RPS 1315 - Imaging of the aorta, pulmonary, and coronary arteries

Author Block: T. Higashigawa¹, Y. Ichikawa¹, K. Nakajima², T. Kobayashi¹, K. Domae¹, A. Yamazaki¹, N. Kato¹, H. Sakuma¹; ¹Tsu/JP, ²Ise/JP
Purpose: To evaluate the diagnostic performance of low-energy virtual monochromatic CT imaging (VMI) combined with deep-learning image reconstruction (DLIR) for the detection of endoleaks.
Methods or Background: A cohort of 71 patients after endovascular aortic repair who underwent dynamic contrast-enhanced CT between March 2022 and August 2023 were studied. Raw data were reconstructed using three different methods: 70-keV VMI using conventional hybrid iterative reconstruction (HIR [ASiR-V50%]), and 40- and 70-keV VMI using DLIR (TrueFidelity-H). Contrast-to-noise ratio (CNR) of the endoleaks on venous phase CT was calculated. Three observers assessed the presence or absence of endoleak on a 5-point scale, taking into account the confidence level: score-1, endoleaks are definitely not present; score-2, probably not present; score-3, may be present; score-4, probably present; score-5, definitely present. A score of 3 or higher was considered positive for endoleak.
Results or Findings: Endoleaks were observed in 41 (58%) of 71 subjects. The CNRs of endoleaks were significantly higher in 40-keV DLIR (17.1±9.8) compared to 70-keV HIR (6.4±3.8; P<0.001) and 70-keV DLIR (10.2±6.2; P<0.001). ROC analysis for endoleak detection showed that AUC for 40-keV DLIR (0.92-0.99) was the largest for all observers (70-keV DLIR, 0.91-0.97; 70-keV HIR, 0.88-0.96). The percentage of patients with endoleaks who were correctly identified with a confidence level of ≥ score-4 in 40-keV VMI with DLIR was significantly higher compared to those in 70-keV VMI with HIR in one observer (Observer1, 85%(35/41) vs 73%(30/41), P=0.02; Observer2, 85%(35/41) vs 78%(32/41), P=0.20; Observer3, 98%(40/41) vs 90%(37/41), P=0.10, respectively).
Conclusion: The 40-keV VMI combined with DLIR reconstruction method improves the CNR of endoleaks and may help to correctly identify endoleaks with higher confidence compared to 70-keV VMI with HIR.
Limitations: The limitations of the study is the relatively small study population.
Funding for this study: No funding was received for this study.
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: The study was approved by institutional review board (approval number; H2019-207).

Author Block: J. Lu, L. Shen, Z. Zhao, Z. Bi, M. Zeng, M. M. Wang; Shanghai/CN
Purpose: To investigate the effects of deep learning reconstruction (DLR) on the image quality in ultra-low dose CT pulmonary angiography (CTPA), compared to hybrid iterative reconstruction (HIR) at routine dose.
Methods or Background: This study prospectively included 130 patients with suspected pulmonary embolism (PE) who underwent CTPA examination in two hospitals from April to July 2024. The noise index of routine dose (RD) group and ultra-low dose (ULD) group was set to 10 and 20, respectively. The CT images of RD group were reconstructed using HIR, while ULD group were reconstructed with HIR and DLR. Pulmonary CT value, signal-to-noise ratio (SNR), and contrast-to-noise ratio (CNR) were assessed as quantitative criteria of image quality. Two senior radiologists independently evaluated the overall image noise, pulmonary artery visibility, and diagnostic confidence based on a 5‑point Likert scale (5, best; 1, worst). The Mann-Whitney U test and the Wilcoxon signed rank test was used for statistical analysis.
Results or Findings: No statistically significant difference was found in the clinical data between the two groups (p>0.05). The ULD-DLR group exhibited higher SNRs and CNRs in all seven pulmonary arteries compared to the RD-HIR group (both p<0.05). The overall image noise and diagnostic confidence of the ULD-DLR images were significantly better than that in the RD-HIR images and ULD-HIR images (both p<0.001). The effective dose in the RD group and ULD group were 2.84±0.49mSv and 0.70±0.21mSv, respectively, representing a reduction of approximately 75% in the ULD group (p<0.001).
Conclusion: DLR can significantly reduce the radiation dose of CTPA examination without compromising the diagnosis of PE. Even at ultra-low radiation dose, its image quality is still better than HIR at routine dose.
Limitations: Not applicable.
Funding for this study: No.
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: Shanghai Geriatrics Medical Center Ethics Committee (B2024-009)

