-axis spatial resolution at lower contrast levels, lower radiation dose, and higher denoising strength.

Sparse-view computed tomography (CT) has been attracting attention for its reduced radiation dose and scanning time. However, analytical image reconstruction methods suffer from streak artifacts due to insufficient projection views. Recently, various deep learning-based methods have been developed to solve this ill-posed inverse problem. Despite their promising results, they are easily overfitted to the training data, showing limited generalizability to unseen systems and patients. In this work, we propose a novel streak artifact reduction algorithm that provides a system- and patient-specificsolution. Motivated by the fact that streak artifacts are deterministic errors, we regenerate the same artifacts from a prior CT image under the same system geometry. This prior image need not be perfect but should contain patient-specific information and be consistent with full-view projection data for accurate regeneration of the artifacts. To this end, we use a coordinate-based neural representation that often causes image blur but can greatly suppress the streak artifacts while having multiview consistency. By employing techniques in neural radiance fields originally proposed for scene representations, the neural representation is optimized to the measured sparse-view projection data via self-supervised learning. Then, we subtract the regenerated artifacts from the analytically reconstructed original image to obtain the final correctedimage. To validate the proposed method, we used simulated data of extended cardiac-torso phantoms and the 2016 NIH-AAPM-Mayo Clinic Low-Dose CT Grand Challenge and experimental data of physical pediatric and head phantoms. The performance of the proposed method was compared with a total variation-based iterative reconstruction method, naive application of the neural representation, and a convolutional neural network-based method. In visual inspection, it was observed that the small anatomical features were best preserved by the proposed method. The proposed method also achieved the best scores in the visual information fidelity, modulation transfer function, and lung nodulesegmentation. The results on both simulated and experimental data suggest that the proposed method can effectively reduce the streak artifacts while preserving small anatomical structures that are easily blurred or replaced with misleading features by the existing methods. Since the proposed method does not require any additional training datasets, it would be useful in clinical practice where the large datasets cannot becollected.

This paper introduces a new accurate two-dimensional (2D) computer tomography reconstruction method, differentiated backprojection (DBP) and Hilbert filtering method. This method consists of two steps. First, form a backprojected image after taking the derivative of the parallel projection data. Second, apply Hilbert filtering along certain lines in the DBP image. Different from traditional filtered backprojection (FBP) algorithm, region of interest (ROI) can be reconstructed from truncated projection data in certain situation with this new method. A modified method is also presented. We carry out these three methods: DBP and Hilbert filtering method, its modified method, and FBP algorithm. Simulation results are presented that illustrate similar reconstructed image quality using this new method compared to FBP algorithm, and that show the capability to accurately reconstruct ROI from truncated projections.

Dose and image quality are two important parameters in CT Scan examination. The radiation dose produced at the CT scan is far greater than the radiation dose of the conventional radiological examination, so the utilization must be as good as possible. On Siemens CT Scan there is an automatic exposure control software, with expectations that the radiation dose received by the patient can be reduced because the software is an auto mA that is adjusted to the thickness of the object, but the impact of using the software causes a decrease in image quality, so it needs to be done image improvement using a back projection filter (BPF). Experimental research by applying a back projection filter to the Scan Head image on axial sections. Image assessment using 2 radiologists with data analysis using statistical processing and image post processing. CT Scan image information Head axial pieces experience differences in image information between before and after the back projection, with the mean value in each group experiencing an increase in image quality in the group after a back projection filter. Where the image after the back projection filter by applying automatic exposure control with kV 110 is the most optimal kV.

We evaluated the feasibility of sub-millisievert (mSv) coronary CT angiography (CCTA) using low tube voltage, prospective ECG gating, and a knowledge-based iterative model reconstruction algorithm. Twenty-four non-obese healthy subjects (M:F 13:11; mean age 50.2±7.8years) were enrolled. Three sets of CT images were reconstructed using three different reconstruction methods: filtered back projection (FBP), iterative reconstruction (IR), and knowledge-based iterative model reconstruction (IMR). The scanning parameters were as follows: step-and-shoot axial scanning, 80kVp, and 200mAs. On the three sets of CT images, the attenuation and image noise values were measured at the aortic root. The signal-to-noise ratio (SNR) and the contrast-to-noise ratio (CNR) were calculated at the proximal right coronary artery and the left main coronary artery. The qualitative image quality of the CCTA with IMR was assessed using a 4-point grading scale (grade 1, poor; grade 4, excellent). The mean radiation dose of the CCTA was 0.89±0.09mSv. The attenuation values with IMR were not different from those of other reconstruction methods. The image noise with IMR was significantly lower than with IR and FBP. Compared to FBP, the noise reduction rate of IMR was 69%. The SNR and CNR of CCTA with IMR were significantly higher than with FBP or IR. On the qualitative analysis with IMR, all included segments were diagnostic (grades 2, 3, and 4), and the mean image quality score was 3.6±0.6. In conclusion, CCTA with low tube voltage, prospective ECG gating, and an IMR algorithm might be a feasible method that allows for sub-millisievert radiation doses and good image quality when used with non-obese subjects.

