Towards safer imaging: A comparative study of deep learning-based denoising and iterative reconstruction in intraindividual low-dose CT scans using an in-vivo large animal model.

PURPOSE: Computed tomography (CT) scans are a significant source of medically induced radiation exposure. Novel deep learning-based denoising (DLD) algorithms have been shown to enable diagnostic image quality at lower radiation doses than iterative reconstruction (IR) methods. However, most comparative studies employ low-dose simulations due to ethical constraints. We used real intraindividual animal scans to investigate the dose-reduction capabilities of a DLD algorithm in comparison to IR. MATERIALS AND METHODS: Fourteen veterinarian-sedated alive pigs underwent 2 CT scans on the same 3rd generation dual-source scanner with two months between each scan. Four additional scans ensued each time, with mAs reduced to 50 %, 25 %, 10 %, and 5 %. All scans were reconstructed ADMIRE levels 2 (IR2) and a novel DLD algorithm, resulting in 280 datasets. Objective image quality (CT numbers stability, noise, and contrast-to-noise ratio) was measured via consistent regions of interest. Three radiologists independently rated all possible dataset combinations per time point for subjective image quality (-1 = inferior, 0 = equal, 1 = superior). The points were averaged for a semiquantitative score, and inter-rater agreement was measured using Spearman's correlation coefficient and adequately corrected mixed-effects modeling analyzed objective and subjective image quality. RESULTS: Neither dose-reduction nor reconstruction method negatively impacted CT number stability (p > 0.999). In objective image quality assessment, the lowest radiation dose achievable by DLD when comparing noise (p = 0.544) and CNR (p = 0.115) to 100 % IR2 was 25 %. Overall, inter-rater agreement of the subjective image quality ratings was strong (r ≥ 0.69, mean 0.93 ± 0.05, 95 % CI 0.92-0.94; each p < 0.001), and subjective assessments corroborated that DLD at 25 % radiation dose was comparable to 100 % IR2 in image quality, sharpness, and contrast (p ≥ 0.281). CONCLUSIONS: The DLD algorithm can achieve image quality comparable to the standard IR method but with a significant dose reduction of up to 75%. This suggests a promising avenue for lowering patient radiation exposure without sacrificing diagnostic quality.

Novel Deep Learning Denoising Enhances Image Quality and Lowers Radiation Exposure in Interventional Bronchial Artery Embolization Cone Beam CT.

OBJECTIVES: In interventional bronchial artery embolization (BAE), periprocedural cone beam CT (CBCT) improves guiding and localization. However, a trade-off exists between 6-second runs (high radiation dose and motion artifacts, but low noise) and 3-second runs (vice versa). This study aimed to determine the efficacy of an advanced deep learning denoising (DLD) technique in mitigating the trade-offs related to radiation dose and image quality during interventional BAE CBCT. MATERIALS AND METHODS: This study included BMI-matched patients undergoing 6-second and 3-second BAE CBCT scans. The dose-area product values (DAP) were obtained. All datasets were reconstructed using standard weighted filtered back projection (OR) and a novel DLD software. Objective image metrics were derived from place-consistent regions of interest, including CT numbers of the Aorta and lung, noise, and contrast-to-noise ratio. Three blinded radiologists performed subjective assessments regarding image quality, sharpness, contrast, and motion artifacts on all dataset combinations in a forced-choice setup (-1 = inferior, 0 = equal; 1 = superior). The points were averaged per item for a total score. Statistical analysis ensued using a properly corrected mixed-effects model with post hoc pairwise comparisons. RESULTS: Sixty patients were assessed in 30 matched pairs (age 64 ± 15 years; 10 female). The mean DAP for the 6 s and 3 s runs was 2199 ± 185 µGym² and 1227 ± 90 µGym², respectively. Neither low-dose imaging nor the reconstruction method introduced a significant HU shift (p ≥ 0.127). The 3 s-DLD presented the least noise and superior contrast-to-noise ratio (CNR) (p < 0.001). While subjective evaluation revealed no noticeable distinction between 6 s-DLD and 3 s-DLD in terms of quality (p ≥ 0.996), both outperformed the OR variants (p < 0.001). The 3 s datasets exhibited fewer motion artifacts than the 6 s datasets (p < 0.001). CONCLUSIONS: DLD effectively mitigates the trade-off between radiation dose, image noise, and motion artifact burden in regular reconstructed BAE CBCT by enabling diagnostic scans with low radiation exposure and inherently low motion artifact burden at short examination times.

