Clinical need-driven, deep-learning-enabled medical image analysis of the lung
Speaker: Dr Wei-Ning Lee
Associate Professor, Department of Electrical and Electronic Engineering, HKU
Facilitator: Professor S.C. Wong
Associate Dean (Development and External Relations), Faculty of Engineering, HKU
Date: September 8, 2021 (Wednesday)
Time: 18:00 – 19:00 (HK Time) | Via Zoom
Medium of the talk: English
Registration: https://bit.ly/3xS4RXT
In this talk, Dr Lee will share her team’s ongoing research works that aim to address two clinical needs in interventions for lung diagnosis—planning of navigation bronchoscopy by computed tomography (CT) and guiding of pleural fluid sampling by ultrasound imaging. CT is the imaging modality of choice for examining the lung in greater detail. A three-dimensional (3D) airway map reconstructed from CT images is the prerequisite for navigation bronchoscopy procedure, which is used for safer sampling of pulmonary nodules. However, existing commercial software and methods for 3D airway reconstruction remain suboptimal because they often cannot detect higher generations (i.e., bronchioles) of the airways. Ultrasound imaging of the lung (or, thoracic ultrasound) has become more widely used. Its diagnostic role in the healthcare pathway for lung diseases, particularly in the critical care sector, has recently been revisited and gained unprecedented attention primarily because of COVID-19. Thoracic ultrasound is clinically used to confirm the presence of effusion and to locate the best site for effusion fluid sampling (i.e., thoracentesis). However, distinguishing pleura from inherent image artifacts that look like pleural layers is critical for safe drawing of fluid and requires extensive clinical experience. The shared image analysis approach to meeting these two clinical needs is segmentation. She will give a brief overview of the two medical imaging modalities for the lung and present our developed deep learning models dedicated for segmentation and tracking of the pleura from ultrasound images and of airways from CT images.
All are welcome!