Registration of bronchoscopic image and CT virtual image with coarse-to-fine strategy

Endobronchial navigation system in the transbronchial lung biopsy is an important tool during the operation, including the alignment of bronchoscopic images and preoperative CT image, to help the precise positioning of target lesions. In this study, a new method with coarse-to-fine strategy is proposed to achieve this goal. First of all, with the aid of optical flow method, epipolar geometry analysis between the two consequential images is used to obtain the initial position of the bronchoscopy. Next, in the stage of image similarity comparison, the property of image saturation and brightness is employed to effectively eliminate the reflection caused by the endoscope light source and the bubbles generated by saliva interference. The experimental results also show that the proposed method can provide clinicians the spatial information needed for bronchoscopic follow-up.