Bronchoscopic video analysis methods for bronchial lesion detection

Open Access
- Author:
- Daneshpajooh, Vahid
- Graduate Program:
- Electrical Engineering
- Degree:
- Doctor of Philosophy
- Document Type:
- Dissertation
- Date of Defense:
- July 18, 2024
- Committee Members:
- Robert Collins, Major Field Member
William Higgins, Chair & Dissertation Advisor
Vishal Monga, Major Field Member
Madhavan Swaminathan, Program Head/Chair
Sri-Rajasekhar Kothapalli, Outside Unit & Field Member - Keywords:
- lung cancer
bronchoscopy
narrow-band imaging
bronchial lesion
deep learning
detection
tracking
medical imaging
vessel characterization
lesion enhancement
image-guided intervention - Abstract:
- Early detection of cancer is crucial for lung cancer patients, as it determines disease prognosis. Lung cancer typically starts as bronchial lesions evolving along the airway walls. Bronchoscopy is an effective minimally invasive procedure for detecting lesions. Recent research has indicated that narrow-band imaging (NBI) bronchoscopy enables more effective bronchial lesion detection than other modalities. Unfortunately, NBI bronchoscopy tends to be overly tedious for the physician to use routinely for early lung cancer detection, because of a lack of effective tools for facilitating an exam of the complex airway tree. To address this problem, this dissertation proposes an interactive video analysis system for NBI bronchoscopy airway exams that offers: 1) navigational guidance to facilitate efficient bronchoscopic airway exams; 2) real-time lesion detection to enable automatic analysis of the incoming video; and 3) visualization tools to provide a comprehensive assessment of an airway exam. In particular, we develop an automatic two-stage real-time method for bronchial lesion detection in NBI airway exam. Given a patient's NBI video, the first method stage entails a deep-learning-based object detection network coupled with a multi-frame abnormality measure to locate candidate lesions on each video frame. The second method stage then draws upon a Siamese network and a Kalman filter to track candidate lesions over multiple frames to arrive at final lesion decisions. Additionally, we enhance the visual quality of the detected lesions to better reveal underlying vasculature and characterize suspicious vascular features, facilitating a thorough review of bronchial findings. At the system level, we integrated these methods into the system for real-time detection and localization of lesions as the physician performs a systematic guided bronchoscopic navigation through the airways. Specifically, the system generates an airway exam profile for lesions, enabling more efficient documentation of bronchial findings than current clinical practices. The profile links the patient's pre-operative 3D chest CT scan with the airway exam results, providing comprehensive visualizations for the physician to make final diagnostic decisions. We present results demonstrating the system's performance through a series of human studies. We successfully analyzed lung cancer patient NBI airway exam videos, achieving a lesion detection precision of 90%, recall of 93%, and F-1 score of 90%. Our system summarized NBI bronchoscopic procedures using only 11% of video frames on average to denote bronchial lesions. The system successfully completed airway exam profiles for lesions, detailing their precise locations on the 3D airway tree and associated navigational instructions for the physician to reach the lesion during follow-up examinations.