Towards Intelligent Endoscopes for Automated Cancer Diagnostics

The objective of this project is to translate and integrate current robotic vision systems and micro-electro-mechanical systems technologies that can expand the capabilities of conventional endoscopes. Endoscopes are essentially handheld, fiber optic microscopes which are used by medical doctors to examine a patient’s internal organs in a minimally invasive manner. Through a previous collaboration with Stanford Medical School which provides us with video clips from real, clinical endoscopic examinations, we will translate leading-edge inertial sensor functionality and apply digital signal processing techniques in order to develop practical enhancements for healthcare examinations. A typical “wish-list” for a MD is the ability to use an endoscope to (1) quickly examine vital internal organs for cancerous growths with the widest possible field-of-view, (2) automatically record the precise location of suspected growths, and (3) automatically analyze suspected sites for known cancer characteristics. We will use software-based methods such as simultaneous location and mapping (SLAM) and pattern matching algorithms , along with inertial sensor augmentation, in order to reconstruct precise 2-D and 3-D representations from endoscopic video of an internal organ such as the human bladder.