Image Quality and Quantitative PET Parameters of Low-Dose [18F]FDG PET in a Long Axial Field-of-View PET/CT Scanner.

PET/CT scanners with a long axial field-of-view (LAFOV) provide increased sensitivity, enabling the adjustment of imaging parameters by reducing the injected activity or shortening the acquisition time. This study aimed to evaluate the limitations of reduced [18F]FDG activity doses on image quality, lesion detectability, and the quantification of lesion uptake in the Biograph Vision Quadra, as well as to assess the benefits of the recently introduced ultra-high sensitivity mode in a clinical setting. A number of 26 patients who underwent [18F]FDG-PET/CT (3.0 MBq/kg, 5 min scan time) were included in this analysis. The PET raw data was rebinned for shorter frame durations to simulate 5 min scans with lower activities in the high sensitivity (HS) and ultra-high sensitivity (UHS) modes. Image quality, noise, and lesion detectability (n = 82) were assessed using a 5-point Likert scale. The coefficient of variation (CoV), signal-to-noise ratio (SNR), tumor-to-background ratio (TBR), and standardized uptake values (SUV) including SUVmean, SUVmax, and SUVpeak were evaluated. Subjective image ratings were generally superior in UHS compared to the HS mode. At 0.5 MBq/kg, lesion detectability decreased to 95% (HS) and to 98% (UHS). SNR was comparable at 1.0 MBq/kg in HS (5.7 ± 0.6) and 0.5 MBq/kg in UHS (5.5 ± 0.5). With lower doses, there were negligible reductions in SUVmean and SUVpeak, whereas SUVmax increased steadily. Reducing the [18F]FDG activity to 1.0 MBq/kg (HS/UHS) in a LAFOV PET/CT provides diagnostic image quality without statistically significant changes in the uptake parameters. The UHS mode improves image quality, noise, and lesion detectability compared to the HS mode.

Diagnosis of an Acute Anterior Wall Infarction in Dual-Energy CT.

Due to its high morbidity and mortality, myocardial infarction is the leading cause of death worldwide. Against this background, rapid diagnosis is of immense importance. Especially in case of an atypical course, the correct diagnosis may be delayed and thus lead to increased mortality rates. In this report, we present a complex case of acute coronary syndrome. A triple-rule-out CT examination was performed in dual-energy CT (DECT) mode. While pulmonary artery embolism and aortic dissection could be ruled out with conventional CT series, the presence of anterior wall infarction was only detectable on DECT reconstructions. Subsequently, adequate and rapid therapy was then initiated leading to survival of the patient.

How Real Are Computed Tomography Low Dose Simulations? An Investigational In-Vivo Large Animal Study.

OBJECTIVES: CT low-dose simulation methods have gained significant traction in protocol development, as they lack the risk of increased patient exposure. However, in-vivo validations of low-dose simulations are as uncommon as prospective low-dose image acquisition itself. Therefore, we investigated the extent to which simulated low-dose CT datasets resemble their real-dose counterparts. MATERIALS AND METHODS: Fourteen veterinarian-sedated alive pigs underwent three CT scans on the same third generation dual-source scanner with 2 months between each scan. At each time, three additional scans ensued, with mAs reduced to 50%, 25%, and 10%. All scans were reconstructed using wFBP and ADMIRE levels 1-5. Matching low-dose datasets were generated from the 100% scans using reconstruction-based and DICOM-based simulations. Objective image quality (CT numbers stability, noise, and signal-to-noise ratio) was measured via consistent regions of interest. Three radiologists independently rated all possible dataset combinations per time point for subjective image quality (-1=inferior, 0=equal, 1=superior). The points were averaged for a semiquantitative score, and inter-rater-agreement was measured using Spearman's correlation coefficient. A structural similarity index (SSIM) analyzed the voxel-wise similarity of the volumes. Adequately corrected mixed-effects analysis compared objective and subjective image quality. Multiple linear regression with three-way interactions measured the contribution of dose, reconstruction mode, simulation method, and rater to subjective image quality. RESULTS: There were no significant differences between objective and subjective image quality of reconstruction-based and DICOM-based simulation on all dose levels (p≥0.137). However, both simulation methods produced significantly lower objective image quality than real-dose images below 25% mAs due to noise overestimation (p<0.001; SSIM≤89±3). Overall, inter-rater-agreement was strong (r≥0.68, mean 0.93±0.05, 95% CI 0.92-0.94; each p<0.001). In regression analysis, significant decreases in subjective image quality were observed for lower radiation doses (b ≤ -0.387, 95%CI -0.399 to -0.358; p<0.001) but not for reconstruction modes, simulation methods, raters, or three-way interactions (p≥0.103). CONCLUSION: Simulated low-dose CT datasets are subjectively and objectively indistinguishable from their real-dose counterparts down to 25% mAs, making them an invaluable tool for efficient low-dose protocol development.

