Deep Learning Reconstruction Algorithm Research Articles

High-Resolution Single-Shot Fast Spin-Echo MR Imaging with Deep Learning Reconstruction Algorithm Can Improve Repeatability and Reproducibility of Follicle Counting.

To investigate the diagnostic performance of high-resolution single-shot fast spin-echo (SSFSE) imaging with deep learning (DL) reconstruction algorithm on follicle counting and compare it with original SSFSE images and conventional fast spin-echo (FSE) images. This study included 20 participants (40 ovaries) with clinically confirmed polycystic ovary syndrome (PCOS) who underwent high-resolution ovary MRI, including three-plane T2-weighted FSE sequences and slice-matched T2-weighted SSFSE sequences. A DL reconstruction algorithm was applied to the SSFSE sequences to generate SSFSE-DL images, and the original SSFSE images were also saved. Subjective evaluations such as the blurring artifacts, subjective noise, and clarity of the follicles on the SSFSE-DL, SSFSE, and conventional FSE images were independently conducted by two observers. Intra-class correlation coefficients and Bland-Altman plots were used to present the repeatability and reproducibility of the follicle number per ovary (FNPO) based on the three types of images. SSFSE-DL images showed less blurring artifact, subjective noise, and better clarity of the follicles than SSFSE and FSE (p < 0.05). For the repeatability of the FNPO, SSFSE-DL showed the highest intra-observer (ICC = 0.930; 95% CI: 0.878-0.962) and inter-observer (ICC = 0.914; 95% CI: 0.843-0.953) agreements. The inter-observer 95% limits of agreement (LOA) for SSFSE-DL, SSFSE, and FSE ranged from -3.7 to 4.5, -4.4 to 7.0, and -7.1 to 7.6, respectively. The intra-observer 95% LOA for SSFSE-DL, SSFSE, and FSE ranged from -3.5 to 4.0, -5.1 to 6.1, and -5.7 to 4.2, respectively. The absolute values of intra-observer and inter-observer differences for SSFSE-DL were significantly lower than those for SSFSE and FSE (p < 0.05). Compared with the original SSFSE images and the conventional FSE images, high-resolution SSFSE images with DL reconstruction algorithm can better display follicles, thus improving FNPO assessment.

First Results of a New Deep Learning Reconstruction Algorithm on Image Quality and Liver Metastasis Conspicuity for Abdominal Low-Dose CT

The study's aim was to assess the impact of a deep learning image reconstruction algorithm (Precise Image; DLR) on image quality and liver metastasis conspicuity compared with an iterative reconstruction algorithm (IR). This retrospective study included all consecutive patients with at least one liver metastasis having been diagnosed between December 2021 and February 2022. Images were reconstructed using level 4 of the IR algorithm (i4) and the Standard/Smooth/Smoother levels of the DLR algorithm. Mean attenuation and standard deviation were measured by placing the ROIs in the fat, muscle, healthy liver, and liver tumor. Two radiologists assessed the image noise and image smoothing, overall image quality, and lesion conspicuity using Likert scales. The study included 30 patients (mean age 70.4 ± 9.8 years, 17 men). The mean CTDIvol was 6.3 ± 2.1 mGy, and the mean dose-length product 314.7 ± 105.7 mGy.cm. Compared with i4, the HU values were similar in the DLR algorithm at all levels for all tissues studied. For each tissue, the image noise significantly decreased with DLR compared with i4 (p < 0.01) and significantly decreased from Standard to Smooth (-26 ± 10%; p < 0.01) and from Smooth to Smoother (-37 ± 8%; p < 0.01). The subjective image assessment confirmed that the image noise significantly decreased between i4 and DLR (p < 0.01) and from the Standard to Smoother levels (p < 0.01), but the opposite occurred for the image smoothing. The highest scores for overall image quality and conspicuity were found for the Smooth and Smoother levels.

Improved image quality with deep learning reconstruction - a study on a semi-anthropomorphic upper-abdomen phantom.

