University of Twente Student Theses
As of Friday, 8 August 2025, the current Student Theses repository is no longer available for thesis uploads. A new Student Theses repository will be available starting Friday, 15 August 2025.
Automated Liver Vein Segmentation in Preoperative Roadmapping Using Non-Contrast MRI
Sallevelt, J.C.J. (2025) Automated Liver Vein Segmentation in Preoperative Roadmapping Using Non-Contrast MRI.
|
PDF
21MB |
| Abstract: | Background: Budd-Chiari Syndrome (BCS) is a rare hepatic vascular disorder characterised by obstruction of hepatic venous outflow. Detailed venous imaging is crucial for preoperative planning. Current imaging techniques often rely on ionising radiation combined with iodine contrast medium, posing health risks to patients. Magnetic Resonance Angiography (MRA) is typically performed using gadolinium-based contrast agents, which also carry potential risks. This study investigates whether a deep learning model can accurately segment the hepatic veins on non-contrast MRI using a balanced Steady-State Free Precession (b-SSFP) sequence. Methods: An nnU-Net v2 model was trained on 23 annotated datasets acquired at Radboud University Medical Center (Nijmegen, Netherlands) using a 3 Tesla MRI scanner. Model validation was performed on data from four healthy volunteers scanned with a 1.5T MRI at the University of Twente (Enschede, Netherlands). Quantitative analysis was performed by calculating an average Dice score, sensitivity, and specificity. For qualitative analysis, a 3D reconstruction was created for visual assessment. Additionally, a 3D-printed silicone liver phantom with hollow veins was developed, and a catheter-based intervention was simulated as an in-vitro experiment to assess clinical applicability. Results: The model achieved an average Dice score of 0.88, sensitivity of 0.85 and a specificity of 1.00 on the independent validation set. Visual analysis showed accurate segmentation of the major hepatic veins. Discrepancies between AI- and manual segmentation were found in smaller peripheral branches, which can be filtered out through post-processing. Conclusion: The trained model can accurately segment veins on non-contrast MRI. This technique has potential for clinical implementation in preoperative roadmapping and, in the future, intraoperative guidance, offering a safer alternative to contrast-enhanced imaging. |
| Item Type: | Essay (Bachelor) |
| Faculty: | TNW: Science and Technology |
| Subject: | 31 mathematics, 33 physics, 42 biology, 44 medicine, 50 technical science in general |
| Programme: | Biomedical Technology BSc (56226) |
| Link to this item: | https://purl.utwente.nl/essays/106988 |
| Export this item as: | BibTeX EndNote HTML Citation Reference Manager |
Show download statistics for this publication
Show download statistics for this publicationRepository Staff Only: item control page