Application of a Deep Learning Algorithm for Combined Super-Resolution and Partial Fourier Reconstruction Including Time Reduction in T1-Weighted Precontrast and Postcontrast Gradient Echo Imaging of Abdominopelvic MR Imaging.

Purpose: The purpose of this study was to test the technical feasibility and the impact on the image quality of a deep learning-based super-resolution reconstruction algorithm in 1.5 T abdominopelvic MR imaging. Methods: 44 patients who underwent abdominopelvic MRI were retrospectively included, of which 4 had to be subsequently excluded. After the acquisition of the conventional volume interpolated breath-hold examination (VIBEStd), images underwent postprocessing, using a deep learning-based iterative denoising super-resolution reconstruction algorithm for partial Fourier acquisitions (VIBESR). Image analysis of 40 patients with a mean age of 56 years (range 18−84 years) was performed qualitatively by two radiologists independently using a Likert scale ranging from 1 to 5, where 5 was considered the best rating. Results: Image analysis showed an improvement of image quality, noise, sharpness of the organs and lymph nodes, and sharpness of the intestine for pre- and postcontrast images in VIBESR compared to VIBEStd (each p < 0.001). Lesion detectability was better for VIBESR (p < 0.001), while there were no differences concerning the number of lesions. Average acquisition time was 16 s (±1) for the upper abdomen and 15 s (±1) for the pelvis for VIBEStd, and 15 s (±1) for the upper abdomen and 14 s (±1) for the pelvis for VIBESR. Conclusion: This study demonstrated the technical feasibility of a deep learning-based super-resolution algorithm including partial Fourier technique in abdominopelvic MR images and illustrated a significant improvement of image quality, noise, and sharpness while reducing TA.

AI Denoising Improves Image Quality and Radiological Workflows in Pediatric Ultra-Low-Dose Thorax Computed Tomography Scans.

(1) This study evaluates the impact of an AI denoising algorithm on image quality, diagnostic accuracy, and radiological workflows in pediatric chest ultra-low-dose CT (ULDCT). (2) Methods: 100 consecutive pediatric thorax ULDCT were included and reconstructed using weighted filtered back projection (wFBP), iterative reconstruction (ADMIRE 2), and AI denoising (PixelShine). Place-consistent noise measurements were used to compare objective image quality. Eight blinded readers independently rated the subjective image quality on a Likert scale (1 = worst to 5 = best). Each reader wrote a semiquantitative report to evaluate disease severity using a severity score with six common pathologies. The time to diagnosis was measured for each reader to compare the possible workflow benefits. Properly corrected mixed-effects analysis with post-hoc subgroup tests were used. Spearman's correlation coefficient measured inter-reader agreement for the subjective image quality analysis and the severity score sheets. (3) Results: The highest noise was measured for wFBP, followed by ADMIRE 2, and PixelShine (76.9 ± 9.62 vs. 43.4 ± 4.45 vs. 34.8 ± 3.27 HU; each p < 0.001). The highest subjective image quality was measured for PixelShine, followed by ADMIRE 2, and wFBP (4 (4−5) vs. 3 (4−5) vs. 3 (2−4), each p < 0.001) with good inter-rater agreement (r ≥ 0.790; p ≤ 0.001). In diagnostic accuracy analysis, there was a good inter-rater agreement between the severity scores (r ≥ 0.764; p < 0.001) without significant differences between severity score items per reconstruction mode (F (5.71; 566) = 0.792; p = 0.570). The shortest time to diagnosis was measured for the PixelShine datasets, followed by ADMIRE 2, and wFBP (2.28 ± 1.56 vs. 2.45 ± 1.90 vs. 2.66 ± 2.31 min; F (1.000; 99.00) = 268.1; p < 0.001). (4) Conclusions: AI denoising significantly improves image quality in pediatric thorax ULDCT without compromising the diagnostic confidence and reduces the time to diagnosis substantially.