To assess image quality of a deep learning reconstruction (DLR) algorithm across dose levels using a semi-anthropomorphic upper-abdominal phantom, and compare with filtered back projection (FBP) and hybrid iterative reconstruction (IR). CT scans obtained at five dose levels (CTDIvol 5, 10, 15, 20 and 25 mGy) were reconstructed with FBP, hybrid IR (IR50, IR70 and IR90) and DLR of low (DLL), medium (DLM) and high strength (DLH) in 0.625mm and 2.5mm slices. CT number, homogeneity, noise, contrast, contrast-to-noise ratio (CNR), noise texture deviation (NTD; a measure of IR-specific artifacts), noise power spectrum (NPS) and task-based transfer function (TTF) were compared between reconstruction algorithms. CT numbers were highly consistent across reconstruction algorithms. Image noise was significantly reduced with higher levels of DLR. Noise texture (NPS and NTD) was with DLR maintained at comparable levels to FBP, contrary to increasing levels of hybrid IR. Images reconstructed with DLR of low and high strength in 0.625mm slices showed similar noise characteristics to 2.5mm slice FBP and IR50, respectively. Dose-reduction potential based on image noise with IR50 as reference was estimated to 35% for DLM and 74% for DLH. The novel DLR algorithm demonstrates robust noise reduction with maintained noise texture characteristics despite higher algorithm strength, and may have overcome important limitations of IR. There may be potential for dose reduction and additional benefit from thin-slice reconstruction.

Spatial resolution, noise properties, and detectability index of a deep learning reconstruction algorithm for dual-energy CT of the abdomen.

Iterative reconstruction (IR) has increasingly replaced traditional reconstruction methods in computed tomography (CT). The next paradigm shift in image reconstruction is likely to come from artificial intelligence, with deep learning reconstruction (DLR) solutions already entering the clinic. An enduring disadvantage to IR has been a change in noise texture, which can affect diagnostic confidence. DLR has demonstrated the potential to overcome this issue and has recently become available for dual-energy CT. To evaluate the spatial resolution, noise properties, and detectability index of a commercially available DLR algorithm for dual-energy CT of the abdomen and compare it to single-energy (SE) CT. An oval 25cmx35cm custom-made phantom was scanned on a GE Revolution CT scanner (GE Healthcare, Waukesha, WI) at two dose levels (13 and 5mGy) and two iodine concentrations (8 and 2mg/mL), using three typical abdominal scan protocols: dual-energy (DE), SE 80kV (SE-80kV) and SE 120kV (SE-120kV). Reconstructions were performed with three strengths of IR (ASiR-V: AR0%, AR50%, AR100%) and three strengths of DLR (TrueFidelity: low, medium, high). The DE acquisitions were reconstructed as mono-energetic images between 40 and 80keV. The noise power spectrum (NPS), task transfer function (TTF), and detectability index (d') were determined for the reconstructions following the recommendations of AAPM Task Group 233. Noise magnitude reductions (relative to AR0%) for the SE protocols were on average (-29%, -21%) for (AR50%, TF-M), while for DE-70keV were (-28%, -43%). There was less reduction in mean frequency (fav ) for DLR than for IR, with similar results for SE and DE imaging. There was, however, a substantial change in the NPS shape when using DE with DLR, quantifiable by a marked reduction in the peak frequency (fpeak ) that was absent in SE mode. All protocols and reconstructions (including AR0%) exhibited slight to moderate shifts towards lower spatial frequencies at the lower dose (<12% in fav ). Spatial resolution was consistently superior for DLR compared to IR for SE but not for DE. All protocols and reconstructions (including AR0%) showed decreased resolution with reduced dose and iodine concentration, with less decrease for DLR compared to IR. DLR displayed a higher d' than IR. The effect of energy was large: d' increased with lower keV, and SE-80kV had higher d' than SE-120kV. Using DE with DLR could provide higher d' than SE-80kV at the higher dose but not at lower dose. DE imaging with DLR maintained spatial resolution and reduced noise magnitude while displaying less change in noise texture than IR. The d' was also higher with DLR than IR, suggesting superiority in detectability of iodinated contrast. Despite these trends being consistent with those previously established for SE imaging, there were some noteworthy differences. For DE imaging there was no improvement in resolution compared to IR and a change in noise texture. DE imaging with low keV and DLR had superior detectability to SE DLR at the high dose but was not better than SE-80kV at low dose.