<strong>Purpose:</strong> To investigate the dose saving potential of iterative reconstruction (IR) in a computed tomography (CT) examination of the thorax. <br/>
\n<strong>Materials and Methods:</strong> An anthropomorphic chest phantom containing various configurations of simulated lesions (5, 8, 10 and 12mm; +100, -630 and -800 Hounsfield Units, HU) was imaged on a modern CT system over a tube current range (20, 40, 60 and 80mA). Images were reconstructed with (IR) and filtered back projection (FBP). An ATOM 701D (CIRS, Norfolk, VA) dosimetry phantom was used to measure organ dose. Effective dose was calculated. Eleven observers (15.11±8.75 years of experience) completed a free response study, localizing lesions in 544 single CT image slices. A modified jackknife alternative free-response receiver operating characteristic (JAFROC) analysis was completed to look for a significant effect of two factors: reconstruction method and tube current. Alpha was set at 0.05 to control the Type I error in this study. <br/>
\n<strong>Results:</strong> For modified JAFROC analysis of reconstruction method there was no statistically significant difference in lesion detection performance between FBP and IR when figures-of-merit were averaged over tube current (F(1,10)=0.08, p = 0.789). For tube current analysis, significant differences were revealed between multiple pairs of tube current settings (F(3,10) = 16.96, p<0.001) when averaged over image reconstruction method.<br/>
\n<strong>Conclusion:</strong> The free-response study suggests that lesion detection can be optimized at 40mA in this phantom model, a measured effective dose of 0.97mSv. In high-contrast regions the diagnostic value of IR, compared to FBP, is less clear.

X-ray computed tomography (CT) using sparse projection views is a recent approach to reduce the radiation dose. However, due to the insufficient projection views, an analytic reconstruction approach using the filtered back projection (FBP) produces severe streaking artifacts. Recently, deep learning approaches using large receptive field neural networks such as U-Net have demonstrated impressive performance for sparse-view CT reconstruction. However, theoretical justification is still lacking. Inspired by the recent theory of deep convolutional framelets, the main goal of this paper is, therefore, to reveal the limitation of U-Net and propose new multi-resolution deep learning schemes. In particular, we show that the alternative U-Net variants such as dual frame and tight frame U-Nets satisfy the so-called frame condition which makes them better for effective recovery of high frequency edges in sparse-view CT. Using extensive experiments with real patient data set, we demonstrate that the new network architectures provide better reconstruction performance.

BackgroundThe arms-down position increases computed tomography (CT) radiation dose. Iterative reconstruction (IR) could enhance image quality without increasing radiation dose in patients with arms-down position.AimTo investigate the feasibility of reduced-dose CT with IR for patients with inappropriate arm positioningMethodsTwenty patients who underwent two-phase abdominopelvic CT including standard-dose and reduced-dose CT (performed with 80% of the radiation dose of the standard protocol) with their arms positioned in the abdominal area were included in this study. Reduced-dose CT images were reconstructed using filtered back projection (FBP), hybrid IR, and iterative model reconstruction (IMR). These images were compared with standard-dose CT images reconstructed with FBP. Objective image noise in the liver and subcutaneous fat was measured by standard deviation for the quantitative analysis. Then, two radiologists qualitatively assessed beam hardening artifacts, artificial texture, noise, sharpness, and overall image quality in consensus.ResultsReduced-dose CT with all IR levels had lower objective image noise compared to standard-dose CT with FBP reconstruction (P < 0.05). Quantitatively measured beam hardening artifacts were similar in reduced-dose CT with iDose levels 5–6 and fewer with IMR compared to standard-dose CT. In the qualitative analysis, beam hardening artifacts and noise decreased as the IR levels increased. However, artificial texture was significantly aggravated with iDose 5–6 and IMR, and overall image quality significantly worsened with IMR.ConclusionsIR algorithms can reduce beam hardening artifacts in a reduced-dose CT setting in patients with arms-down position, and an intermediate level of hybrid IR allows radiologists to obtain the best image quality. Because the retrospective and single-center nature of our study limited the number of patients, multicenter prospective clinical studies are required to validate our results.

