Scholarlink

This study proposes an intuitive and high-precision visualization system that integrates machine learning-based segmentation of CT images with mixed reality (MR) technology. Traditional surface rendering techniques often have limitations, such as including irrelevant structures due to similar densities, dependence on threshold settings, and loss of anatomical detail. To address these issues, we utilized MONAI Label to automatically segment key anatomical structures from abdominal CT scans and constructed 3D surface models for each segmented organ. The system incorporates a spatial reality display (SONY ELF-SR2) and a gesture-based motion capture device (Leap Motion Controller 2) to enable users to manipulate 3D models intuitively without physical contact. Reconstructed cross-sectional images and segmented annotations are displayed as 3D textures and overlaid on the surface models using stencil buffer techniques. Users can freely select and hide specific organs via finger gestures, and dynamically adjust cross-sectional positions to explore internal structures. Experimental results demonstrated that the proposed system significantly enhances spatial understanding of complex anatomical configurations. The system is particularly suited for preoperative planning, medical education, and patient communication by enabling selective visualization and intuitive manipulation. The results of this study are expected to serve as a foundation for new three-dimensional visualization technology that facilitates understanding of complex internal structures in clinical settings, enabling safer and more accurate diagnosis and surgical support.

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

Spatial Representation of Three-Dimensional X-ray CT Segmentation Data Using Machine Learning and Mixed Reality

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Hiroki Kase,
  • Katsuyuki Takagi,
  • Toru Aoki

摘要

This study proposes an intuitive and high-precision visualization system that integrates machine learning-based segmentation of CT images with mixed reality (MR) technology. Traditional surface rendering techniques often have limitations, such as including irrelevant structures due to similar densities, dependence on threshold settings, and loss of anatomical detail. To address these issues, we utilized MONAI Label to automatically segment key anatomical structures from abdominal CT scans and constructed 3D surface models for each segmented organ. The system incorporates a spatial reality display (SONY ELF-SR2) and a gesture-based motion capture device (Leap Motion Controller 2) to enable users to manipulate 3D models intuitively without physical contact. Reconstructed cross-sectional images and segmented annotations are displayed as 3D textures and overlaid on the surface models using stencil buffer techniques. Users can freely select and hide specific organs via finger gestures, and dynamically adjust cross-sectional positions to explore internal structures. Experimental results demonstrated that the proposed system significantly enhances spatial understanding of complex anatomical configurations. The system is particularly suited for preoperative planning, medical education, and patient communication by enabling selective visualization and intuitive manipulation. The results of this study are expected to serve as a foundation for new three-dimensional visualization technology that facilitates understanding of complex internal structures in clinical settings, enabling safer and more accurate diagnosis and surgical support.