Combination of Whole-Body Baseline CT Radiomics and Clinical Parameters to Predict Response and Survival in a Stage-IV Melanoma Cohort Undergoing Immunotherapy.

BACKGROUND: This study investigated whether a machine-learning-based combination of radiomics and clinical parameters was superior to the use of clinical parameters alone in predicting therapy response after three months, and overall survival after six and twelve months, in stage-IV malignant melanoma patients undergoing immunotherapy with PD-1 checkpoint inhibitors and CTLA-4 checkpoint inhibitors. METHODS: A random forest model using clinical parameters (demographic variables and tumor markers = baseline model) was compared to a random forest model using clinical parameters and radiomics (extended model) via repeated 5-fold cross-validation. For this purpose, the baseline computed tomographies of 262 stage-IV malignant melanoma patients treated at a tertiary referral center were identified in the Central Malignant Melanoma Registry, and all visible metastases were three-dimensionally segmented (n = 6404). RESULTS: The extended model was not significantly superior compared to the baseline model for survival prediction after six and twelve months (AUC (95% CI): 0.664 (0.598, 0.729) vs. 0.620 (0.545, 0.692) and AUC (95% CI): 0.600 (0.526, 0.667) vs. 0.588 (0.481, 0.629), respectively). The extended model was not significantly superior compared to the baseline model for response prediction after three months (AUC (95% CI): 0.641 (0.581, 0.700) vs. 0.656 (0.587, 0.719)). CONCLUSIONS: The study indicated a potential, but non-significant, added value of radiomics for six-month and twelve-month survival prediction of stage-IV melanoma patients undergoing immunotherapy.

AI Denoising Significantly Enhances Image Quality and Diagnostic Confidence in Interventional Cone-Beam Computed Tomography.

(1) To investigate whether interventional cone-beam computed tomography (cbCT) could benefit from AI denoising, particularly with respect to patient body mass index (BMI); (2) From 1 January 2016 to 1 January 2022, 100 patients with liver-directed interventions and peri-procedural cbCT were included. The unenhanced mask run and the contrast-enhanced fill run of the cbCT were reconstructed using weighted filtered back projection. Additionally, each dataset was post-processed using a novel denoising software solution. Place-consistent regions of interest measured signal-to-noise ratio (SNR) per dataset. Corrected mixed-effects analysis with BMI subgroup analyses compared objective image quality. Multiple linear regression measured the contribution of "Radiation Dose", "Body-Mass-Index", and "Mode" to SNR. Two radiologists independently rated diagnostic confidence. Inter-rater agreement was measured using Spearman correlation (r); (3) SNR was significantly higher in the denoised datasets than in the regular datasets (p < 0.001). Furthermore, BMI subgroup analysis showed significant SNR deteriorations in the regular datasets for higher patient BMI (p < 0.001), but stable results for denoising (p > 0.999). In regression, only denoising contributed positively towards SNR (0.6191; 95%CI 0.6096 to 0.6286; p < 0.001). The denoised datasets received overall significantly higher diagnostic confidence grades (p = 0.010), with good inter-rater agreement (r ≥ 0.795, p < 0.001). In a subgroup analysis, diagnostic confidence deteriorated significantly for higher patient BMI (p < 0.001) in the regular datasets but was stable in the denoised datasets (p ≥ 0.103).; (4) AI denoising can significantly enhance image quality in interventional cone-beam CT and effectively mitigate diagnostic confidence deterioration for rising patient BMI.

AI Denoising Significantly Improves Image Quality in Whole-Body Low-Dose Computed Tomography Staging.

