RST2G: Residual-Guided Spatiotemporal Transformer Graph Fusion Enhancement for Breast Cancer Segmentation in DCE-MRI.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Oncology

Background:

• Accurate breast tumor segmentation in dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) is crucial for breast cancer management.
• Tumor heterogeneity and complex contrast enhancement dynamics pose significant challenges for current segmentation methods.

Purpose of the Study:

• To develop a novel framework, the residual-guided spatiotemporal transformer with graph fusion enhancement (RST2G), for precise breast tumor segmentation in DCE-MRI.
• To enhance the capture of inter-temporal kinetic information and spatiotemporal dependencies for improved segmentation accuracy.

Main Methods:

• The RST2G framework utilizes pre-contrast, post-contrast, and residual MRI data.
• A weight-sharing hybrid encoder combines convolutional neural networks and vision transformers for feature extraction.
• Spatiotemporal graph enhancement is achieved through modality-specific graphs and inter-slice/inter-temporal attention mechanisms.

Main Results:

• RST2G demonstrated superior performance compared to state-of-the-art 2D, 3D, and 4D segmentation methods on two public DCE-MRI datasets.
• The framework effectively captures complex spatiotemporal tumor characteristics.

Conclusions:

• The proposed RST2G framework offers a significant advancement in precise breast tumor segmentation using DCE-MRI.
• RST2G shows potential for improving clinical breast cancer diagnosis, treatment planning, and monitoring.