Deep Learning Image Reconstruction for CT: Technical Principles and Clinical Prospects.

Filtered back projection (FBP) has been the standard CT image reconstruction method for 4 decades. A simple, fast, and reliable technique, FBP has delivered high-quality images in several clinical applications. However, with faster and more advanced CT scanners, FBP has become increasingly obsolete. Higher image noise and more artifacts are especially noticeable in lower-dose CT imaging using FBP. This performance gap was partly addressed by model-based iterative reconstruction (MBIR). Yet, its "plastic" image appearance and long reconstruction times have limited widespread application. Hybrid iterative reconstruction partially addressed these limitations by blending FBP with MBIR and is currently the state-of-the-art reconstruction technique. In the past 5 years, deep learning reconstruction (DLR) techniques have become increasingly popular. DLR uses artificial intelligence to reconstruct high-quality images from lower-dose CT faster than MBIR. However, the performance of DLR algorithms relies on the quality of data used for model training. Higher-quality training data will become available with photon-counting CT scanners. At the same time, spectral data would greatly benefit from the computational abilities of DLR. This review presents an overview of the principles, technical approaches, and clinical applications of DLR, including metal artifact reduction algorithms. In addition, emerging applications and prospects are discussed.

COMBINING HI-RESOLUTION SCAN MODE WITH DEEP LEARNING RECONSTRUCTION ALGORITHMS IN CARDIAC CT.

To investigate the impact of combining the high-resolution (Hi-res) scan mode with deep learning image reconstruction (DLIR) algorithm in CT. Two phantoms (Catphan600® and Lungman, small, medium, large size) were CT scanned using combinations of Hi-res/standard mode and high-definition (HD)/standard kernels. Images were reconstructed with ASiR-V and three levels of DLIR. Spatial resolution, noise and contrast-to-noise ratio (CNR) were assessed. The radiation dose was recorded. The spatial resolution increased using Hi-res & HD. Image noise in the Catphan600® (69%) and the Lungman (10-70%) significantly increased when Hi-res & HD was applied. DLIR reduced the mean noise (54%). The CNR was reduced (64%) for Hi-res & HD. The radiation dose increased for both small (+70%) and medium (+43%) Lungman phantoms but decreased slightly for the large ones (-3%) when Hi-res was applied. In conclusion, the Hi-res scan mode improved the spatial resolution. The HD kernel significantly increased the image noise. DLIR improved the image noise and CNR and did not affect the spatial resolution.

Effect of Deep Learning-based Contrast-enhanced CT Image Reconstruction on the Image Quality of the Biliary System

Objective To evaluate the effect of a deep learning reconstruction (DLR) method on the visibility of contrast-enhanced CT images of the biliary system by comparing it with different iterative reconstruction algorithms including the adaptive iterative dose reduction 3D (AIDR 3D) algorithm,forward projected model based iterative reconstruction solution (FIRST),and filtered back projection (FBP) algorithm. Methods A total of 30 patients subjected to abdominal contrast-enhanced CT and diagnosed with dilatation of common bile duct or extrahepatic bile duct were retrospectively included in this study.The images of the portal phase were reconstructed via four different algorithms (FBP,AIDR 3D,FIRST,and DLR).Signal to noise ratio (SNR) and contrast to noise ratio (CNR) of the dilated bile duct,liver parenchyma,measurable bile duct lesions,and image noise were compared between the four datasets.In subjective analyses,two radiologists independently scored the image quality (best:4 points,second:3 points;third:2 points;fourth:1 point) of the four datasets based on the noise and image visual quality of the biliary system.The Friedman and the Bonferroni-Dunn post-hoc tests were performed for comparison. Results The DLR images (bile duct:4.42±0.87;liver parenchyma:3.78±1.47) yielded higher CNR than the FBP (bile duct:2.21±1.02,P<0.001;liver parenchyma:1.43±1.29,P<0.001),AIDR 3D (bile duct:2.81±0.91,P=0.024;liver parenchyma:2.39±1.94,P=0.278),and FIRST (bile duct:2.51±1.24,P<0.001;liver parenchyma:2.45±1.81,P=0.003) images.Furthermore,the DLR images had higher SNR (bile duct:1.39±0.85,liver parenchyma:9.75±1.90) than the FBP (bile duct:0.86±0.63,P<0.001;liver parenchyma:3.31±1.12,P<0.001) and FIRST (bile duct:1.01±0.61,P=0.013;liver parenchyma:5.73±1.37,P<0.001) images,and showed lower noise (10.51±3.53) than the FBP(4.10±3.92,P<0.001),AIDR 3D (15.72±2.41,P=0.032),and FIRST (17.20±3.82,P<0.001) images.SNR and CNR showed no significant differences between FIRST and AIDR 3D images (all P>0.05).DLR images [4(4,4)] obtained higher score than FPB [1(1,1),P<0.001],AIDR3D[3 (2,3),P=0.029],and FIRST[2 (2,3),P<0.001] images. Conclusion DLR algorithm improved the subjective and objective quality of the contrast-enhanced CT image of the biliary system.

