Navigated bronchoscopy using intraoperative fluoroscopy and preoperative CT

Abstract

Bronchoscopic biopsies for diagnosis of lung cancer are usually done with the help of intraoperative fluoroscopy. But fluoroscopy images lack 3D information and do not provide a clear view of the bronchi or lesions. Our goal is to enhance the physician's view by overlaying the intraoperative fluoroscopy images with both 2D and 3D airway visualizations from preoperatively taken CT scans. The presented system provides automatic airway segmentation and skeletonization as well as automatic 2D/3D alignment of fluoroscopy to CT. The results are used for correctly overlaying the airways and visualizing bronchoscopic paths. The only additional equipment needed is a specifically designed pattern of steel spheres and sticks on acrylic glass, which is fixed on the patient table. It is used for estimating the C-arm pose during image acquisition and allows 2D/3D image alignment in clinically feasible time (<6 s) and accuracy (mTRE<0.33mm on simulated data). No interference with the physician's standard bronchoscopy procedure is introduced and no additional radiation exposure is required. On the contrary, by improving the physician's view and orientation inside the bronchial tree, a faster and more target-oriented guidance to the site of interest is possible

Hierarchical Discriminative Framework for Detecting Tubular Structures in 3D Images

“…Despite that, the results suggest that the algorithm is suitable for clinical setting, e.g. for bronchoscopic navigation [5,15], since most detected centerlines are within the airway boundaries. The tree extraction algorithm is efficient.…”

“…1 for examples). Despite these challenges, the segmented trees are required to be complete and have low number of false branches to make the results accurate for diagnostic and interventional procedures [5,9,15]. In this paper, we propose an efficient algorithm that addresses the challenges above and produces accurate tree detection.…”