RPS 415 - Technical innovations in vascular imaging

Author Block: S. Gruendemann, T. Frenzel, J. Lohrke, J. Boyken, G. Jost, M. Berger, H-F. Ulbrich, H. Pietsch; Berlin/DE
Purpose: Gadoquatrane is a tetrameric extracellular macrocyclic gadolinium-based contrast agent (GBCA) with a T1 relaxivity of 11.8 L/(mmol Gd*s) at 1.41 T in human plasma, which has been submitted for regulatory approval in several countries. This study evaluated its stability in comparison with approved macrocyclic GBCAs in vitro and in vivo.
Methods or Background: All assays were conducted at equimolar Gd concentrations. Dissociation kinetics were measured for gadoquatrane, gadoteridol, gadobutrol, gadoterate, and gadopiclenol at pH 1.2 and 37°C using a complexometric assay. Stability in human plasma at pH 7.4 and 37°C was analyzed by ion exchange chromatography with ICP-MS. Rats received a single injection of gadoquatrane, gadobutrol, or gadopiclenol (0.6 mmol Gd/kg; human equivalent 0.1 mmol Gd/kg), and Gd distribution in bone was quantified by laser ablation ICP-MS one week later.
Results or Findings: At pH 1.2, dissociation half-lives were 28.6 days (gadoquatrane), 14.2 days (gadopiclenol), 2.7 days (gadoterate), 14.1 h (gadobutrol), and 2.2 h (gadoteridol). After 15 days in plasma (pH 7.4), no detectable Gd release was observed for gadoquatrane or gadoterate, while gadobutrol, gadoteridol, and gadopiclenol released 0.12%, 0.20%, and 0.20%, respectively. In rats, bone marrow Gd levels were similar across compounds (2.3-3.0 nmol/g). In epiphysis, concentrations were 1.2 nmol/g for gadoquatrane and gadobutrol, and 2.2 nmol/g for gadopiclenol. In diaphysis, levels were 0.5, 1.0, and 2.7 nmol/g, respectively. Elemental imaging showed lowest Gd in mineralized bone for gadoquatrane (<1 nmol/g) compared to gadobutrol and gadopiclenol.
Conclusion: Gadoquatrane demonstrated the highest kinetic inertness under acidic conditions and no measurable Gd release under physiological conditions in plasma. Its high stability was supported by very low Gd concentrations in mineralized bone in rats.
Limitations: These preclinical data do not allow direct conclusions regarding safety in humans.
Funding for this study: None
Has your study been approved by an ethics committee? Not applicable
Ethics committee - additional information:

Author Block: G. Jost, J. Boyken, J. Lohrke, S. Gruendemann, H. Pietsch; Berlin/DE
Purpose: There is ongoing scientific research on the presence of gadolinium (Gd) in bones after administration of Gd-based contrast agents (GBCAs). Gadoquatrane is a novel, macrocyclic GBCA with significantly higher r1-relaxivity, enabling MRI at reduced Gd dose. Gadoquatrane is in clinical development and was recently investigated in phase III studies at a dose of 0.04 mmol Gd/kg. Our study evaluates the presence of Gd in rat femurs compared to standard doses of gadopiclenol and gadobutrol.
Methods or Background: Rats (Wistar-Han, n=40 per GBCA) received a single intravenous injection of gadoquatrane (0.24 mmolGd/kg), gadopiclenol (0.3 mmolGd/kg) or gadobutrol (0.6 mmolGd/kg), representing human doses of 0.04, 0.05 and 0.1 mmol Gd/kg, respectively. After a treatment-free period of 1,4,22 and 52 weeks the Gd concentrations of the dissected femur (n=10 rats) were measured by inductively coupled plasma mass spectrometry (ICP-MS) and laser-ablation ICP-MS imaging (LA-ICP-MS).
Results or Findings: The Gd concentrations in bone epiphysis and diaphysis were comparable after administration of gadobutrol and gadopiclenol, whereas 2-to 6-fold lower concentrations were observed after injection of gadoquatrane. All GBCAs showed a limited Gd wash-out over the one-year follow up. In bone diaphysis the Gd concentration (geometric mean in nmol/g) after 1 and 52 weeks was 0.2±1.4 and 0.2±1.3 (gadoquatrane), 1.0±1.9 and 0.6±1.9 (gadopiclenol), 0.6±1.8 and 0.5±1.2 (gadobutrol). LA-ICP-MS imaging revealed different Gd distribution patterns. A local accumulation of Gd in the cortical bone was present after injection of gadopiclenol, which was barely visible after injection of gadobutrol. No local accumulation was detected for gadoquatrane.
Conclusion: The presence and distribution of Gd in the rat femur after a single injection of human equivalent clinical Gd doses differ among the three GBCAs. For gadoquatrane considerably lower Gd concentrations without local accumulations were detected.
Limitations: Preclinical study
Funding for this study: None
Has your study been approved by an ethics committee? Not applicable
Ethics committee - additional information:

Author Block: F. Wegner¹, M. R. Meyer², V. Gesché³, P. Borm⁴, A. Dell¹, G. Männel¹, D. Wendt¹, T. Friedrich¹, M. M. Sieren¹; ¹Lübeck/DE, ²Duisburg/DE, ³Aachen/DE, ⁴Düsseldorf/DE
Purpose: Commercially available endovascular stents are typically manufactured from metals. Due to their ferromagnetic properties, these devices can cause significant imaging artefacts, potentially limiting follow-up assessments with MRI and posing challenges for future MRI-guided EVAR procedures. The aim of this study was to develop a non-metallic aortic stent graft and evaluate its imaging characteristics in MRI.
Methods or Background: The stent struts were printed in a Z-shaped configuration using a custom-built 3D-printer with PEEK polymer filaments. The printer employed a rotational principle to apply the polymer onto a tubular graft textile. The stent graft was placed via a tube of 20 mm in diameter in a 3D-printed aortic phantom (material: Vero/Agilus, Shore 70A). The phantom was filled with gadolinium-based contrast agent (dilution 1:200) and subsequently imaged using a clinical MRI-scanner. To evaluate potential stent artefacts, the SNRs inside the stent and outside the stent were calculated. The stent strut thickness displayed on MR-images was measured at six locations and averaged.
Results or Findings: The rotational printing process enabled a stable connection between the stent struts and the textile. The resulting stent measured 30 mm in outer diameter and 100 mm in length. The strut thickness was measured by a caliper to be 1.6 mm. After the placement in the phantom, the stent fully expanded and showed no macroscopic loss of integrity. The stent struts were clearly delineable in MRI with a measured strut thickness of 1.7 ± 0.1 mm. The SNR inside the stent was 86.4 and outside 88.1, indicating no relevant stent-induced artefacts.
Conclusion: A non-metallic stent graft can be successfully manufactured using a rotational 3D-printing process with PEEK polymer, enabling artefact-free stent imaging in MRI.
Limitations: Only static measurements have been performed.
Funding for this study: Federal Ministry of Education and Research (BMBF), grant number 13GW0608F
Has your study been approved by an ethics committee? Not applicable
Ethics committee - additional information:

Author Block: P. Elfers¹, K. Lüdtke-Buzug², A. Malhotra², J. Ackers², J. C. Engster², D. Melenberg², M. Ahlborg², T. Friedrich², F. Wegner¹; ¹Lübeck/DE, ²Luebeck/DE
Purpose: Magnetic Particle Imaging (MPI) is a preclinical imaging modality on the verge to first human studies. It is based on the real-time visualisation of magnetic nanoparticles (MNPs). MPI is particularly well suited for cardiovascular imaging and enables artefact-free imaging of stent lumina. To visualise stents during MPI-guided interventions and to prevent interferences during follow-up examinations, a bioresorbable coating is required that degrades after a defined period. The aim of this work was to develop a coating based on polylactic acid (PLA) and MNPs to make stents temporarily visible in MPI.
Methods or Background: Glass capillaries were coated with a PLA-MNP-mixture by dip-coating. The degradation of the coating was investigated in a static water bath experiment at 37 °C. For this purpose, each sample was stored individually in a glass vial filled with isotonic saline solution. Ten different exposure durations (one hour to 28 days) were investigated and the samples were removed for analysis at the respective time. Before and after the water bath, the samples were dried, weighed, their MPI signal was measured, and their morphology was examined by using micro-CT.
Results or Findings: A continuous mass reduction of the markers was observed after exposure to the water bath (approx. 6 % after one hour, approx. 90 % after 28 days). Micro-CT showed only a slight volume decrease, whereas cavities formed inside the markers with increasing exposure time. A detectable MPI signal was present throughout the entire observation period.
Conclusion: PLA seems to be a suitable polymer for degradable stent markers. The observed properties provide a promising basis for future applications of MPI in cardiovascular and periinterventional imaging.
Limitations: The samples were investigated under static conditions only. Flow experiments are required to transfer the results.
Funding for this study: This work was supported in part by the Clinician Scientist Program of the University of Lübeck under Grant CS10-2021. The Fraunhofer IMTE and this work are supported by the EU, the State Schleswig-Holstein, Germany, and by Internal Programs (Grants 12420002/LPWE1.1.1/1536, 12524009/LPW21L/2.2/262 and 139–600251).
Has your study been approved by an ethics committee? Not applicable
Ethics committee - additional information:

Author Block: P. Osztrogonacz, J. Csőre, A. Szentiványi, Á. Bérczi, A. Hüttl, C. Csobay-Novák; Budapest/HU
Purpose: During physician-modified endograft (PMEG) preparation, reloading the stent graft into its delivery sheath is a time-sensitive step. Traditionally, this is performed using the tourniquet (T) technique. At our quaternary aortic referral center, valve stent crimper (C) has emerged as a potential alternative, offering a controlled and possibly more reproducible method.
Our study aimed to compare the C and T techniques in terms of reloading time.
Methods or Background: A Medtronic Valiant 38x200mm thoracic stent graft was reloaded a total of 18 times in a bench-top setting using either the T-technique (n=9) or the C-technique (n=9). Reloading times were recorded in minutes. Statistical analysis was performed using the Mann–Whitney U test to compare reloading times between groups. Results are reported in median (interquartile ranges (IQR)).
Results or Findings: The C-group demonstrated significantly faster reloading times, with a median time of 3.98 minutes (3.27–5.23), compared to 11.48 minutes (8.82–17.35) in the T-group (p = .001). The C group also exhibited narrower variability, suggesting improved reproducibility.
Conclusion: The crimper-technique offers a faster and more consistent method for thoracic stent graft reloading when compared to the traditional tourniquet technique. Its use may improve procedural efficiency and reproducibility in complex endovascular aortic aneurysm repair.
Limitations: Bench-top design, single stent graft type assessed, sample size may not detect rare occurrences of infolding / device failure
Funding for this study: None.
Has your study been approved by an ethics committee? Not applicable
Ethics committee - additional information:

Author Block: W. Hu, J. Zhang; Lanzhou/CN
Purpose: To develop and evaluate a GAN-based method for synthesizing contrast-enhanced CCTA (Syn-CCTA) from NCCTA images, and to determine whether Syn-CCTA achieves sufficient image quality for clinical diagnostic use.
Methods or Background: This retrospective study included 1,528 patients from four centers. Each patient underwent CCTA. Subsequently, we developed an end-to-end GAN model to generate Syn-CCTA images from NCCTA images in the PACS system and transmitted them back to the PACS system. The visual quality of the Syn-CCTA images was compared with that of the original CCTA images using a three-point scale. Additionally, image fidelity was quantitatively assessed using metrics such as the structural similarity index (SSIM), peak signal-to-noise ratio (PSNR), and normalized mean absolute error (NMAE), including t-tests and Mann-Whitney tests for continuous variables and Chi-square tests for dichotomous or categorical variables.
Results or Findings: Demographic data (age, gender, location of calcification) of all patients were not statistically significant. There was no statistical difference between the visual scores of Syn-CCTA images and real CCTA images (p=0.86), SSIM=0.853, PSNR=30.5dB, NMAE=0.022 for Syn-CCTA images.
Conclusion: The diagnostic efficacy of synthetic Syn-CCTA images generated using the end-to-end GAN model is comparable to conventional CCTA images. This finding underscores the potential of GAN-based methodologies to produce diagnostically valuable contrast-enhanced images without using actual contrast agents, offering significant implications for future research and clinical practice in non-invasive cardiovascular imaging.
Limitations: This study was limited by its single-center design and relatively small sample size. Additionally, the synthesized Syn-CCTA images were not validated against clinical outcomes or expert diagnostic performance, which warrants further multicenter prospective studies.
Funding for this study: This study has received funding by the Youth Science and Technology Foundation of Gansu Province (25JRRA623), Gansu Province Clinical Research Center for Functional and Molecular Imaging. Grant (No. 21JR7RA438). Key Project of Gansu Province United Research Foundation (No. 25JRRA1266).
Has your study been approved by an ethics committee? Not applicable
Ethics committee - additional information:

Author Block: W. Hu, J. Zhang; Lanzhou/CN
Purpose: To evaluate the diagnostic performance of high-resolution slow-flow phase-contrast MRA (HRSPC-MRA) in noninvasively detecting the arterial supply and fistula location in spinal dural arteriovenous fistula (SDAVF), using digital subtraction angiography (DSA) as the reference standard.
Methods or Background: This prospective study included 21 patients with clinically suspected and surgically confirmed SDAVF. All patients underwent HRSPC-MRA prior to surgery using the following parameters: FOV 380×380 mm, matrix 480×480, voxel size 0.8×0.8×0.8 mm³. A double dose of gadopentetate dimeglumine (Magnevist) was administered at 4 ml/s via high-pressure injector, followed by saline. Maximum intensity projection (MIP) images were used to identify the arterial feeder(s) and fistula location(s). DSA findings served as the reference. The number and location of fistulas were compared between modalities. Agreement was assessed using Bland-Altman analysis; t-tests, Mann-Whitney U, and chi-square tests were applied as appropriate.
Results or Findings: HRSPC-MRA detected 25 fistulas versus 28 with DSA (p = 0.256). There was no significant difference in demographic or clinical variables. HRSPC-MRA showed high agreement with DSA in identifying both the number and location of fistulas and arterial feeders, with consistent findings on Bland-Altman plots.
Conclusion: HRSPC-MRA provides a noninvasive, contrast-enhanced alternative to DSA for identifying arterial supply and fistula location in SDAVF with comparable diagnostic performance. It holds clinical value for preoperative planning.
Limitations: The small sample size and single-center design may limit generalizability. Further validation in larger, multicenter cohorts is warranted
Funding for this study: This study has received funding by the Youth Science and Technology Foundation of Gansu Province (25JRRA623), Gansu Province Clinical Research Center for Functional and Molecular Imaging. Grant (No. 21JR7RA438). Key Project of Gansu Province United Research Foundation (No. 25JRRA1266).
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: The Second Hospital of Lanzhou University(A2024-0110)

