Abstract
One of the main bottlenecks of training neural networks in a segmentation task in medical imaging is the scarcity of annotated data, and the rare occurrence of some configurations. This is the case, for example, in the TopAneu2026 machine learning challenge, where participants are asked to detect Intracranial Aneurysms (ICAs) in 3D MRA. In this work, we use a new version of a synthetic procedural model, Vascular Model V2 (VaMosV2), based on previously published VaMos model, that generates full 3D synthetic MRA volumes with precise voxel-wise segmentation masks of vessels and ICAs. Our model takes a real 3D MRA and its vessel and ICA segmentation as input, and generates synthetic examples of 3D volumes. It can modify the vascular geometry and add or remove vessels and ICAs. We detail the main improvements of VaMosV2 in this work and show how it can be used to perform data augmentation on the training dataset of the TopAneu2026 data challenge. We plan to open source VaMosV2 code by the end of TopAneu2026 challenge.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/TopAneu_031.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=WsA9bRDqPp
BibTex
@InProceedings{MarTan_VaMosV2_MICCAISAT2026,
author = { Marchand, Tanguy AND Rouet, Laurence AND Bourcier, Romain AND Autrusseau, Florent},
title = { { VaMosV2: A Multi-step Procedural Model That Generates Full 3D Synthetic MRA Volumes for Data Augmentation in the TopAneu ICA Segmentation Challenge } },
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}
}