(1) Background: To evaluate the effects of an AI-based denoising post-processing software solution in low-dose whole-body computer tomography (WBCT) stagings; (2) Methods: From 1 January 2019 to 1 January 2021, we retrospectively included biometrically matching melanoma patients with clinically indicated WBCT staging from two scanners. The scans were reconstructed using weighted filtered back-projection (wFBP) and Advanced Modeled Iterative Reconstruction strength 2 (ADMIRE 2) at 100% and simulated 50%, 40%, and 30% radiation doses. Each dataset was post-processed using a novel denoising software solution. Five blinded radiologists independently scored subjective image quality twice with 6 weeks between readings. Inter-rater agreement and intra-rater reliability were determined with an intraclass correlation coefficient (ICC). An adequately corrected mixed-effects analysis was used to compare objective and subjective image quality. Multiple linear regression measured the contribution of "Radiation Dose", "Scanner", "Mode", "Rater", and "Timepoint" to image quality. Consistent regions of interest (ROI) measured noise for objective image quality; (3) Results: With good-excellent inter-rater agreement and intra-rater reliability (Timepoint 1: ICC ≥ 0.82, 95% CI 0.74-0.88; Timepoint 2: ICC ≥ 0.86, 95% CI 0.80-0.91; Timepoint 1 vs. 2: ICC ≥ 0.84, 95% CI 0.78-0.90; all p ≤ 0.001), subjective image quality deteriorated significantly below 100% for wFBP and ADMIRE 2 but remained good-excellent for the post-processed images, regardless of input (p ≤ 0.002). In regression analysis, significant increases in subjective image quality were only observed for higher radiation doses (≥0.78, 95%CI 0.63-0.93; p < 0.001), as well as for the post-processed images (≥2.88, 95%CI 2.72-3.03, p < 0.001). All post-processed images had significantly lower image noise than their standard counterparts (p < 0.001), with no differences between the post-processed images themselves. (4) Conclusions: The investigated AI post-processing software solution produces diagnostic images as low as 30% of the initial radiation dose (3.13 ± 0.75 mSv), regardless of scanner type or reconstruction method. Therefore, it might help limit patient radiation exposure, especially in the setting of repeated whole-body staging examinations.

Diagnostic Performance of a Contrast-Enhanced Ultra-Low-Dose High-Pitch CT Protocol with Reduced Scan Range for Detection of Pulmonary Embolisms.

(1) Background: To evaluate the diagnostic performance of a simulated ultra-low-dose (ULD), high-pitch computed tomography pulmonary angiography (CTPA) protocol with low tube current (mAs) and reduced scan range for detection of pulmonary embolisms (PE). (2) Methods: We retrospectively included 130 consecutive patients (64 ± 16 years, 69 female) who underwent clinically indicated high-pitch CTPA examination for suspected acute PE on a 3rd generation dual-source CT scanner (SOMATOM FORCE, Siemens Healthineers, Forchheim, Germany). ULD datasets with a realistic simulation of 25% mAs, reduced scan range (aortic arch-basal pericardium), and Advanced Modeled Iterative Reconstruction (ADMIRE®, Siemens Healthineers, Forchheim, Germany) strength 5 were created. The effective radiation dose (ED) of both datasets (standard and ULD) was estimated using a dedicated dosimetry software solution. Subjective image quality and diagnostic confidence were evaluated independently by three reviewers using a 5-point Likert scale. Objective image quality was compared using noise measurements. For assessment of diagnostic accuracy, patients and pulmonary vessels were reviewed binarily for affection by PE, using standard CTPA protocol datasets as the reference standard. Percentual affection of pulmonary vessels by PE was computed for disease severity (modified Qanadli score). (3) Results: Mean ED in ULD protocol was 0.7 ± 0.3 mSv (16% of standard protocol: 4.3 ± 1.7 mSv, p < 0.001, r > 0.5). Comparing ULD to standard protocol, subjective image quality and diagnostic confidence were comparably good (p = 0.486, r > 0.5) and image noise was significantly lower in ULD (p < 0.001, r > 0.5). A total of 42 patients (32.2%) were affected by PE. ULD protocol had a segment-based false-negative rate of only 0.1%. Sensitivity for detection of any PE was 98.9% (95% CI, 97.2-99.7%), specificity was 100% (95% CI, 99.8-100%), and overall accuracy was 99.9% (95% CI, 98.6-100%). Diagnoses correlated strongly between ULD and standard protocol (Chi-square (1) = 42, p < 0.001) with a decrease in disease severity of only 0.48% (T = 1.667, p = 0.103). (4) Conclusions: Compared to a standard CTPA protocol, the proposed ULD protocol proved reliable in detecting and ruling out acute PE with good levels of image quality and diagnostic confidence, as well as significantly lower image noise, at 0.7 ± 0.3 mSv (84% dose reduction).