Improved image quality and dose reduction in abdominal CT with deep-learning reconstruction algorithm: a phantom study.

To assess the impact of a new artificial intelligence deep-learning reconstruction (Precise Image; AI-DLR) algorithm on image quality against a hybrid iterative reconstruction (IR) algorithm in abdominal CT for different clinical indications. Acquisitions on phantoms were performed at 5 dose levels (CTDIvol: 13/11/9/6/1.8 mGy). Raw data were reconstructed using level 4 of iDose4 (i4) and 3 levels of AI-DLR (Smoother/Smooth/Standard). Noise power spectrum (NPS), task-based transfer function (TTF) and detectability index (d') were computed: d' modelled detection of a liver metastasis (LM) and hepatocellular carcinoma at portal (HCCp) and arterial (HCCa) phases. Image quality was subjectively assessed on an anthropomorphic phantom by 2 radiologists. From Standard to Smoother levels, noise magnitude and average NPS spatial frequency decreased and the detectability (d') of all simulated lesions increased. For both inserts, TTF values were similar for all three AI-DLR levels from 13 to 6 mGy but decreased from Standard to Smoother levels at 1.8 mGy. Compared to the i4 used in clinical practice, d' values were higher using theSmoother and Smooth levels and close for the Standard level. For all dose levels, except at 1.8 mGy, radiologists considered images satisfactory for clinical use for the 3 levels of AI-DLR, but rated images too smooth using the Smoother level. Use ofthe Smooth and Smoother levels of AI-DLR reducesthe image noise and improves the detectability of lesions and spatial resolution for standard and low-dose levels. Using the Smooth level is apparently the best compromise between the lowest dose level and adequate image quality. • Evaluation of the impact of a new artificial intelligence deep-learning reconstruction (AI-DLR) on image quality and dose compared to a hybrid iterative reconstruction (IR) algorithm. • The Smooth and Smoother levels of AI-DLR reduced theimage noise and improved thedetectability of lesions and spatial resolution for standard and low-dose levels. • The Smooth level seems the best compromise between the lowest dose level and adequate image quality.

The impact of deep learning reconstruction on image quality and coronary CT angiography-derived fractional flow reserve values.

