Integrated video analysis system for autofluorescence bronchoscopy

Early detection of lung cancer allows for more effective treatment and helps increase the likelihood of patient survival. This fact has inspired the search for biomarkers that can help indicate disease development and cancer risk. One important minimally invasive method for identifying potential biomarkers entails performing an airway exam using bronchoscopy. More specifically, autofluorescence bronchoscopy (AFB) is notable for its high sensitivity in detecting candidate early cancer lesions along the airway wall. The task of performing an AFB airway exam to identify such lesions, however, proves to be very tedious, error prone and overly dependent on physician skill. This is due to the lack of sufficient tools to facilitate efficient, accurate analysis of the airway exam’s video stream. We propose an integrated interactive system for autofluorescence bronchoscopy. The system takes a patient’s three-dimensional (3D) chest computed tomography (CT) scan and a live bronchoscopy video stream as inputs and provides the following capabilities: 1) guidance assistance for performing the airway exam; 2) automatic video analysis to produce real-time detection and segmentation of candidate lesions; 3) subsequent lesion tracking over the video sequence to identify key frames that denote the most representative locations of detected lesions; 4) visualization and interaction tools to view lesion detection outcomes and to make final lesion confirmation decisions; 5) graphical tools for showing a detected lesion’s precise anatomical location within the 3D airway tree. Through these capabilities, the system has the capacity to deliver a comprehensive assessment of suspect lesions throughout an AFB airway exam. Utilizing the exam videos and CT scans from lung cancer patients, we demonstrate the potential of our system for real-time, systematic analysis of a patient’s major airways.