Robust camera localisation with depth reconstruction for bronchoscopic navigation.

Area of Science:

• Medical Imaging
• Computer-Aided Surgery
• Pulmonology

Background:

• Bronchoscopy is crucial for diagnosing and treating lung diseases.
• Accurate bronchoscope localization to CT models is vital for targeting lesions.
• Current vision-based methods struggle with illumination and surface texture variations.

Purpose of the Study:

• To develop a robust camera localization technique for bronchoscopic navigation.
• To improve the accuracy of registering the bronchoscope's position to CT bronchial models.
• To overcome limitations of existing intensity- and gradient-based registration methods.

Main Methods:

• Utilized depth information derived from endobronchial images using surface shading.
• Estimated camera pose by maximizing similarity between recovered depth and CT model projection.
• Compared normalized cross-correlation and mutual information as similarity measures.

Main Results:

• The depth-based tracking approach demonstrated superior performance over existing methods.
• Achieved smaller pose estimation errors for the bronchoscopic camera.
• Showed increased robustness to illumination artifacts and surface texture variations.

Conclusions:

• A reliable depth-based camera localization technique for bronchoscopic navigation was developed.
• The proposed framework offers significant clinical value for improved navigation.
• Evaluations confirmed the technique's reliability and effectiveness in phantom and in vivo studies.