Author Block: G. Lin, W. Chen, M. Chen, J. Ji; Lishui/CN
Purpose: To investigate the value of deep transfer learning (DTL) and radiomics features derived from three-dimensional high-resolution MR vessel wall imaging (3D HRMR-VWI) in identifying symptomatic vertebrobasilar artery plaques.
Methods or Background: This retrospective study included 331 patients with intracranial vertebrobasilar atherosclerosis. Patients were classified as symptomatic or asymptomatic based on the presence of posterior circulation symptoms within two weeks prior to imaging. Plaque features were extracted from non-contrast and contrast-enhanced 3D HRMR-VWI sequences. Handcrafted radiomics (HCR) and DTL features were obtained, and the top ten features were selected. Deep learning radiomics (DLR) features were constructed using HCR, DTL, and combined features. Seven machine learning models were trained and validated to identify symptomatic plaques. The best-performing model was further interpreted using SHapley Additive exPlanations (SHAP) analysis.
Results or Findings: Models based on combined HCR and DTL features achieved the highest predictive performance. After recursive feature elimination with a support vector machine (SVM-RFE), the SVM classifier using combined non-contrast and contrast-enhanced sequences yielded optimal results. In the validation set, the model achieved an area under the curve (AUC) of 0.996, with a Youden index of 0.920, accuracy of 0.976, sensitivity of 0.988, and specificity of 0.932. High predictive performance was confirmed in the independent test set. SHAP analysis indicated that features from contrast-enhanced 3D HRMR-VWI contributed most strongly to model predictions.
Conclusion: The DLR features derived from 3D HRMR-VWI enable accurate identification of symptomatic vertebrobasilar plaques, offering a potential tool for risk stratification and clinical decision support.
Limitations: Our study's retrospective nature may introduce inherent biases. Therefore, our proposed model requires further validation in larger prospective cohorts.
Funding for this study: This work was supported by the National Key Research and Development Program of China (2024YFC2417600) and the Zhejiang Medicine and Health Science and Technology Project (2025KY495, 2024KY568).
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: Institutional Review Board approval was obtained in the Fifth Affiliated Hospital of Wenzhou Medical University (No. 2025-130).

Author Block: Y. Lin, S. Xie; Beijing/CN
Purpose: To develop, externally test, and evaluate clinical acceptability of a deep learning (DL) segmentation model of pulmonary vessels for detecting chronic thromboembolic pulmonary hypertension (CTEPH) at CT pulmonary angiography (CTPA).
Methods or Background: A multicenter, retrospective study was conducted involving 621 consecutive patients who underwent right heart catheterization (RHC), including individuals with confirmed CTEPH, CTED, and normal controls. Among them, 503 patients were assigned to the model development and internal test cohort, and the remaining 118 formed the external validation cohort.
Using CTPA, the pulmonary vasculature was reconstructed via a DL method to separate arteries from veins and compute vascular volumes based on cross-sectional area (CSA) and tortuosity for each branch. Statistically significant features were selected to develop the segmentation model. Diagnostic performance was evaluated using AUC, and the correlation between predicted and actual mean pulmonary arterial pressure (mPAP) was assessed across two institutions.
Results or Findings: The CTEPH cohort in the training set showed significantly larger total vessel volume, large vessel volume, total arterial volume, and greater arterial and venous tortuosity compared to both the CTED and control cohorts (all P < 0.001). Seventeen features were selected to develop a segmentation model, which demonstrated excellent diagnostic performance for identifying CTEPH using RHC results as the gold standard. The model achieved high AUC values (0.96–0.99) in both the training and internal test sets, and maintained strong accuracy in external validation. Additionally, predicted mPAP showed good agreement with real mPAP across two institutions (R² = 0.91 and 0.77, respectively).
Conclusion: Automated quantification methods using CT imaging can provide a imaging marker for identification of CTEPH and CTED.
Limitations: The findings are limited by a small sample size and multiple comparisons, which increased the risk of false-positive results.
Funding for this study: Beijing Natural Science Foundation
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: This project was approved by the Ethics Committee of China-Japan Friendship Hospital