Simulated Radiation Dose Reduction in Whole-Body CT on a 3rd Generation Dual-Source Scanner: An Intraindividual Comparison.

To evaluate the effect of radiation dose reduction on image quality and diagnostic confidence in contrast-enhanced whole-body computed tomography (WBCT) staging. We randomly selected March 2016 for retrospective inclusion of 18 consecutive patients (14 female, 60 ± 15 years) with clinically indicated WBCT staging on the same 3rd generation dual-source CT. Using low-dose simulations, we created data sets with 100, 80, 60, 40, and 20% of the original radiation dose. Each set was reconstructed using filtered back projection (FBP) and Advanced Modeled Iterative Reconstruction (ADMIRE®, Siemens Healthineers, Forchheim, Germany) strength 1-5, resulting in 540 datasets total. ADMIRE 2 was the reference standard for intraindividual comparison. The effective radiation dose was calculated using commercially available software. For comparison of objective image quality, noise assessments of subcutaneous adipose tissue regions were performed automatically using the software. Three radiologists blinded to the study evaluated image quality and diagnostic confidence independently on an equidistant 5-point Likert scale (1 = poor to 5 = excellent). At 100%, the effective radiation dose in our population was 13.3 ± 9.1 mSv. At 20% radiation dose, it was possible to obtain comparably low noise levels when using ADMIRE 5 (p = 1.000, r = 0.29). We identified ADMIRE 3 at 40% radiation dose (5.3 ± 3.6 mSv) as the lowest achievable radiation dose with image quality and diagnostic confidence equal to our reference standard (p = 1.000, r > 0.4). The inter-rater agreement for this result was almost perfect (ICC ≥ 0.958, 95% CI 0.909-0.983). On a 3rd generation scanner, it is feasible to maintain good subjective image quality, diagnostic confidence, and image noise in single-energy WBCT staging at dose levels as low as 40% of the original dose (5.3 ± 3.6 mSv), when using ADMIRE 3.

AI Lung Segmentation and Perfusion Analysis of Dual-Energy CT Can Help to Distinguish COVID-19 Infiltrates from Visually Similar Immunotherapy-Related Pneumonitis Findings and Can Optimize Radiological Workflows.

(1) To explore the potential impact of an AI dual-energy CT (DECT) prototype on decision making and workflows by investigating its capabilities to differentiate COVID-19 from immunotherapy-related pneumonitis. (2) Methods: From 3 April 2020 to 12 February 2021, DECT from biometrically matching patients with COVID-19, pneumonitis, and inconspicuous findings were selected from our clinical routine. Three blinded readers independently scored each pulmonary lobe analogous to CO-RADS. Inter-rater agreement was determined with an intraclass correlation coefficient (ICC). Averaged perfusion metrics per lobe (iodine uptake in mg, volume without vessels in ml, iodine concentration in mg/mL) were extracted using manual segmentation and an AI DECT prototype. A generalized linear mixed model was used to investigate metric validity and potential distinctions at equal CO-RADS scores. Multinomial regression measured the contribution "Reader", "CO-RADS score", and "perfusion metrics" to diagnosis. The time to diagnosis was measured for manual vs. AI segmentation. (3) Results: We included 105 patients (62 ± 13 years, mean BMI 27 ± 2). There were no significant differences between manually and AI-extracted perfusion metrics (p = 0.999). Regardless of the CO-RADS score, iodine uptake and concentration per lobe were significantly higher in COVID-19 than in pneumonitis (p < 0.001). In regression, iodine uptake had a greater contribution to diagnosis than CO-RADS scoring (Odds Ratio (OR) = 1.82 [95%CI 1.10-2.99] vs. OR = 0.20 [95%CI 0.14-0.29]). The AI prototype extracted the relevant perfusion metrics significantly faster than radiologists (10 ± 1 vs. 15 ± 2 min, p < 0.001). (4) Conclusions: The investigated AI prototype positively impacts decision making and workflows by extracting perfusion metrics that differentiate COVID-19 from visually similar pneumonitis significantly faster than radiologists.