Scholarlink

Purpose <p>Segmentation and three-dimensional model creation of the hepatic vasculature obtained from preoperative CT images can improve surgery planning. A comparative study is carried out to evaluate different methods for generating meshes of the hepatic vasculature.</p> Methods <p>Seven segmentation methods and 3 degrees of smoothing of the final mesh are compared both quantitatively and qualitatively. Portal venous CT images of 9 patients are segmented in DICOM to Print. The volume, surface and number of triangles are calculated and compared to a ground truth mesh, and 3 raters evaluate the meshes using a head-mounted virtual reality display.</p> Results <p>The results for quantitative metrics reach intraclass correlation coefficients of up to 0.79. Considerable differences between employed segmentation methods are identified when analyzing the qualitative evaluation by the raters.</p> Conclusion <p>The study highlights the need for a combination of automatic algorithms and manual adjustment or addition to obtain accurate and usable meshes for surgical planning.</p>

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

Comparative Study of Segmentation Algorithms for the Generation of CT-Based Preoperative Planning Meshes of the Liver Vessels

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年7月7日

查看全文
  • Juan Pedraja,
  • Alexander P. Seiffert,
  • Teresa Corpas-del Moral,
  • Saul Higuera,
  • Juan Andrés Echeverri,
  • Enrique J. Gómez,
  • Patricia Sánchez-González

摘要

Purpose

Segmentation and three-dimensional model creation of the hepatic vasculature obtained from preoperative CT images can improve surgery planning. A comparative study is carried out to evaluate different methods for generating meshes of the hepatic vasculature.

Methods

Seven segmentation methods and 3 degrees of smoothing of the final mesh are compared both quantitatively and qualitatively. Portal venous CT images of 9 patients are segmented in DICOM to Print. The volume, surface and number of triangles are calculated and compared to a ground truth mesh, and 3 raters evaluate the meshes using a head-mounted virtual reality display.

Results

The results for quantitative metrics reach intraclass correlation coefficients of up to 0.79. Considerable differences between employed segmentation methods are identified when analyzing the qualitative evaluation by the raters.

Conclusion

The study highlights the need for a combination of automatic algorithms and manual adjustment or addition to obtain accurate and usable meshes for surgical planning.