Abstract
Rib fractures are prevalent traumatic injuries that may lead to severe complications if undetected. Precise detection and classification of fracture types on CT scans are therefore essential for timely clinical intervention. However, existing methods struggle with subtle fracture patterns and the intricate curvilinear anatomy of ribs, resulting in limited detection and classification accuracy. To address this, we propose MVANet, a two-stage Multi-View Aggregation framework for rib fracture diagnosis. In the first stage, an anatomy-aware detection module localizes candidate fracture regions by leveraging rib masks as spatial priors, effectively suppressing background false positives. In the second stage, a trajectory-aligned multi-view sampler (TAMS) extracts geometry-aligned 2D projections along each candidate’s rib centerline, which are encoded by a self-supervisedly adapted vision foundation model. A latent-query multi-view attention (LQMV-Attn) module further aggregates these multi-view features into a discriminative fracture-level representation for fine-grained classification. Experiments on our in-house MaskRib dataset and the public RibFrac dataset demonstrate that MVANet consistently outperforms existing methods in both detection sensitivity and classification accuracy, highlighting its potential for clinical translation. Code and models are publicly available at https://github.com/preciousluo/Beyond-Single-View.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/MLMI_028.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=5PoSqbTrmU
BibTex
@InProceedings{LuoLei_Beyond_MICCAISAT2026,
author = { Luo, Lei AND Luo, Tao AND Yang, Jiancheng AND Wu, Dijia AND Cui, Zhiming},
title = { { Beyond Single View: Anatomy-Aware Rib Fracture Diagnosis with Multi-View Aggregation } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 17272},
month = {pending},
page = {pending}
}