Advanced gastrointestinal tract organ differentiation using an integrated swin transformer U-Net model for cancer care

The segmentation of gastrointestinal (GI) organs, including the stomach, small intestine, and large intestine, is crucial for radio oncologists to plan effective cancer therapy. This study presents an innovative semantic segmentation approach that integrates the Swin Transformer Block with the U-Net model to delineate healthy GI organs accurately using MRI data. The paper presents a novel approach that merges the Swin Transformer and U-Net models to leverage global context learning capabilities and fine-grained spatial resolution. Incorporating this integration greatly enhances the model’s capacity to achieve precise and comprehensive semantic segmentation, specifically in accurately outlining the gastrointestinal tract in MRI data. It utilizes the Swin Transformer, incorporating a shift-based windowing technique to gather contextual information efficiently while ensuring scalability. This novel architecture effectively balances local and global contexts, improving performance across various computer vision tasks, especially in medical imaging for segmenting the gastrointestinal tract. The model was trained and tested on the UW Madison GI Tract dataset, which comprises 38,496 MRI images from actual cancer cases. By leveraging the self-attention mechanisms of the Swin Transformer to capture global context and long-term dependencies, this approach combines the strengths of both models. The proposed architecture achieved a loss of 0.0949, a dice coefficient of 0.9190, and an Intersection over Union (IoU) score of 0.8454, demonstrating its effectiveness in providing high accuracy and robust performance. This technology holds significant potential for integration into clinical processes, enhancing the precision of radiation therapy for GI cancer patients.