Abstract
Mechanical thrombectomy is a minimally invasive endovascular procedure for the treatment of acute ischemic stroke caused by vessel occlusions. The procedure relies on real-time fluoroscopic guidance, yet stent retrievers are often difficult to visualize because of their low radiopacity. Supervised learning could support automatic device localization, but annotated clinical fluoroscopy data are scarce due to privacy constraints and costly pixel-level annotation. To address this limitation, we propose an automatic pipeline for generating synthetic neurovascular fluoroscopy sequences. The pipeline simulates randomized mechanical thrombectomy devices within patient head CT volumes to generate annotated fluoroscopy sequences. Device geometry, curvature, material properties, position, and deployment configuration vary randomly to increase dataset diversity. This process produces a large-scale annotated fluoroscopy dataset suitable for supervised training of segmentation models. The sequences model microcatheter retraction and stent deployment. For stent retriever segmentation, RF-DETR achieved a Dice score of 0.683 with clinical-only training and 0.593 with synthetic-only training. Synthetic pretraining followed by clinical fine-tuning improved performance to a Dice score of 0.738, while training on a combined clinical and synthetic dataset achieved a comparable Dice score of 0.736. These results indicate that synthetic fluoroscopy is useful for stent retriever segmentation, both as a pretraining source and when combined with clinical data for training, although a synthetic-to-clinical domain gap remains.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/SWITCH_020.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: Not Submitted
Link to Open Review
Open Review Page: https://openreview.net/forum?id=uBoRTVuTGB
BibTex
@InProceedings{ShoMer_An_MICCAISAT2026,
author = { Shoaei Taklimi, Mersad AND Atayi, Fateme AND Scheuplein, Joshua AND Buessen, Anne Tjorven AND Tasharofi, Farid AND Preuhs, Elisabeth AND Denzinger, Felix AND Maier, Andreas},
title = { { An Automated Framework for Synthetic Neurovascular Fluoroscopy Generation and Stent Retriever Segmentation } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 17255},
month = {pending},
page = {pending}
}