To explore the impact of deep learning reconstruction (DLR) on image quality and machine learning-based coronary CT angiography (CTA)-derived fractional flow reserve (CT-FFRML) values. Thirty-three consecutive patients with known or suspected coronary artery disease who underwent coronary CTA and subsequent invasive coronary angiography were enrolled. DLR was compared with filtered back projection (FBP), statistical-based iterative reconstruction (SBIR), model-based iterative reconstruction (MBIR) Cardiac, and MBIR Cardiac sharp for objective image qualities of coronary CTA. Invasive fractional flow reserve (FFR) and quantitative flow ratio (QFR) were used as the reference standards. The diagnostic performances of different reconstruction approach-based CT-FFRML were calculated. A total of 182 lesions in 33 patients were enrolled for analysis. The image quality of DLR was superior to the others. There were no significant differences in the CT-FFRML values among these five approaches (all p > 0.05). Of the 182 lesions, 17 had invasive FFR results, and 70 had QFR results. Using FFR as a reference, MBIR Cardiac, MBIR Cardiac sharp, and DLR achieved equal diagnostic performance, slightly higher than the other reconstruction approaches (MBIR Cardiac, MBIR Cardiac sharp, and DLR: AUC = 0.82, FBP and AIDR: AUC = 0.78, all p > 0.05). Using QFR as a reference, the AUCs of FBP, SBIR, MBIR Cardiac, MBIR Cardiac sharp, and DLR were 0.83, 0.81, 0.86, 0.84, and 0.83, respectively (all p > 0.05). Our study showed that the DLR algorithm improved image quality, but there were no significant differences in the CT-FFRML values and diagnostic performance among different reconstruction approaches. • Deep learning-based image reconstruction (DLR) improves the image quality of coronary CTA. • CT-FFRML values and diagnostic performance of DLR revealed no significant differences compared to other reconstruction approaches.

Clinical acceptance of deep learning reconstruction for abdominal CT imaging: objective and subjective image quality and low-contrast detectability assessment.

To evaluate the image quality and clinical acceptance of a deep learning reconstruction (DLR) algorithm compared to traditional iterative reconstruction (IR) algorithms. CT acquisitions were performed with two phantoms and a total of nine dose levels. Images were reconstructed with two types of IR algorithms, DLR and filtered-back projection. Spatial resolution, image texture, mean noise value, and objective and subjective low-contrast detectability were compared. Ten senior radiologists evaluated the clinical acceptance of these algorithms by scoring ten CT exams reconstructed with the DLR and IR algorithms evaluated. Compared to MBIR, DLR yielded a lower noise and a higher low-contrast detectability index at low doses (CTDIvol ≤ 2.2 and ≤ 4.5mGy, respectively). Spatial resolution and detectability at higher doses were better with MBIR. Compared to HIR, DLR yielded a higher spatial resolution, a lower noise, and a higher detectability index. Despite these differences in algorithm performance, significant differences in subjective low-contrast performance were not found (p ≥ 0.005). DLR texture was finer than that of MBIR and closer to that of HIR. Radiologists preferred DLR images for all criteria assessed (p < 0.0001), whereas MBIR was rated worse than HIR (p < 0.0001) in all criteria evaluated, except for noise (p = 0.044). DLR reconstruction time was 12 times faster than that of MBIR. DLR yielded a gain in objective detection and noise at lower dose levels with the best clinical acceptance among the evaluated reconstruction algorithms. • DLR yielded improved objective low-contrast detection and noise at lower dose levels. • Despite the differences in objective detectability among the algorithms evaluated, there were no differences in subjective detectability. • DLR presented significantly higher clinical acceptability scores compared to MBIR and HIR.

Application of deep-learning reconstruction algorithm for enhanced CT scan of upper abdomen under different radiation doses: focus on noise, contrast-to-noise ratio and image quality

Application of deep-learning reconstruction algorithm for enhanced CT scan of upper abdomen under different radiation doses: focus on noise, contrast-to-noise ratio and image quality

A deep-learning reconstruction algorithm that improves the image quality of low-tube-voltage coronary CT angiography

