Abstract:The accurate segmentation of brain tumor is significant in clinical practice. Convolutional Neural Network (CNN)-based methods have made great progress in brain tumor segmentation due to powerful local modeling ability. However, brain tumors are frequently pattern-agnostic, i.e. variable in shape, size and location, which can not be effectively matched by traditional CNN-based methods with local and regular receptive fields. To address the above issues, we propose a shape-scale co-awareness network (S2CA-Net) for brain tumor segmentation, which can efficiently learn shape-aware and scale-aware features simultaneously to enhance pattern-agnostic representations. Primarily, three key components are proposed to accomplish the co-awareness of shape and scale. The Local-Global Scale Mixer (LGSM) decouples the extraction of local and global context by adopting the CNN-Former parallel structure, which contributes to obtaining finer hierarchical features. The Multi-level Context Aggregator (MCA) enriches the scale diversity of input patches by modeling global features across multiple receptive fields. The Multi-Scale Attentive Deformable Convolution (MS-ADC) learns the target deformation based on the multiscale inputs, which motivates the network to enforce feature constraints both in terms of scale and shape for optimal feature matching. Overall, LGSM and MCA focus on enhancing the scale-awareness of the network to cope with the size and location variations, while MS-ADC focuses on capturing deformation information for optimal shape matching. Finally, their effective integration prompts the network to perceive variations in shape and scale simultaneously, which can robustly tackle the variations in patterns of brain tumors. The experimental results on BraTS 2019, BraTS 2020, MSD BTS Task and BraTS2023-MEN show that S2CA-Net has superior overall performance in accuracy and efficiency compared to other state-of-the-art methods. Code: https://github.com/jiangyu945/S2CA-Net.

Structure-aware Shape Correspondence Network for 3D Shape Synthesis.

3D Graph-S<SUP>2</SUP>Net: Shape-Aware Self-ensembling Network for Semi-supervised Segmentation with Bilateral Graph Convolution

Shape-Scale Co-Awareness Network for 3D Brain Tumor Segmentation.

Electors Voting for Fast Automatic Shape Correspondence

SADRNet: Self-Aligned Dual Face Regression Networks for Robust 3D Dense Face Alignment and Reconstruction

StructureNet: Hierarchical Graph Networks for 3D Shape Generation

Anisotropic Multi-Scale Graph Convolutional Network for Dense Shape Correspondence

NeuSDFusion: A Spatial-Aware Generative Model for 3D Shape Completion, Reconstruction, and Generation

Unsupervised Template-assisted Point Cloud Shape Correspondence Network

Learning Local Neighboring Structure for Robust 3D Shape Representation

Learning Part Generation and Assembly for Structure-aware Shape Synthesis

Learning Descriptor Networks for 3D Shape Synthesis and Analysis

Networks for 3 D Shape Synthesis and Analysis

DAE-Net: Deforming Auto-Encoder for fine-grained shape co-segmentation

Assembly-Based 3D Modeling Using Graph Convolutional Neural Networks.

Deep Shape Constrained Network for Robust Face Alignment.

Unsupervised 3D Learning for Shape Analysis Via Multiresolution Instance Discrimination

Unsupervised 3D Learning for Shape Analysis via Multiresolution Instance Discrimination

Shape Completion using 3D-Encoder-Predictor CNNs and Shape Synthesis

STD-Net: Structure-preserving and Topology-adaptive Deformation Network for Single-View 3D Reconstruction