Purpose:To evaluate nodule detection in an anthropomorphic chest phantom in computed tomography (CT) images reconstructed with adaptive iterative dose reduction 3D (AIDR3D) and filtered back projection (FBP) over a range of tube current–time product (mAs).Methods:Two phantoms were used in this study: (i) an anthropomorphic chest phantom was loaded with spherical simulated nodules of 5, 8, 10, and 12 mm in diameter and +100, −630, and −800 Hounsfield units electron density; this would generate CT images for the observer study; (ii) a whole-body dosimetry verification phantom was used to ultimately estimate effective dose and risk according to the model of the BEIR VII committee. Both phantoms were scanned over a mAs range (10, 20, 30, and 40), while all other acquisition parameters remained constant. Images were reconstructed with both AIDR3D and FBP. For the observer study, 34 normal cases (no nodules) and 34 abnormal cases (containing 1–3 nodules, mean 1.35 ± 0.54) were chosen. Eleven observers evaluated images from all mAs and reconstruction methods under the free-response paradigm. A crossed-modality jackknife alternative free-response operating characteristic (JAFROC) analysis method was developed for data analysis, averaging data over the two factors influencing nodule detection in this study: mAs and image reconstruction (AIDR3D or FBP). A Bonferroni correction was applied and the threshold for declaring significance was set at 0.025 to maintain the overall probability of Type I error at α = 0.05. Contrast-to-noise (CNR) was also measured for all nodules and evaluated by a linear least squares analysis.Results:For random-reader fixed-case crossed-modality JAFROC analysis, there was no significant difference in nodule detection between AIDR3D and FBP when data were averaged over mAs [F(1, 10) = 0.08, p = 0.789]. However, when data were averaged over reconstruction methods, a significant difference was seen between multiple pairs of mAs settings [F(3, 30) = 15.96, p < 0.001]. Measurements of effective dose and effective risk showed the expected linear dependence on mAs. Nodule CNR was statistically higher for simulated nodules on images reconstructed with AIDR3D (p < 0.001).Conclusions:No significant difference in nodule detection performance was demonstrated between images reconstructed with FBP and AIDR3D. mAs was found to influence nodule detection, though further work is required for dose optimization.

For a given imaging task and patient size, the optimal selection of x-ray tube potential (kV) and tube current-rotation time product (mAs) is pivotal in achieving the maximal radiation dose reduction while maintaining the needed diagnostic performance. Although contrast-to-noise (CNR)-based strategies can be used to optimize kV/mAs for computed tomography (CT) imaging systems employing the linear filtered backprojection (FBP) reconstruction method, a more general framework needs to be developed for systems using the nonlinear statistical model-based iterative reconstruction (MBIR) method. The purpose of this paper is to present such a unified framework for the optimization of kV/mAs selection for both FBP- and MBIR-based CT systems. The optimal selection of kV and mAs was formulated as a constrained optimization problem to minimize the objective function, Dose(kV,mAs), under the constraint that the achievable detectability index d'(kV,mAs) is not lower than the prescribed value of d'R for a given imaging task. Since it is difficult to analytically model the dependence of d' on kV and mAs for the highly nonlinear MBIR method, this constrained optimization problem is solved with comprehensive measurements of Dose(kV,mAs) and d'(kV,mAs) at a variety of kV-mAs combinations, after which the overlay of the dose contours and d' contours is used to graphically determine the optimal kV-mAs combination to achieve the lowest dose while maintaining the needed detectability for the given imaging task. As an example, d' for a 17 mm hypoattenuating liver lesion detection task was experimentally measured with an anthropomorphic abdominal phantom at four tube potentials (80, 100, 120, and 140 kV) and fifteen mA levels (25 and 50-700) with a sampling interval of 50 mA at a fixed rotation time of 0.5 s, which corresponded to a dose (CTDIvol) range of [0.6, 70] mGy. Using the proposed method, the optimal kV and mA that minimized dose for the prescribed detectability level of d'R=16 were determined. As another example, the optimal kV and mA for an 8 mm hyperattenuating liver lesion detection task were also measured using the developed framework. Both an in vivo animal and human subject study were used as demonstrations of how the developed framework can be applied to the clinical work flow. For the first task, the optimal kV and mAs were measured to be 100 and 500, respectively, for FBP, which corresponded to a dose level of 24 mGy. In comparison, the optimal kV and mAs for MBIR were 80 and 150, respectively, which corresponded to a dose level of 4 mGy. The topographies of the iso-d' map and the iso-CNR map were the same for FBP; thus, the use of d'- and CNR-based optimization methods generated the same results for FBP. However, the topographies of the iso-d' and iso-CNR map were significantly different in MBIR; the CNR-based method overestimated the performance of MBIR, predicting an overly aggressive dose reduction factor. For the second task, the developed framework generated the following optimization results: for FBP, kV = 140, mA = 350, dose = 37.5 mGy; for MBIR, kV = 120, mA = 250, dose = 18.8 mGy. Again, the CNR-based method overestimated the performance of MBIR. Results of the preliminary in vivo studies were consistent with those of the phantom experiments. A unified and task-driven kV/mAs optimization framework has been developed in this work. The framework is applicable to both linear and nonlinear CT systems such as those using the MBIR method. As expected, the developed framework can be reduced to the conventional CNR-based kV/mAs optimization frameworks if the system is linear. For MBIR-based nonlinear CT systems, however, the developed task-based kV/mAs optimization framework is needed to achieve the maximal dose reduction while maintaining the desired diagnostic performance.

