Scholarlink

Hip arthroscopy, a minimally invasive procedure for diagnosing and treating hip disorders, is gaining traction due to technological advancements. However, surgeons face challenges in accurately visualizing bone surfaces during the procedure, hindering their ability to effectively guide surgical interventions. Computer-Assisted Orthopedic Surgery (CAOS) systems, which incorporate imaging modalities like ultrasound (US), offer a promising solution to address these limitations. This research contributes to the field of CAOS by proposing a novel algorithm for femoral head bone surface segmentation in US images. Leveraging local phase features for image denoising and a rigid object filtering process, the algorithm overcomes inherent noise and simplifies segmentation tasks. The proposed method achieves promising results, demonstrated by metrics like accuracy (0.7978) and low Hamming Loss (0.2022). Additionally, metrics like F-score (0.8269) and recall (0.8377) indicate robust bone surface delineation. While the absolute y-centroid difference is 32.125 pixels, the percentual error remains low at 4.4010%.

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

Femoral Head Surface Segmentation Using Ultrasound Images as Input Source: A Local Phase and Rigid Filtering Approach

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Eduardo de Avila-Armenta,
  • José María Celaya-Padilla,
  • Robert B. A. Adamson,
  • Gamaliel Moreno-Chávez,
  • Antonio Martinez-Torteya,
  • Manuel A. Soto-Murillo,
  • Jorge I. Galván-Tejada,
  • Carlos E. Galván-Tejada,
  • Erika Acosta-Cruz,
  • Osvaldo Moreno-Terrazas,
  • Miguel A. Cid-Baez,
  • Diana L. Jácome-Cadena

摘要

Hip arthroscopy, a minimally invasive procedure for diagnosing and treating hip disorders, is gaining traction due to technological advancements. However, surgeons face challenges in accurately visualizing bone surfaces during the procedure, hindering their ability to effectively guide surgical interventions. Computer-Assisted Orthopedic Surgery (CAOS) systems, which incorporate imaging modalities like ultrasound (US), offer a promising solution to address these limitations. This research contributes to the field of CAOS by proposing a novel algorithm for femoral head bone surface segmentation in US images. Leveraging local phase features for image denoising and a rigid object filtering process, the algorithm overcomes inherent noise and simplifies segmentation tasks. The proposed method achieves promising results, demonstrated by metrics like accuracy (0.7978) and low Hamming Loss (0.2022). Additionally, metrics like F-score (0.8269) and recall (0.8377) indicate robust bone surface delineation. While the absolute y-centroid difference is 32.125 pixels, the percentual error remains low at 4.4010%.