Author Block: M. Dewilde, W. Coudyzer, A. Laenen, H. Bosmans, G. Maleux; Leuven/BE
Purpose: To prospectively compare aortic image quality by adapting contrast volume and kiloVoltage (kV) in patients referred for preoperative aortic computed tomography (aCT).
Methods or Background: Eighty prospectively included patients were assigned into 3 groups: 50% of the contrast dose calculated on body surface area (BSA) and heart rate (HR) (group 1, n=56); 50% of the contrast dose calculated on BSA and HR and additional kV reduction (group 2, n=11); 50% of contrast dose calculated on BSA and HR and additional contrast dilution 80% contrast & 20% NaCl (group 3, n=13). Quantitative (measurement of Hounsfield units) analysis at different anatomical aortic levels and qualitative image analysis by 2 radiologists using a visual score (1 = inadequate; 5 = excellent) was performed.
Results or Findings: Mean contrast dose injected was 46.1 ml, 28.3 ml and 35.0 ml for group 1, 2 and 3 respectively, with a significant difference between group 1 and 2 (P=<0.001) and between group 1 and 3 (P=<0.001); no difference between group 2 and 3 (P=0.072). Mean qualitative scores were 4.35/5, 2.82/5 and 3.46/5 for group 1, 2 and 3 respectively. No patient needed repeat imaging for inadequate aortic CT-imaging. Interobserver agreement was moderate for group 1 and 3 (0.577 and 0.576 resp.) and fair for group 2 (0.282) with consistent difference in scoring.
Conclusion: Meaningful contrast dose reduction in preoperative aCT while maintaining diagnostic efficacy is feasible through utilization of a contrast injection algorithm incorporating patient’s HR and BSA, coupled with adjusting kV values.
Limitations: Limited number of patients (n=80). Only patients with a pre operative angio CT were included.
Funding for this study: No funding was received for this study.
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: The institutional ethics committee approved this study (S58042).

Author Block: C. R. G. L. O. M. Talei Franzesi, D. Ippolito, C. Maino, P. N. Franco, D. G. Gandola, R. Corso; Milan/IT
Purpose: To evaluate image quality and radiation dose reduction of deep learning reconstruction algorithm in CT angiography (CTA) studies performed for TAVI planning, compared with low dose CTA reconstructed with hybrid iterative algorithm
Methods or Background: Fifty six patients candidates for TAVI were enrolled in this study and 26 patients (study-group) were examined with 128 MDCT scanner, with 80 kV, automated mAs dose-modulation and 50 mL of contrast media (CM), combined with a new deep learning reconstruction algorithm (Precise Image); while a control group of 32 patients were evaluated with 256 MDCT (100 KV; automated mAs; 50 mL of CM) reconstructed with hybrid iterative reconstruction algorithm (iDose4).
Subjective (using a 4-point Likert scale) and objective image quality (vascular enhancement, SNR and CNR in different aortic levels and in the iliac arteries) were evaluated and the radiation dose exposure of both groups (CTDIvol and DLP) was calculated
Results or Findings: Study group with deep learning algorithm demonstrated significantly higher mean attenuation values (p<.05) in all the measurements compared to the control group with model based algorithm (aortic root 621HU vs 314 HU; external iliac arteries 537HU vs 335HU). Mean DLP and CTDI of study group was significantly lower than in control group (DLP: 395 mGy*cm vs 1600 mGy*cm, p<0.001; CTDI: 8.03 mGy vs 23.5 mGy, p<0.001), with an overall radiation dose reduction of about 75%. Furthermore, study group showed a significant decrease of image noise with an increase of image quality
Conclusion: Deep learning based CT reconstruction algorithm combined with low Kv setting allows to significantly reduce radiation dose exposure and increase the image quality in CTA protocol for TAVI planning, in comparison with low dose CTA reconstructed with hybrid iterative algorithm
Limitations: None
Funding for this study: None
Has your study been approved by an ethics committee? Not applicable
Ethics committee - additional information: None

