Multi-View Deformable Convolution Meets Visual Mamba for Coronary Artery Segmentation

Xiaochan Yuan; Pai Zeng

cs.CVMar 23, 2026

Multi-View Deformable Convolution Meets Visual Mamba for Coronary Artery Segmentation

Authors: Xiaochan Yuan, Pai Zeng

Organizations: Sichuan Agricultural University

Abstract

Accurate segmentation of coronary arteries from computed tomography angiography (CTA) images is of paramount clinical importance for the diagnosis and treatment planning of cardiovascular diseases. However, coronary artery segmentation remains challenging due to the inherent multi-branching and slender tubular morphology of the vasculature, compounded by severe class imbalance between foreground vessels and background tissue. Conventional convolutional neural network (CNN)-based approaches struggle to capture long-range dependencies among spatially distant vascular structures, while Vision Transformer (ViT)-based methods incur prohibitive computational overhead that hinders deployment in resource-constrained clinical settings. Motivated by the recent success of state space models (SSMs) in efficiently modeling long-range sequential dependencies with linear complexity, we propose MDSVM-UNet, a novel two-stage coronary artery segmentation framework that synergistically integrates multidirectional snake convolution (MDSConv) with residual visual Mamba (RVM). In the encoding stage, we introduce MDSConv, a deformable convolution module that learns adaptive offsets along three orthogonal anatomical planes -- sagittal, coronal, and axial -- thereby enabling comprehensive multi-view feature fusion that faithfully captures the elongated and tortuous geometry of coronary vessels. In the decoding stage, we design an RVM-based upsampling decoder block that leverages selective state space mechanisms to model inter-slice long-range dependencies while preserving linear computational complexity. Furthermore, we propose a progressive two-stage segmentation strategy: the first stage performs coarse whole-image segmentation to guide intelligent block extraction, while the second stage conducts fine-grained block-level segmentation to recover vascular details and suppress false positives..

Figures & tables

Figure 1: Overview of the proposed two-stage coronary artery segmentation framework. In Stage 1, the input CTA volume is downsampled and segmented using MDSVM-UNet to produce a coarse segmentation map. The coarse results guide the extraction of $64\times 64\times 64$ voxel blocks from the original resolution. In Stage 2, MDSVM-UNet performs fine-grained block-level segmentation, and the results from all blocks are merged to produce the final output.

Figure 2: Detailed architecture of MDSVM-UNet. (a) The overall encoder-decoder network with UNet++-style dense skip connections. The encoder comprises four MDSConv blocks and one RVM block, while the decoder consists of three RVM blocks and a convolutional output layer. (b) The Residual Visual Mamba (RVM) layer with residual connection and scaling factor. (c) The Vision State Space Module (VSSM) with dual-branch parallel processing.

Figure 3: Architecture of the Multidirectional Snake Convolution (MDSConv) layer. Input features are processed by standard convolution and three axis-specific snake convolutions (along $x$ , $y$ , $z$ axes), followed by concatenation and fusion.

\begin{cases}(x_{i+c},y_{i+c},z_{i+c})=(x_{i+c},y_{i}+\sum_{i}^{i+c}\Delta y,z_{i}+\sum_{i}^{i+c}\Delta z)\\ (x_{i-c},y_{i-c},z_{i-c})=(x_{i-c},y_{i}+\sum_{i}^{i-c}\Delta y,z_{i}+\sum_{i}^{i-c}\Delta z)\end{cases}

(2)

Table 4

\begin{cases}(x_{j+c},y_{j+c},z_{j+c})=(x_{j}+\sum_{j}^{j+c}\Delta x,y_{j+c},z_{j}+\sum_{j}^{j+c}\Delta z)\\ (x_{j-c},y_{j-c},z_{j-c})=(x_{j}+\sum_{j}^{j-c}\Delta x,y_{j-c},z_{j}+\sum_{j}^{j-c}\Delta z)\end{cases}

(3)

