SPG-CDENet: Spatial Prior-Guided Cross Dual Encoder Network for Multi-Organ Segmentation

TL;DR

SPG-CDENet addresses the challenge of accurate multi-organ segmentation amid large anatomical variability by introducing a two-stage paradigm that leverages spatial priors. The first stage uses a Spatial Prior Network to localize ROI, guiding a Crossing Dual Encoder Network that fuses global and ROI-focused local features via a Symmetric Cross-Attention module and a flow-based decoder. Empirical results on Synapse and ACDC show state-of-the-art performance with strong boundary accuracy and robustness, supported by thorough ablations validating each component. The approach offers a plug-and-play ROI prior mechanism and a scalable framework for precise organ delineation with potential extension to 3D segmentation.

Abstract

Multi-organ segmentation is a critical task in computer-aided diagnosis. While recent deep learning methods have achieved remarkable success in image segmentation, huge variations in organ size and shape challenge their effectiveness in multi-organ segmentation. To address these challenges, we propose a Spatial Prior-Guided Cross Dual Encoder Network (SPG-CDENet), a novel two-stage segmentation paradigm designed to improve multi-organ segmentation accuracy. Our SPG-CDENet consists of two key components: a spatial prior network and a cross dual encoder network. The prior network generates coarse localization maps that delineate the approximate ROI, serving as spatial guidance for the dual encoder network. The cross dual encoder network comprises four essential components: a global encoder, a local encoder, a symmetric cross-attention module, and a flow-based decoder. The global encoder captures global semantic features from the entire image, while the local encoder focuses on features from the prior network. To enhance the interaction between the global and local encoders, a symmetric cross-attention module is proposed across all layers of the encoders to fuse and refine features. Furthermore, the flow-based decoder directly propagates high-level semantic features from the final encoder layer to all decoder layers, maximizing feature preservation and utilization. Extensive qualitative and quantitative experiments on two public datasets demonstrate the superior performance of SPG-CDENet compared to existing segmentation methods. Furthermore, ablation studies further validate the effectiveness of the proposed modules in improving segmentation accuracy.

Figures (6)

• Figure 1: Conceptual comparison of the three most popular models used for medical image segmentation, where (a) classical U-Netronneberger2015unet; (b) Swin U-Netcao2021swinunet; (c) ParaTransCNN encodersun2024paratranscnn; (d)Hierarchical FCNroth2017hierarchical; (e)Our SPG-CDENet
• Figure 2: Overview of the SPG-CDENet
• Figure 3: Symmetric Cross-Attention Module
• Figure 4: Overview of the Flow-based Decoder. The decoder takes three inputs — semantic features from the previous decoder layer $F^{i-1}_{Dec}$, the encoder $F^{GE(6-1-i)}_{Enc}$ , and the Flow Block layer $F^{flow(i-1)}_{Dec}$, respectively.
• Figure 5: Qualitative results of different models on the Synapse dataset, from left to right: Input image; U-Netronneberger2015unet; Att-UNetoktay2018attentionunet; TransUNetchen2021transunet; Swin U-Netcao2021swinunet; R50 U-Netchen2021transunet; R50 Att-UNetchen2021transunet; FCTtragakis2023fullyconvolutional; MISSFormerhuang2022missformer; SPG-CDENet (Ours); Ground Truth.