Author Block: M. Both, C. Jochum, J. H. Schirmer, E. A. Strathmann, P. Langguth, S. Sandra Freitag-Wolf, C. Von Der Burchard, O. Jansen, M. Salehi Ravesh; Kiel/DE
Purpose: The application of imaging techniques for early detection of thoracic aortic aneurysms in patients with giant cell arteritis (GCA), including the identification and monitoring of subgroups at high risk for this condition, is still the subject of debate. We investigated whether aortic stiffness could be quantified based on MRI and used as a potential biomarker for post-inflammatory damage.
Methods or Background: Ten GCA patients in clinical remission and 36 healthy volunteers (HVs) were examined using non-contrast-enhanced cine-balanced steady-state free precession (bSSFP) MRI technique to determine the distensibility and diameter of the ascending (AAo), descending (DAo), and arch (AArch) segments of the thoracic aorta. In addition, changes in aortic diameters during follow-up in GCA patients and the impact of demographic and clinical characteristics on the aortic elasticity were investigated.
Results or Findings: Distensibility was significantly higher in the AArch (p=0.039) and in the DAo (p=0.004) than in the AAo in HVs, but not in GCA patients. Aortic distensibility was significantly lower in patients than in HVs in the AArch (0.89 vs. 2.15, p=0.035). Age was an additional predictor of aortic stiffening in the AAo (p=0.029) and DAo (p=0.001) of HVs. In patients with GCA, the diameter increased at an above-average rate in all aortic segments (AAo 1.04 mm/year, AArch 1.12 mm/year, DAo 0.95 mm/year) compared to baseline MRI.
Conclusion: The bSSFP MRI technique revealed functional and structural differences in the thoracic aorta of patients with GCA as a potential marker for weakness of the thoracic aortic wall.
Limitations: The small size of our patient group is the main limitation due to its single-center design. Another limitation relates to the fact that most of our study patients presented with predominantly cranial symptoms, some without proof of aortitis on MRI.
Funding for this study: None
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: The Ethics Committee at the Faculty of Medicine of Kiel University approved this study (No. D577/18).

Author Block: A. M. C. Boehner, B. Salam, A. Jacob, A. Isaak, C. C. Pieper, D. Kütting; Bonn/DE
Purpose: Coronary Computed Tomography Angiography (CCTA) often requires extensive preparation, contributing to prolonged in-room time. This study aims to assess the impact of pre-examination preparation, including the administration of IV beta-blockers and ECG lead placement outside the examination room, on reducing in-room time for coronary CT scans.
Methods or Background: A prospective study with 139 patients was conducted, comparing standard in-room preparation (control cohort) with receiving preparation outside (intervention cohort) and mostly omitted preparation, scanned via a spiral acquisition protocol (spiral cohort). Patients' heart rates were regulated in the intervention group before entering the examination room. Key measures included: patient entering the examination-room, installation of patient monitoring, heart rate adjustment, first scan, heart rate during scanning and image quality.
Results or Findings: The intervention cohort demonstrated significantly (P<0.0001) reduced in-room time compared to the control cohort (984±347s vs. 704±308s). The spiral cohort performed best and displayed the lowest variability (583±103s, P<0.0001). The heart rates of the spiral cohort was highest with 69±21bpm (P<0.04), but the intervention cohort did not differ from the control cohort (60±7bpm vs. 59±6bpm, P=0.88). The rate of non-diagnostic segments remained low across all groups (control: 3.5%, intervention: 1.5%, spiral: 2.0%).
Conclusion: Pre-examination preparation outside of the examination room, including installation of patient monitoring and administration of IV beta-blockers, significantly reduces in-room time for CCTA without compromising image quality. Alternatively, a spiral image acquisition protocol allows for the omission of most preparatory steps. Both approaches offer a feasible strategy to streamline workflow and enhance efficiency in cardiac imaging departments.
Limitations: The limitation regarding the spiral cohort is the need to preselect patients with a coronary-calcium-score <400. Some radiological departments may lack the premises to implement our approach.
Funding for this study: The study was conducted in collaboration with Siemens Healthineers.
Has your study been approved by an ethics committee? Not applicable
Ethics committee - additional information: Administrational

Author Block: L. D. Grünewald, V. Koch, S. Mahmoudi, J. Gotta, P. Reschke, J-E. Scholtz, S. Martin, C. Booz, T. Vogl; Frankfurt/DE
Purpose: To approximate the Agatston score in contrast-enhanced CTs using a volumetric approach without distortions from contrast agents.
Methods or Background: The aorta of 1276 patients (886 men, 390 women; median age 67 years; interquartile range 57-76) without prior surgical interventions who underwent contrast-enhanced multi-phase CT between January 2018 and December 2023 were retrospectively analyzed. For all patients, the Agatston score was derived from unenhanced CT scans for the thoracic and abdominal aorta. The number and volume of plaques were assessed for the thoracic and abdominal aorta in arterial and venous contrast phases using thresholds ranging from 100 to 1000 to assess the influence of contrast agents. Correlations with the Agatston score were calculated, and linear regression was used to identify the optimal threshold.
Results or Findings: Median aortic enhancement was 46 HU (unenhanced), 323 HU (arterial), and 120 HU (venous). In venous phases, a threshold of 300 HU yielded the highest correlations with the Agatston score (thoracic: r=0.91; abdominal: r=0.93; p<0.001). In arterial phases, a threshold of 900 HU provided the best correlation (thoracic: r=0.72; abdominal: r=0.65; p<0.001). Linear regression confirmed these thresholds, but showed only moderate predictive power (R²=0.66 for venous, R²=0.52 for arterial phases). Dynamic thresholding resulted in poor correlation (r=0.26; p<0.001) and low predictive value (R²=0.07).
Conclusion: Plaque volume assessment using optimized HU thresholds can reliably approximate the Agatston score in contrast-enhanced CTs, offering a robust assessment without contrast-induced distortion. This approach is particularly valuable when non-contrast images are unavailable, such as in staging or pre-TAVR evaluations.
Limitations: Patient collective with high plaque burden.
Funding for this study: No funding was received for this study.
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: Waiver due to retrospective nature