Table 5

Method	Params (M)	Loss	DSC $\uparrow$	HD $\downarrow$	AHD $\downarrow$
Method	Params (M)	Loss
FCN Lei et al. (2020)	20.0	$\mathcal{L}_{\text{dice}}$	0.5755	42.8031	1.6885
		$\mathcal{L}_{\text{focal}}$	0.5573	37.0958	1.7619
3D U-Net Çiçek et al. (2016)	25.0	$\mathcal{L}_{\text{dice}}$	0.6599	31.1518	1.1167
		$\mathcal{L}_{\text{focal}}$	0.6521	30.9527	1.1966
TS ConvGRU Kong et al. (2020)	35.2	$\mathcal{L}_{\text{dice}}$	0.6878	30.3350	1.4336

Table 1: Quantitative comparison of single-stage segmentation methods on the ImageCAS dataset. Bold values indicate the best performance in each column.

\uparrow

indicates higher is better;

\downarrow

indicates lower is better.

Method	Params (M)	Loss	DSC $\uparrow$	HD $\downarrow$	AHD $\downarrow$
Method	Params (M)	Loss
Two-Stage 3D U-Net Çiçek et al. (2016)	23.3	$\mathcal{L}_{\text{dice}}$	0.7201	40.9693	3.0686
		$\mathcal{L}_{\text{focal}}$	0.7183	32.7724	4.4687
Two-Stage DSU-Net Qi et al. (2023)	29.0	$\mathcal{L}_{\text{dice}}$	0.7790	30.1524	0.9092
		$\mathcal{L}_{\text{focal}}$	0.7720	36.4390	1.7171
Two-Stage LightM-UNet Liao et al. (2024)	5.5	$\mathcal{L}_{\text{dice}}$	0.8079	28.1864	0.9488

Table 2: Quantitative comparison of two-stage segmentation methods on the ImageCAS dataset. Bold values indicate the best performance in each column. “–” indicates experiments not conducted.

Method	DSC $\uparrow$	HD $\downarrow$	AHD $\downarrow$
Baseline	0.7824	36.3886	1.7116
+ MDSConv	0.8241	27.5515	0.9046
+ RVM	0.8236	27.6082	0.9192
+ MDSConv + RVM	0.8365	27.8430	0.9023

Table 3: Ablation study results on the ImageCAS dataset. Bold values indicate the best performance.

Explore similar work

Aug 25, 2026

MoE-based Feature Adapter for Prompt-free Binary Coronary Artery Segmentation in X-ray Angiography

Accurate segmentation of coronary arteries in X-ray angiography videos is essential for quantitative coronary analysis and image-guided interventions. However, accurate segmentation remains challenging because coronary vessels are thin and exhibit low contrast, while the presence of catheters, guidewires, and complex anatomical background structures can further interfere with vessel delineation. Existing U-Net- and Transformer-based models provide strong baselines, but their shared feature-adaptation pathways may be insufficient for heterogeneous angiographic appearances. In this paper, we propose a prompt-free mixture-of-experts (MoE) feature adapter for binary coronary artery segmentation. Built upon parameter-efficient Vision Transformer adapters, the proposed method uses multiple lightweight experts with input-dependent top-

k

routing to adaptively refine vessel-related features while limiting active computational cost. Experiments on MOSXAV and external evaluation on XACV show that the proposed method outperforms representative baselines and improves cross-dataset generalisation. These results suggest that MoE-based adapter learning is effective for robust coronary artery segmentation in X-ray angiography videos.

Lin Xi, Yingliang Ma

University College London, United Kingdom · University of East Anglia, United Kingdom · King’s College London, United Kingdom

Coronary Artery Segmentation Coronary Angiography Mixture of Experts Vision Foundation Model Adaptation

Ask PDF

Apr 7, 2025

MSA-UNet3+: Multi-Scale Attention UNet3+ with New Supervised Prototypical Contrastive Loss for Coronary DSA Image Segmentation

