Scholarlink

Trauma to the ribcage necessitates prompt and precise diagnosis to ensure effective treatment. Rib fractures, commonly resulting from incidents such as motor vehicle accidents, falls, or sports injuries, can lead to severe complications, including damage to vital organs like the heart and lungs. These injuries require immediate medical attention to prevent life-threatening outcomes. Manual annotation of rib fractures in CT scans remains a labor-intensive and error-prone process, heavily reliant on the radiologist‘s expertise. We propose a deep learning framework that combines a SWIN transformer-based generator and generative adversarial networks (GANs) to enhance rib fracture segmentation in CT scans. Our model significantly improves segmentation accuracy, achieving a Dice score of 71.55, outperforming the state-of-the-art FracNet+ by more than 4% on the RibFrac benchmark. This framework demonstrates the potential for more efficient and accurate automated detection of rib fractures, contributing to better clinical workflows. While the results are promising, there remains considerable scope for further exploration in the 3D Medical Image Domain, especially with the advent of newer AI models that could further enhance detection and segmentation capabilities.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Transformer-GAN Enhanced Rib Fracture Segmentation: Integrating Swin UNET3D with Adversarial Learning

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Samarth Turmari,
  • Chinmay Sultanpuri,
  • Sanskruti Kagawade,
  • Narendra Kaliwal,
  • Sneha Varur,
  • Channabasappa Muttal

摘要

Trauma to the ribcage necessitates prompt and precise diagnosis to ensure effective treatment. Rib fractures, commonly resulting from incidents such as motor vehicle accidents, falls, or sports injuries, can lead to severe complications, including damage to vital organs like the heart and lungs. These injuries require immediate medical attention to prevent life-threatening outcomes. Manual annotation of rib fractures in CT scans remains a labor-intensive and error-prone process, heavily reliant on the radiologist‘s expertise. We propose a deep learning framework that combines a SWIN transformer-based generator and generative adversarial networks (GANs) to enhance rib fracture segmentation in CT scans. Our model significantly improves segmentation accuracy, achieving a Dice score of 71.55, outperforming the state-of-the-art FracNet+ by more than 4% on the RibFrac benchmark. This framework demonstrates the potential for more efficient and accurate automated detection of rib fractures, contributing to better clinical workflows. While the results are promising, there remains considerable scope for further exploration in the 3D Medical Image Domain, especially with the advent of newer AI models that could further enhance detection and segmentation capabilities.