Navigation systems based on registration of endoscopic and CT-derived virtual images for bronchofiberoscopic...

Bronchofiberoscopy is an essential diagnostic procedure in patients with lung cancer. Sampling methods employed during endoscopy of the respiratory tract are performed with the aim of diagnosis confirmation and staging. Transbronchial needle aspiration may be used for evaluation of lymph nodes neighbouring with trachea and bronchi. Many efforts have been undertaken to increase the sensitivity of this procedure including the application of endobronchial ultrasonography. In recent years several research groups have proposed models of navigating systems to provide computer assistance during bronchoscopic interventions. Although they have used different techniques, their objective was the same - enabling tracking location and movement of bronchofiberoscope tip with reference to previously-acquired computed tomography (CT) images. Since a fiber-optic bronchoscope is a rather long and flexible device, determination of its tip location is not an easy task. The adoption of optical tracking methods used in neurosurgery or laparoscopic surgery to endoscopy of the tracheobronchial tree is usually not possible. Another obstacle is related to the fact that bronchofiberoscopes usually have only one operational channel. This feature considerably limits the feasibility of navigation systems based on the use of small electromagnetic sensing devices or USG probes. The sources of positioning errors in such systems are respiratory movements and the lack of external referential coordinate system associated with the tracheobronchial tree.A promising option for development of a bronchoscopic guidance system is the application of image registration algorithms. Such an approach encompasses registration of endoscopic images to views derived from advanced imaging methods, e.g. CT. In the first step, reconstruction of a three-dimensional, endoluminal views is performed. Next, the position of the virtual camera in a CT-derived virtual model is determined using a complex multi-level image registering algorithm. This stage is the source of the greatest differences between proposed methods and a potential field for considerable improvements.This paper presents a review of recent as well as classical image registering methods used in navigation systems for bronchofiberoscopy. Details related to the selection of crucial elements of navigation systems based on image registering techniques, such as transformation, similarity criteria, matching and global optimization, are specified.