PurposeTo assess the image quality (IQ) of low tube voltage coronary CT angiography (CCTA) images reconstructed with deep learning image reconstruction (DLIR). MethodsAccording to body mass index (BMI), eighty patients who underwent 70kVp CCTA (Group A, N = 40, BMI ≤ 26 kg/m2) or 80kVp CCTA (Group B, N = 40, BMI > 26 kg/m2) were prospectively included. All images were reconstructed with four algorithms, including filtered back-projection (FBP), adaptive statistical iterative reconstruction-Veo at a level of 50% (ASiR-V50%), and DLIR at medium (DLIR-M) and high (DLIR-H) levels. Image noise, signal-to-noise ratio (SNR), contrast-to-noise ratio (CNR) and edge rise distance (ERD) within aorta root and coronary arteries were calculated. The IQ was subjectively evaluated by using a 5-point scale. ResultsCompared with FBP, ASiR-V50% and DLIR-M, DLIR-H led to the lowest noise (Group A: 24.7 ± 5.0HU; Group B, 21.6 ± 2.8 HU), highest SNR (Group A, 24.9 ± 5.0; Group B, 28.0 ± 5.8), CNR (Group A, 42.2 ± 15.2; Group B, 43.6 ± 10.5) and lowest ERD (Group A, 1.49 ± 0.30 mm; Group B, 1.50 ± 0.22 mm) with statistical significance (all P < 0.05). For the objective assessment, the percentages of 4 and 5 IQ scores were significantly higher for DLIR-H (Group A, 93.8%; Group B,90.0%) and DLIR-M (Group A, 85.6%; Group B,86.9 %) compared to ASiR-V50% (Group A, 58.8%; Group B, 58.8%) and FBP (Group A, 34.4%; Group B, 33.1%) algorithms (all P < 0.05). ConclusionThe application of DLIR significantly improves both objective and subjective IQ in low tube voltage CCTA compared with ASiR-V and FBP, which may promote a further radiation dose reduction in CCTA.

Evaluation of Apparent Noise on CT Images Using Moving Average Filters.

This study aims to devise a simple method for evaluating the magnitude of texture noise (apparent noise) observed on computed tomography (CT) images scanned at a low radiation dose and reconstructed using iterative reconstruction (IR) and deep learning reconstruction (DLR) algorithms, and to evaluate the apparent noise in CT images reconstructed using the filtered back projection (FBP), IR, and two types of DLR (AiCE Body and AiCE Body Sharp) algorithms. We set a square region of interest (ROI) on CT images of standard- and obese-sized low-contrast phantoms, slid different-sized moving average filters in the ROI vertically and horizontally in steps of 1 pixel, and calculated the standard deviation (SD) of the mean CT values for each filter size. The SD of the mean CT values was fitted with a curve inversely proportional to the filter size, and an apparent noise index was determined from the curve-fitting formula. The apparent noise index of AiCE Body Sharp images for a given mAs value was approximately 58, 23, and 18% lower than that of the FBP, AIDR 3D, and AiCE Body images, respectively. The apparent noise index was considered to reflect noise power spectrum values at lower spatial frequency. Moreover, the apparent noise index was inversely proportional to the square roots of the mAs values. Thus, the apparent noise index could be a useful indicator to quantify and compare texture noise on CT images obtained with different scan parameters and reconstruction algorithms.

Metal artefact reduction in the oral cavity using deep learning reconstruction algorithm in ultra-high-resolution computed tomography: a phantom study.

This study aimed to improve the impact of the metal artefact reduction (MAR) algorithm for the oral cavity by assessing the effect of acquisition and reconstruction parameters on an ultra-high-resolution CT (UHRCT) scanner. The mandible tooth phantom with and without the lesion was scanned using super-high-resolution, high-resolution (HR), and normal-resolution (NR) modes. Images were reconstructed with deep learning-based reconstruction (DLR) and hybrid iterative reconstruction (HIR) using the MAR algorithm. Two dental radiologists independently graded the degree of metal artefact (1, very severe; 5, minimum) and lesion shape reproducibility (1, slight; 5, almost perfect). The signal-to-artefact ratio (SAR), accuracy of the CT number of the lesion, and image noise were calculated quantitatively. The Tukey-Kramer method with a p-value of less than 0.05 was used to determine statistical significance. The HRDLR visual score was better than the NRHIR score in terms of degree of metal artefact (4.6 ± 0.5 and 2.6 ± 0.5, p < 0.0001) and lesion shape reproducibility (4.5 ± 0.5 and 2.9 ± 1.1, p = 0.0005). The SAR of HRDLR was significantly better than that of NRHIR (4.9 ± 0.4 and 2.1 ± 0.2, p < 0.0001), and the absolute percentage error of the CT number in HRDLR was lower than that in NRHIR (0.8% in HRDLR and 23.8% in NRIR). The image noise of HRDLR was lower than that of NRHIR (15.7 ± 1.4 and 51.6 ± 15.3, p < 0.0001). Our study demonstrated that the combination of HR mode and DLR in UHRCT scanner improved the impact of the MAR algorithm in the oral cavity.