Accurate segmentation of coronary Digital Subtraction Angiography (DSA) images is essential for diagnosing and treating coronary artery disease (CAD). Despite advances in deep learning, challenges such as high intra-class variance and class imbalance limit precise vessel delineation. Existing approaches for coronary DSA segmentation cannot effectively address these issues. Furthermore, existing segmentation network encoders do not directly generate semantic embeddings, which could enable the decoder to reconstruct segmentation masks more effectively. We propose a Supervised Prototypical Contrastive Loss (SPCL) that combines supervised and prototypical contrastive learning to enhance coronary DSA image segmentation. The supervised contrastive loss enforces semantic embeddings in the encoder, improving feature differentiation. The prototypical contrastive loss enables the model to focus on the foreground class while alleviating high intra-class variance and class imbalance by concentrating only on hard-to-classify background samples. We implement the proposed SPCL within MSA-UNet3+, a Multi-Scale Attention-Enhanced UNet3+ architecture. The architecture integrates a Multi-Scale Attention Encoder (M-encoder), a Multi-Scale Dilated Bottleneck (MSD-Bottleneck) for multi-scale feature extraction, and a Contextual Attention Fusion Module (CAFM) to preserve fine-grained details while improving contextual understanding. Experiments on a private coronary DSA dataset demonstrate that MSA-UNet3+ outperforms state-of-the-art methods, achieving the highest Dice coefficient and F1-score while significantly reducing ASD and ACD. The framework provides precise vessel segmentation for accurate identification of coronary stenosis and supports informed diagnostic and therapeutic decisions. The code will be released at https://github.com/rayanmerghani/MSA-UNet3plus.

Rayan Merghani Ahmed, Adnan Iltaf, Mohamed Elmanna +5

Shenzhen Institutes of Advanced Technology (SIAT), Chinese Academy of Sciences (CAS), Shenzhen, 518055, China · University of Chinese Academy of Sciences (UCAS), Beijing, 101408, China · Department of Biomedical Engineering and Systems, Faculty of Engineering, Cairo University, Cairo, 12613, Egypt +2

Contrastive Learning Medical Imaging Coronary Angiography Multi-Scale Feature Fusion Medical Image Segmentation Coronary Artery Segmentation+1

Ask PDF

Jun 29, 2026

HTC-SGA Former: A Hybrid Transformer-CNN Network with Self-Guided Attention and a New Boundary-Weighted Adaptive Loss for Coronary DSA Vessel Segmentation

Accurate coronary Digital Subtraction Angiography (DSA) vessel segmentation is essential for computer-aided diagnosis and treatment planning of coronary artery disease (CAD). However, thin low-contrast vessels, background interference, and severe vessel-background class imbalance make reliable segmentation of weak distal branches and vessel boundaries challenging. Existing methods struggle to balance global contextual reasoning with preservation of weak vessels, vessel continuity, and fine boundaries. To address these limitations, we propose HTC-SGA Former, a lightweight hybrid Transformer-CNN framework for coronary DSA vessel segmentation. It employs a CNN encoder for local vessel morphology extraction and a Transformer decoder for contextual feature modeling. A Multi-Scale Global-Local Window Attention (MS-GLWA) block performs efficient global-local contextual modeling, while a Self-Guided Feature Attention (SGFA) module enhances weak-vessel responses. In addition, a Boundary-Weighted Adaptive Compound Loss (BWACL) emphasizes thin-vessel boundaries and adaptively balances vessel recovery and boundary refinement. Experiments on private right and left coronary artery DSA subsets show that HTC-SGA Former outperforms 14 state-of-the-art segmentation methods while maintaining a compact architecture with only 0.81M parameters. BWACL also improves performance over binary cross-entropy and Dice losses across four encoder-decoder architectures, demonstrating strong cross-backbone applicability. HTC-SGA Former improves thin-vessel recovery, vessel continuity, and boundary localization through complementary global-local contextual modeling, vessel-focused refinement, and adaptive optimization, supporting reliable and computationally efficient coronary vessel analysis for future computer-assisted cardiovascular interventions.

Rayan Merghani Ahmed, Marwa Omer Mohammed Omer, Mohamed Elmanna +3

Medical Imaging Coronary Angiography Hybrid CNN-Transformer Architectures Medical Image Segmentation Coronary Artery Segmentation Boundary-Aware Segmentation

Ask PDF