BackgroundPositron Emission Tomography (PET), Computed Tomography (CT), PET/CT and Single Photon Emission Tomography (SPECT) are non-invasive imaging tools used for creating two dimensional (2D) cross section images of three dimensional (3D) objects. PET and SPECT have the potential of providing functional or biochemical information by measuring distribution and kinetics of radiolabelled molecules, whereas CT visualizes X-ray density in tissues in the body. PET/CT provides fused images representing both functional and anatomical information with better precision in localization than PET alone.Images generated by these types of techniques are generally noisy, thereby impairing the imaging potential and affecting the precision in quantitative values derived from the images. It is crucial to explore and understand the properties of noise in these imaging techniques. Here we used autocorrelation function (ACF) specifically to describe noise correlation and its non-isotropic behaviour in experimentally generated images of PET, CT, PET/CT and SPECT.MethodsExperiments were performed using phantoms with different shapes. In PET and PET/CT studies, data were acquired in 2D acquisition mode and reconstructed by both analytical filter back projection (FBP) and iterative, ordered subsets expectation maximisation (OSEM) methods. In the PET/CT studies, different magnitudes of X-ray dose in the transmission were employed by using different mA settings for the X-ray tube. In the CT studies, data were acquired using different slice thickness with and without applied dose reduction function and the images were reconstructed by FBP. SPECT studies were performed in 2D, reconstructed using FBP and OSEM, using post 3D filtering. ACF images were generated from the primary images, and profiles across the ACF images were used to describe the noise correlation in different directions. The variance of noise across the images was visualised as images and with profiles across these images.ResultsThe most important finding was that the pattern of noise correlation is rotation symmetric or isotropic, independent of object shape in PET and PET/CT images reconstructed using the iterative method. This is, however, not the case in FBP images when the shape of phantom is not circular. Also CT images reconstructed using FBP show the same non-isotropic pattern independent of slice thickness and utilization of care dose function. SPECT images show an isotropic correlation of the noise independent of object shape or applied reconstruction algorithm. Noise in PET/CT images was identical independent of the applied X-ray dose in the transmission part (CT), indicating that the noise from transmission with the applied doses does not propagate into the PET images showing that the noise from the emission part is dominant. The results indicate that in human studies it is possible to utilize a low dose in transmission part while maintaining the noise behaviour and the quality of the images.ConclusionThe combined effect of noise correlation for asymmetric objects and a varying noise variance across the image field significantly complicates the interpretation of the images when statistical methods are used, such as with statistical estimates of precision in average values, use of statistical parametric mapping methods and principal component analysis. Hence it is recommended that iterative reconstruction methods are used for such applications. However, it is possible to calculate the noise analytically in images reconstructed by FBP, while it is not possible to do the same calculation in images reconstructed by iterative methods. Therefore for performing statistical methods of analysis which depend on knowing the noise, FBP would be preferred.

