Scholarlink

In the last several years, deep learning technology has brought many innovative advancements in the medical imaging sector along with other aspects. Modern analytical tools adopt transformer-based architectural design; this allows a researcher to develop special features to ensure high performance in the image examination process. Patients require accurate head and neck cancer diagnosis at appropriate intervals to receive improved patient care along with proper treatment development. The research analyzes deep learning transformer models that detect head and neck cancer inside computed tomography (CT) and magnetic resonance imaging (MRI) scanning systems. The system prototype utilizes state-of-the-art transformer models, TransUNet and Swin-UNet, to measure the segmentation accuracy for medical images. A validation step utilizes different CT/MRI image series obtained from various HNC patients for performing tests. The research evaluates transformer-based models to determine their capability in detecting HNC and explores performance relationships among multiple imaging modalities. This research supports the creation of computer-aided diagnosis (CAD) technology through its findings about using transformer networks to improve clinical detection of head and neck cancers. The Swin-UNet model reached 99.26% accuracy, 98.97% Precision, 98.91% Recall, and 98.94% F1score during testing of CT images with 80% training and 20% testing split while showing better results than TransUNet and competing imaging models.

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-Based Models for Head and Neck Cancer Detection Across Imaging Modalities

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Sunkara Naga Krishna Mohan Sai,
  • G. Jyotsna,
  • Bolem Ajay Babu

摘要

In the last several years, deep learning technology has brought many innovative advancements in the medical imaging sector along with other aspects. Modern analytical tools adopt transformer-based architectural design; this allows a researcher to develop special features to ensure high performance in the image examination process. Patients require accurate head and neck cancer diagnosis at appropriate intervals to receive improved patient care along with proper treatment development. The research analyzes deep learning transformer models that detect head and neck cancer inside computed tomography (CT) and magnetic resonance imaging (MRI) scanning systems. The system prototype utilizes state-of-the-art transformer models, TransUNet and Swin-UNet, to measure the segmentation accuracy for medical images. A validation step utilizes different CT/MRI image series obtained from various HNC patients for performing tests. The research evaluates transformer-based models to determine their capability in detecting HNC and explores performance relationships among multiple imaging modalities. This research supports the creation of computer-aided diagnosis (CAD) technology through its findings about using transformer networks to improve clinical detection of head and neck cancers. The Swin-UNet model reached 99.26% accuracy, 98.97% Precision, 98.91% Recall, and 98.94% F1score during testing of CT images with 80% training and 20% testing split while showing better results than TransUNet and competing imaging models.