Deep learning\u2013based reconstruction may improve non-contrast cerebral CT imaging compared to other current reconstruction algorithms

ObjectivesTo evaluate image quality and reconstruction times of a commercial deep learning reconstruction algorithm (DLR) compared to hybrid-iterative reconstruction (Hybrid-IR) and model-based iterative reconstruction (MBIR) algorithms for cerebral non-contrast CT (NCCT).MethodsCerebral NCCT acquisitions of 50 consecutive patients were reconstructed using DLR, Hybrid-IR and MBIR with a clinical CT system. Image quality, in terms of six subjective characteristics (noise, sharpness, grey-white matter differentiation, artefacts, natural appearance and overall image quality), was scored by five observers. As objective metrics of image quality, the noise magnitude and signal-difference-to-noise ratio (SDNR) of the grey and white matter were calculated. Mean values for the image quality characteristics scored by the observers were estimated using a general linear model to account for multiple readers. The estimated means for the reconstruction methods were pairwise compared. Calculated measures were compared using paired t tests.ResultsFor all image quality characteristics, DLR images were scored significantly higher than MBIR images. Compared to Hybrid-IR, perceived noise and grey-white matter differentiation were better with DLR, while no difference was detected for other image quality characteristics. Noise magnitude was lower for DLR compared to Hybrid-IR and MBIR (5.6, 6.4 and 6.2, respectively) and SDNR higher (2.4, 1.9 and 2.0, respectively). Reconstruction times were 27 s, 44 s and 176 s for Hybrid-IR, DLR and MBIR respectively.ConclusionsWith a slight increase in reconstruction time, DLR results in lower noise and improved tissue differentiation compared to Hybrid-IR. Image quality of MBIR is significantly lower compared to DLR with much longer reconstruction times.Key Points• Deep learning reconstruction of cerebral non-contrast CT results in lower noise and improved tissue differentiation compared to hybrid-iterative reconstruction.• Deep learning reconstruction of cerebral non-contrast CT results in better image quality in all aspects evaluated compared to model-based iterative reconstruction.• Deep learning reconstruction only needs a slight increase in reconstruction time compared to hybrid-iterative reconstruction, while model-based iterative reconstruction requires considerably longer processing time.

Deep learning reconstruction versus iterative reconstruction for cardiac CT angiography in a stroke imaging protocol: reduced radiation dose and improved image quality.

To assess the radiation dose and image quality of cardiac computed tomography angiography (CCTA) in an acute stroke imaging protocol using a deep learning reconstruction (DLR) method compared to a hybrid iterative reconstruction algorithm. Retrospective analysis of 296 consecutive patients admitted to the emergency department for stroke suspicion. All patients underwent a stroke CT imaging protocol including a non-enhanced brain CT, a brain perfusion CT imaging if necessary, a CT angiography (CTA) of the supra-aortic vessels, a CCTA and a post-contrast brain CT. The CCTA was performed with a prospectively ECG-gated volume acquisition. Among all CT scans performed, 143 were reconstructed with an iterative reconstruction algorithm (AIDR 3D, adaptive iterative dose reduction three dimensional) and 146 with a DLR algorithm (AiCE, advanced intelligent clear-IQ engine). Image noise, signal-to-noise ratio (SNR), contrast-to-noise ratio (CNR), and subjective image quality (IQ) scored from 1 to 4 were assessed. Dose-length product (DLP), volume CT dose index (CTDIvol) and effective dose (ED) were obtained. The radiation dose was significantly lower with AiCE than with AIDR 3D (DLP =106.4±50.0 vs. 176.1±37.1 mGy·cm, CTDIvol =6.9±3.2 vs. 11.5±2.2 mGy, and ED =1.5±0.7 vs. 2.5±0.5 mSv) (P<0.001). The median SNR and CNR were higher [9.9 (IQR, 8.1-12.3); and 12.6 (IQR, 10.5-15.5), respectively], with AiCE than with AIDR 3D [6.5 (IQR, 5.2-8.5); and 8.4 (IQR, 6.7-11.0), respectively] (P<0.001). SNR and CNR were increased by 51% and 49%, respectively, with AiCE compared to AIDR 3D. The image quality was significantly better with AiCE (mean IQ score =3.4±0.7) than with AIDR 3D (mean IQ score =3±0.9) (P<0.001). The use of a DLR algorithm for cardiac CTA in an acute stroke imaging protocol reduced the radiation dose by about 40% and improved the image quality by about 50% compared to an iterative reconstruction algorithm.