The purpose of this study was to survey the radiation dose used in CT urography (CTU) in routine clinical practice, both before and after implementation of a scanning protocol that uses iterative reconstruction (Adaptive Iterative Dose Reduction 3D [AIDR 3D]). We retrospectively surveyed dose reports from consecutive CTU examinations performed in 2011 with the use of 64- and 320-MDCT scanners that were reconstructed with filtered back projection (FBP) and from CTU examinations performed from May 2012 through November 2013 that were reconstructed with the use of AIDR 3D. Findings from these dose reports were then correlated with such patient characteristics as weight and body mass index (BMI; weight in kilograms divided by the square of height in meters). Only dose reports from single-bolus three-phase CTU examinations were included in the study. The volume CT dose index, dose-length product (DLP), and effective dose were surveyed both per examination and per phase by use of published effective dose DLP conversion factors. Image quality was evaluated subjectively for a subset of patients. The two study cohorts included 82 patients (median patient weight, 75.0 kg; median BMI, 25.3) who underwent CTU with FBP and 85 patients (median patient weight, 78.0 kg; median BMI, 24.5) who underwent CTU with AIDR 3D. The median total DLP and median effective dose were 924 mGy · cm and 13.0 mSv, respectively, in the CTU with the FBP cohort and 433 mGy · cm and 6.1 mSv, respectively, in the CTU with the AIDR 3D cohort. The median DLP in the unenhanced, nephrogenic, and excretory phases was 218, 300, and 441 mGy · cm, respectively, in patients undergoing CTU with FBP and 114, 121, and 190 mGy · cm, respectively, in patients undergoing CTU with AIDR 3D. Image quality was diagnostic in both groups, with relatively fewer artifacts noted on scans obtained using CTU with AIDR 3D. Our study presents detailed dose data from three-phase CTU examinations performed both before and after implementation of iterative reconstruction. Implementation of a CTU protocol using iterative reconstruction resulted in a mean effective dose of 6.1 mSv with preservation of clinical diagnostic image quality.

Low RCS targets detection is a challenging problem for marine radar in sea clutter. In this paper, a new detection algorithm is developed based on Radon and inverse Radon transform which accounts for the linear feature of the targets in sea clutter. After Radon transform the SCR is increased and clutter can be suppressed more easily in Radon domain. Then the data in Radon domain can be inverse Radon transformed into time-range domain, and the target track can be detected. The simulation results show that the proposed algorithm makes the detection of very small targets possible.

The objective of this study was to evaluate the radiation dose reduction potential of a novel image-based denoising technique in pediatric abdominopelvic and chest CT examinations and compare it with a commercial iterative reconstruction method. Data were retrospectively collected from 50 (25 abdominopelvic and 25 chest) clinically indicated pediatric CT examinations. For each examination, a validated noise-insertion tool was used to simulate half-dose data, which were reconstructed using filtered back-projection (FBP) and sinogram-affirmed iterative reconstruction (SAFIRE) methods. A newly developed denoising technique, adaptive nonlocal means (aNLM), was also applied. For each of the 50 patients, three pediatric radiologists evaluated four datasets: full dose plus FBP, half dose plus FBP, half dose plus SAFIRE, and half dose plus aNLM. For each examination, the order of preference for the four datasets was ranked. The organ-specific diagnosis and diagnostic confidence for five primary organs were recorded. The mean (± SD) volume CT dose index for the full-dose scan was 5.3 ± 2.1 mGy for abdominopelvic examinations and 2.4 ± 1.1 mGy for chest examinations. For abdominopelvic examinations, there was no statistically significant difference between the half dose plus aNLM dataset and the full dose plus FBP dataset (3.6 ± 1.0 vs 3.6 ± 0.9, respectively; p = 0.52), and aNLM performed better than SAFIRE. For chest examinations, there was no statistically significant difference between the half dose plus SAFIRE and the full dose plus FBP (4.1 ± 0.6 vs 4.2 ± 0.6, respectively; p = 0.67), and SAFIRE performed better than aNLM. For all organs, there was more than 85% agreement in organ-specific diagnosis among the three half-dose configurations and the full dose plus FBP configuration. Although a novel image-based denoising technique performed better than a commercial iterative reconstruction method in pediatric abdominopelvic CT examinations, it performed worse in pediatric chest CT examinations. A 50% dose reduction can be achieved while maintaining diagnostic quality.