Scale-invariant registration of monocular endoscopic images to CT-scans for sinus surgery

We present a scale-invariant registration method for 3D structures reconstructed from a monocular endoscopic camera to pre-operative CT-scans. The presented approach is based on a previously presented method for reconstruction of a scaled 3D model of the environment from unknown camera motion. We use this scaleless reconstruction as input to a PCA-based algorithm that recovers the scale and pose parameters of the camera in the coordinate frame of the CT scan. The result is used in an IGP registration method to refine the registration estimates. The presented approach is used for localization during sinus surgeries. It simplifies the navigation of the instrument by localizing it relative to the CT scan that was used for pre-operative procedure planning. The details of our approach and the experimental results with a phantom of a human skull are presented in this paper.