Author Block: Z. Hu; Guiyang/CN
Purpose: To explore the feasibility of artificial intelligence (AI)-based compressed sensing technology in enhancing imaging efficiency, image quality, and lesion detection efficacy in non-contrast-enhanced renal artery magnetic resonance angiography (MRA), and to provide a basis for optimizing non-invasive renal artery imaging protocols.
Methods or Background: A total of 60 patients with suspected renal artery lesions were enrolled and randomly divided into two groups. The control group underwent scanning with a conventional non-contrast-enhanced renal artery MRA sequence, while the experimental group was scanned using a non-contrast-enhanced MRA sequence optimized by AI compressed sensing technology. The scan time, image signal-to-noise ratio (SNR), display grade of renal artery branches, and detection accuracy of renal artery stenosis (≥50%) were compared between the two groups.
Results or Findings: Compared with the control group, the scan time of the experimental group was shortened by 42.3% (P<0.05); the SNR was increased by 28.6% (P<0.05); the display rate of renal artery branches at grade 3 and above reached 91.7%, which was significantly higher than the 73.3% of the control group (P<0.05); there was no significant difference in the lesion detection accuracy between the two groups (93.3% vs 86.7%, P>0.05).
Conclusion: AI compressed sensing technology can shorten the scan time of non-contrast-enhanced renal artery MRA and improve image quality while ensuring lesion detection efficacy, indicating its feasibility for clinical application.
Limitations: The number of patients is insufficient.
Funding for this study: No
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: This study has been reviewed and approved by the Ethics Committee of [Guiqian International General Hospital] (Approval No.: [KY-2025-137]). The study was conducted in strict compliance with the ethical guidelines of the Declaration of Helsinki.
All participants in this study provided adequate informed consent prior to their participation. We informed each participant in writing of the study's purpose, procedures, potential risks and benefits, confidentiality measures, and their right to withdraw unconditionally at any time. All questions raised by the participants were answered patiently by the researchers. Finally, each participant signed a Written Informed Consent Form.

Author Block: J. Csőre¹, A. Crichton², E. Pomozi², M. Drake², P. Haddad², J. Lamichhane², A. B. Lumsden², T. L. Roy²; ¹Budapest/HU, ²Houston, TX/US
Purpose: Chronic limb-threatening ischemia (CLTI) is often linked to chronic kidney disease (CKD), making vascular imaging in this population challenging. Digital subtraction angiography (DSA) remains the diagnostic reference standard but is invasive and requires iodinated contrast. Quiescent Interval Single Shot (QISS) MRI is a newer non-contrast-enhanced method that mitigates these risks. This study aimed to determine whether QISS MRI identifies a greater number of patent arterial segments than DSA, and whether this advantage is most pronounced in patients with advanced CKD (stages 4/5).
Methods or Background: Patients with CLTI who underwent both QISS MRI and DSA were included. Popliteal and infrapopliteal arteries were classified as patent or occluded across three sub-segments (e.g., AT1=proximal, AT2=mid, AT3=distal). Two blinded reviewers independently evaluated all images. Participants were stratified by GFR into <30, 30-59, ≥60 mL/min/1.73 m² groups. Primary outcome was the proportion of patent arterial segments identified by QISS compared to DSA across these GFR categories. Secondary outcomes included TASC and infrapopliteal GLASS score changes between modalities.
Results or Findings: Among 57 patients (752 vessel segments), QISS MRI demonstrated a significantly higher proportion of patent segments compared to DSA (67.2% vs 57.3%, p<0.001). Differences were significant in the <30 and ≥60 GFR groups, but not in the 30-59 group. TASC and GLASS scores were significantly downgraded using QISS MRI (mean TASC 2.2 vs 2.3, p=0.049; GLASS 2.6 vs 2.2, p<0.001).
Conclusion: QISS MRI identifies more patent vessel segments than DSA, especially in patients with advanced renal dysfunction. As earlier studies using older non-contrast MR techniques have shown comparable outcomes in bypasses to arteries visible only on MRI but not on DSA after three years, these results further support incorporating QISS MRI into routine limb-salvage assessment protocols.
Limitations: Single-center design, modest sample size
Funding for this study: Jerold B. Katz Academy of Translational Science under project number 15790002 (recipient's name: Trisha Roy), the American Heart Association Transformational Award under project ID 17590004 (recipient's name: Trisha Roy), and the National Institutes of Health Research Project grant (R01) under award number R01HL174587 (recipient's name: Trisha Roy).
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: Study ID 15790002