Author Block: A. Janin-Manificat¹, M. Sigovan¹, L. Boussel¹, A. Millon¹, P. C. Douek¹, S. Boccalini²; ¹Lyon/FR, ²Villeurbanne/FR
Purpose: To assess the performance of conventional CT (conv-CT) and spectral CT (spectral-CT) for the detection and characterization of communications between the true (TL) and false lumen (FL) in aortic dissections, using 4D-flow MRI as the reference.
Methods or Background: 18 patients with type A and B aortic dissection who underwent 4D-flow MRI, conv-CT and spectral-CT were included. For each patient, the exams closest in time, without any surgical or endovascular intervention in-between were retrieved and subjectively analysed by two observers, independently for conv-CT and in consensus for MRI and spectral-CT. Communications between the two lumens were identified as: intimal tears on conv-CT; focal alterations in velocities corresponding to jet flows on 4D-flow; both intimal tears and focal changes of contrast concentration corresponding to jet flows on spectral-CT. The number, size, and location of communications were noted. Additionally, the direction of the flow was assessed for spectral-CT and MRI.
Results or Findings: Of the 176 communications detected with 4D-flow, spectral-CT allowed visualisation of 122 (69%) compared to 58 (33%) for Obs1 and 38 (22%) for Obs2 for conv-CT, yielding an accuracy twice as high (63% vs. 29-30%).
On spectral CT, in only 45 cases (26%) the size of the communications could be assessed, in all other cases only jet flows were detected without visible intimal tears.
The flow was unidirectional TL-FL in 2 cases for both MRI and spectral-CT and bidirectional in 5 and 3 cases for the two modalities. In all other cases the flow was in the direction TL-FL.
Conclusion: Spectral-CT outperformed conv-CT for the detection of communications between TL and FL in aortic dissections. Spectral-CT allows direct visualization of flow jets, and their direction, through intimal tears.
Limitations: Low number of patients; time in-between different exams
Funding for this study: No
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: Approved

Author Block: J. Lu, L. Shen, Z. Zhao, Z. Bi, M. Zeng, M. M. Wang; Shanghai/CN
Purpose: To assess the performance of artificial intelligence (AI) software and inexperienced physicians in diagnosing pulmonary embolism (PE) at deep learning reconstruction-based ultra-low dose (ULD) CT pulmonary angiography (CTPA).
Methods or Background: This prospective two-center study contained 210 patients with suspected pulmonary embolism (PE) who underwent CTPA examination, randomizing into two groups with equal proportion of patients. Images in the routine-dose (RD) group were reconstructed using hybrid iterative reconstruction (HIR, AIDR 3D, FC08), while ULD images were reconstructed using HIR and deep learning reconstruction (DLR, AiCE), respectively. A subset of 74 participants (1:1 PE to non-PE ratio) was randomly selected and evaluated by two inexperienced physicians and AI software (Discover PE, uAI). Reference standard was established by expert consensus. The diagnostic accuracy (sensitivity and specificity) of the AI or reader interpretations were compared between methods by bootstrapping.
Results or Findings: There was no statistically significant difference in the patient demographics between two groups. ULD-DLR images exhibited significantly higher objective and subjective image quality compared to both RD-HIR and ULD-HIR images. The AI software exhibited near-perfect accuracy in both ULD-HIR and ULD-DLR sets (sensitivity: 97.30 %, specificity: 100 %). In comparison, two physicians showed a mean sensitivity of 75.68% and specificity of 93.75% in ULD-HIR sets, and a mean sensitivity of 94.59% and specificity of 100.00% in ULD-DLR sets. Inter-observer agreement was moderate for HIR (κ = 0.75) and good for DLR (κ = 0.81). The effective dose of ULD group was significantly lower than the RD group (2.74±0.47 mSv vs. 0.73±0.25 mSv, p<0.001).
Conclusion: DLR can significantly reduce the radiation dose of CTPA examination without compromising the diagnosis of pulmonary embolism even at ultra-low radiation dose. AI software outperforms inexperienced physicians in interpreting ULD images.
Limitations: Not applicable.
Funding for this study: No.
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: Shanghai Geriatrics Medical Center Ethics Committee (B2024-009)