ESR Research Seed Grants 2024

Contrast-enhanced Imaging

• SeqSynth: synthetic neuro-oncological post-contrast sequence synthesis by deep generative modelling

Summary: Imaging surveillance in cancer treatment is a key determinant of downstream neuro-oncological care across adult and paediatric patients, conventionally informed by pre- and post-contrast structural imaging. Contrast agents cannot always be given, such as with patient contraindications from allergy or renal failure. Imaging can also be corrupted by artefacts, making its radiological interpretation difficult. Furthermore, the ability to minimise repeated gadolinium use in these cohorts, of note especially
applicable to paediatric patients, is highly desirable, though realistically not often feasible, for the risk of missing an abnormality from lack of available postcontrast imaging outweighs the risk of repeated gadolinium administration. However, we have recently shown that deep learning models can accurately identify enhancing tumours without contrastenhanced imaging (Figure 1)1. Therefore, deep models can potentially leverage non-contrast MRI sequence information to derive post-contrast information from unenhanced scans. Here, we prototype an enhancing disease triaging system using only unenhanced MRI data, harmonising state-of-the-art practices in deep learning with largescale unselective neuro-oncology data.

James Ruffle
University College London/UK

• Utilizing Deep Learning for Opportunistic Cardiac Output Estimation from Bolus-Tracking Data

Summary: This project explored the feasibility of using deep learning techniques to estimate cardiac output, specifically Ejection Fraction (FE), from standard bolus tracking (BT) sequences from CT scans. Conventionally used only for scan timing, BT data may contain untapped hemodynamic information. We hypothesized that a DL model could provide quantitative and automatic estimations, comparable to standard methods like geometrical measurements derived from Cardiac CT (CCT). A bicentric retrospective dataset of CCT scans with corresponding BT sequences was collected. DL models, specifically hybrid Convolutional-Recurrent Neural Networks (CNN-RNNs) using transfer learning, were developed and trained to classify FE based on the CCT-EF ground truth. Results demonstrated the feasibility of this approach, with models achieving promising performance (AUC up to ~0.90) in internal cross-validation, particularly when using data augmentation. Performance decreased on an independent external dataset, highlighting the need for larger, multi-center datasets for robust generalization in DL solutions. This opportunistic method can enrich vascular CT reports with valuable functional data without additional scans or radiation and further research in this direction is advocated.

Mario Laudazi Hospital Universitario y Politécnico La Fe, Valencia/ES

• Sustainable Iodine Contrast Media for a Healthy Future (SUSTAINED)

Summary: Iodinated contrast media (ICM) are essential radiological tools; they are used to highlight structures, save lives, and are the most administered drugs in radiology. Iodine is a rare element and global supply issues have been reported. Such supply issues are a challenge for healthcare, additionally extracting iodine has high environmental costs and waste iodine has been identified in surface and drinking water. Medical ICM users were surveyed internationally (eight languages) to understand current knowledge, attitudes, and practices regarding ICM sustainability. International survey respondents (n=231) indicated that safe and responsible disposal of ICM will increase in priority in the future. Responses also highlight the need for alternatives to iodine. Formal conservation strategies appeared to be infrequently available at institutional levels. Opportunities exist in terms of the development of methods and processes to recycle and reuse ICM leftovers. Strategies, local and national, are also needed to focus on collection and the potential recycling of ICM excretions. More widely, it would be highly advantageous for stronger direction at a European level on promoting and operationalizing ICM sustainability. Such an approach should also consider the training requirements for all stakeholders in terms of correct ICM usage, but also with focus on sustainability.

Andrew England
University College Cork – National University of Ireland, Cork/IE

• Intra-arterial Computed Tomography liver perfusion (IA CT liver perfusion) as an aid for dose planning of colorectal liver metastasis (mCRC)

Summary: Injection of 99mTc-macro-aggregated albumin (99mTc-MAA) followed by SPECT/CT is used to simulate the Yttrium (90Y) microsphere distribution for radioembolization. The 99mTc-MAA distribution is used for activity planning. While literature shows good agreement between 99mTc-MAA dose distribution for normal liver parenchyma, it indicates poor agreement for tumor tissue in liver metastatic cancer, often resulting in lower than expected absorbed doses after 90Y radioembolization. Therefore, new methods for activity planning for radioembolization are necessary. CT perfusion scanning is a dynamic technique that measures changes in contrast. Although CT liver perfusion with intravenous contrast exists, it requires long scan times due to the liver’s dual blood supply. In interventional radiology it is possible to make a CT liver perfusion scan with intra-arterial contrast, making it possible to measure the arterial perfusion of the liver and tumors in a much shorter scanning time, reducing artefacts and lower radiation burden. The aim of this study is to develop an intra-arterial CT liver perfusion protocol and investigate correlation of perfusion variables with 90Y microsphere distribution.

