EUVIMEDEUVIMED entdecken
Crossref

HoleNet: Neural Network for 5-DoF Skull-Hole-Center Localization for Minimally Invasive Brain–Machine Interfaces Surgery

Angen Ye, Xiaofeng Wang, Yujia Song, Yuqing Li, Shan Yu, Dapeng Zhang

IEEE Transactions on Medical Robotics and Bionics · 2026-08 · 8 (3) · 1086-1098

DOI: 10.1109/tmrb.2026.3699212 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Transactions on Medical Robotics and Bionics, 2026-08. Verfasst von Angen Ye und 5 weiteren Autor:innen. Die Arbeit nennt 54 Literaturangaben. Gefördert von: National Key Research and Development Program of China, National Natural Science Foundation of China, Beijing Natural Science Foundation Haidian Original Innovation Joint Fund Project.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Minimally invasive surgery shows promise in Brain-machine interfaces (BMIs), as it facilitates sampling specific neural signals with a low risk of infection. During the minimally invasive operation, 5-Degree of Freedom (DoF) skull-hole-center localization (3-DoF point, 2-DoF orientation) is desired to insert micro-electrodes. Conventional circular-hole localization methods are based on fitting ellipses, which is numerically unstable in challenging BMIs scenarios. Although learning-based ellipse detection methods alleviate this problem, they require sophisticated circular contour annotations. Unlike the aforementioned ellipse-based approaches, we introduce HoleNet which leverages 2D keypoints representation, requiring minimal annotation effort. Based on stereo 2D keypoints, the geometric and photometric constraints are imposed to match stereo keypoints for triangulation. Besides, we introduce the Stereo Hough Voting module to enhance 3D consistency. Finally, to determine the 2-DoF orientation, a skull-surface plane is regressed based on the proposed coplanarity assumption. We provide theoretical analysis to verify the coplanarity assumption and conduct extensive experiments to validate the proposed method. The results show that the mean distance error of the 3-DoF point is lower than $15{\mathrm {\mu \text {m} }}$ , and the mean relative angle error of the 2-DoF orientation is within 3°, which satisfies the requirement for collision-free in surgery process.

Förderung

  • National Key Research and Development Program of China
  • National Natural Science Foundation of China
  • Beijing Natural Science Foundation Haidian Original Innovation Joint Fund Project

Zitieren

APA

Ye, A., Wang, X., Song, Y., Li, Y., Yu, S., & Zhang, D. (2026). HoleNet: Neural Network for 5-DoF Skull-Hole-Center Localization for Minimally Invasive Brain–Machine Interfaces Surgery. IEEE Transactions on Medical Robotics and Bionics, 8(3), 1086-1098. https://doi.org/10.1109/tmrb.2026.3699212

Vancouver

Ye A, Wang X, Song Y, Li Y, Yu S, Zhang D. HoleNet: Neural Network for 5-DoF Skull-Hole-Center Localization for Minimally Invasive Brain–Machine Interfaces Surgery. IEEE Transactions on Medical Robotics and Bionics. 2026;8(3):1086-1098. doi:10.1109/tmrb.2026.3699212

BibTeX
@article{Ye2026,
  author = {Ye, Angen and Wang, Xiaofeng and Song, Yujia and Li, Yuqing and Yu, Shan and Zhang, Dapeng},
  title = {HoleNet: Neural Network for 5-DoF Skull-Hole-Center Localization for Minimally Invasive Brain–Machine Interfaces Surgery},
  journal = {IEEE Transactions on Medical Robotics and Bionics},
  year = {2026},
  volume = {8},
  number = {3},
  pages = {1086--1098},
  doi = {10.1109/tmrb.2026.3699212}
}
Kostenlos starten →