Related Experiment Video
Updated: Apr 14, 2026
07:35
The Microscopic Transcanal Approach in Stapes Surgery Revisited
Published on: February 16, 2022
3.1K
Monocular Surface Reconstructions in Transcanal Endoscopic Ear Surgery: A Phantom Feasibility Study.
Obinna I Nwosu1, Francis X Creighton2
1Department of Otolaryngology-Head and Neck Surgery, Massachusetts Eye and Ear, Boston, MA.
Summary
Computer vision models can create accurate 3D middle ear reconstructions from single endoscopic images. This technology shows promise for submillimeter accurate surgical navigation in transcanal endoscopic ear surgery.
Area of Science:
- Medical Imaging
- Computer Vision
- Otolaryngology
Background:
- Accurate 3D visualization of middle ear anatomy is crucial for surgical navigation.
- Current methods may be limited by invasiveness or complexity.
Purpose of the Study:
- To assess if a computer vision model can generate 3D surface reconstructions from monocular endoscopic images of the middle ear.
- To determine if these reconstructions achieve clinically acceptable accuracy for surgical navigation when registered to ground-truth imaging.
Main Methods:
- Eight 3D-printed temporal bone phantoms were used.
- Monocular endoscopic images of the middle ear were captured and processed using a surface reconstruction model.
- Reconstructions were registered to ground-truth meshes, and accuracy was quantified using surface reconstruction error (SRE) and target registration error (TRE).
Main Results:
- The mean SRE was 0.58 ± 0.10 mm, with 81.6% of points within 1 mm of ground-truth surfaces.
- The mean TRE was 0.96 ± 0.17 mm, indicating submillimeter registration accuracy.
Conclusions:
- High-fidelity 3D surface reconstructions of middle ear anatomy can be generated from simple 2D monocular endoscopic images.
- These reconstructions achieve submillimeter registration accuracy, demonstrating preclinical feasibility for monocular, endoscope-centric navigation in transcanal endoscopic ear surgery.