Fernando Gómez Muñoz, Charlotte Schneider, Lisa Klompenhouwer
Netherlands Cancer Institute, Amsterdam/NL

• Surveillance Without Contrast – Diagnostic Effectiveness of SMI Ultrasound in Detecting Endoleaks after EVAR

Summary: Endovascular aneurysm repair (EVAR) requires long-term imaging surveillance to identify possible late complications. Computed tomography angiography (CTA) is currently the gold standard but involves exposure to ionizing radiation and nephrotoxic contrast agents. Doppler ultrasound presents a safer and more accessible alternative; however, it has been shown to have lower sensitivity in detecting endoleaks unless it involves ultrasound contrast injection (CEUS). Microvascular ultrasound (MVUS) is a novel, contrast-free modality capable of detecting fine blood flow, making it potentially valuable for post-EVAR surveillance. Between July 2024 and May 2025, 65 patients who underwent EVAR were enrolled in a prospective study. Each patient underwent both MVUS and CEUS examinations within one week after the procedure, and CTA one month later. MVUS and CEUS jointly detected 38 endoleaks, with MVUS alone identifying an additional 10 cases. These additional endoleaks displayed subtle, linear flow patterns oriented perpendicular to the ultrasound beam. CTA confirmed 30 of the 38 endoleaks initially seen by both MVUS and CEUS. Among the MVUS-only findings, two were confirmed by CTA. MVUS demonstrated high sensitivity for subtle endoleaks and may represent a promising contrast-free imaging tool. Its implementation could reduce dependency on CTA, enhancing patient safety in long-term post-EVAR monitoring.

Publications:
Czeczelewski, M. et al.  Diagnostic accuracy of microvascular flow imaging ultrasound for endoleak detection after endovascular aortic aneurysm repair: a systematic review and meta-analysis*. Polish Journal of Radiology. 2024; 89:414–419. doi:10.5114/pjr/190502.

Marcin Czeczelewski
Medical University of Lublin, Lublin/PL

• Longitudinal Assessment of Renal Allografts in Donor-Recipient Pairs by Perfusion Quantification Using Arterial Spin Labeling (ASL-MRI)

Summary: The non-invasive, short-lived nature of the tracer, combined with the rapid acquisition time, makes renal arterial spin labelling (ASL) MRI inherently suited to repeated measurements over a range of timescales that would not be possible with any other perfusion technique in humans. Ten living, related donors and dedicated paediatric recipients underwent renal cortical volume and ASL-MRI measurements prior to and 3, 6, and 12 months after transplantation in order to quantify cRBF. Baseline values were compared to post-transplant values and correlated to estimated glomerular filtration rate (eGFR). In addition, renal plasma flow (RPF) and estimated filtration fraction (eFF) were calculated to approximate filtration. Cortical volume, cRBF, and eGFR demonstrated a highly significant increase at 3 months compared to baseline and subsequently showed consistent values until the end of the observed period. However, at 3 months, four patients, and at 6 months, three patients had an eGFR < 60 ml/min/1.73 m2, suggesting allograft nephropathy, with significantly lower eGFR and cRBF compared to recipients with normal allograft function. By 12 months, the mean eGFR had increased to normal values in all recipients, suggesting normal allograft function, which was clinically reflected in significant patient growth. Estimated FF did not change significantly throughout the entire observed post-transplant period compared to baseline, which excluded hyperfiltration as an adaptive process.Our study identifies notable adaptive changes in cortical perfusion and renal plasma flow as significant biomarkers of renal function reserve activation during the early post-transplant period in clinically stable pediatric recipients.

Tijana Radovic
University Children’s Hospital Belgrade, Belgrade/RS

• Blood-Brain Barrier Dysfunction as a Marker of Tissue Face in Cerebral Venous Thrombosis

Expected Impact: Our project will be the first to assess DCE perfusion in CVT patients. We expect to find BBB dysfunction metrics in the acute phase of CVT that are predictive of tissue fate. Our results have the potential to be used in future models of patient selection for endovascular treatment, thereby possibly contributing to a more personalized approach in CVT. By correctly identifying patients that are more likely to develop definitive brain injury, who potentially could benefit from invasive treatment, we will be able to reduce brain damage and sequela, while avoiding overtreatment of the remaining cases, thus improving disease prognosis.

Sara Daniela Botelho Rosa
Liga dos Amigos do Hospital de Santa Marta (IPSS) Lisbon/PT