Improving Image Quality and Reducing Radiation Dose for Pediatric CT by Using Deep Learning Reconstruction.

Background CT deep learning reconstruction (DLR) algorithms have been developed to remove image noise. How the DLR affects image quality and radiation dose reduction has yet to be fully investigated. Purpose To investigate a DLR algorithm's dose reduction and image quality improvement for pediatric CT. Materials and Methods DLR was compared with filtered back projection (FBP), statistical-based iterative reconstruction (SBIR), and model-based iterative reconstruction (MBIR) in a retrospective study by using data from CT examinations of pediatric patients (February to December 2018). A comparison of object detectability for 15 objects (diameter, 0.5-10 mm) at four contrast difference levels (50, 150, 250, and 350 HU) was performed by using a non-prewhitening-matched mathematical observer model with eye filter (d'NPWE), task transfer function, and noise power spectrum analysis. Object detectability was assessed by using area under the curve analysis. Three pediatric radiologists performed an observer study to assess anatomic structures with low object-to-background signal and contrast to noise in the azygos vein, right hepatic vein, common bile duct, and superior mesenteric artery. Observers rated from 1 to 10 (worst to best) for edge definition, quantum noise level, and object conspicuity. Analysis of variance and Tukey honest significant difference post hoc tests were used to analyze differences between reconstruction algorithms. Results Images from 19 patients (mean age, 11 years ± 5 [standard deviation]; 10 female patients) were evaluated. Compared with FBP, SBIR, and MBIR, DLR demonstrated improved object detectability by 51% (16.5 of 10.9), 18% (16.5 of 13.9), and 11% (16.5 of 14.8), respectively. DLR reduced image noise without noise texture effects seen with MBIR. Radiologist ratings were 7 ± 1 (DLR), 6.2 ± 1 (MBIR), 6.2 ± 1 (SBIR), and 4.6 ± 1 (FBP); two-way analysis of variance showed a difference on the basis of reconstruction type (P < .001). Radiologists consistently preferred DLR images (intraclass correlation coefficient, 0.89; 95% CI: 0.83, 0.93). DLR demonstrated 52% (1 of 2.1) greater dose reduction than SBIR. Conclusion The DLR algorithm improved image quality and dose reduction without sacrificing noise texture and spatial resolution. © RSNA, 2020 Online supplemental material is available for this article.

J-MoDL: Joint Model-Based Deep Learning for Optimized Sampling and Reconstruction.

Modern MRI schemes, which rely on compressed sensing or deep learning algorithms to recover MRI data from undersampled multichannel Fourier measurements, are widely used to reduce the scan time. The image quality of these approaches is heavily dependent on the sampling pattern. We introduce a continuous strategy to optimize the sampling pattern and the network parameters jointly. We use a multichannel forward model, consisting of a non-uniform Fourier transform with continuously defined sampling locations, to realize the data consistency block within a model-based deep learning image reconstruction scheme. This approach facilitates the joint and continuous optimization of the sampling pattern and the CNN parameters to improve image quality. We observe that the joint optimization of the sampling patterns and the reconstruction module significantly improves the performance of most deep learning reconstruction algorithms. The source code is available at https://github.com/hkaggarwal/J-MoDL.