Skip to main content

Colonic Disease Investigation by Robotic Hydro-colonoscopy

Final Report Summary - CODIR (Colonic Disease Investigation by Robotic Hydro-colonoscopy)

Platform Group
Each group has provided novel outcomes for each research of the research strands.
The outcomes from the PG are:
i. NiTi wires as actuators, with the study of the mechanism and control strategies. The research results are not yet published because a patent will be submitted.
ii. A novel robot locomotion concept, using pulsed/ continuous jet propulsion study has been developed. The outcome of this research is not yet published because a patent will be submitted.
iii. Fluid-electromechanical coupled model of a pulsed jet device has been described and a patent will be submitted.
iv. Sensor fusion methods applied to colonoscopy: Uncertainty theory has been applied to estimate the locomotion state of a device from limited direct measurements of the colonic environment.
v. Hydraulically Driven Colonoscope: Colonoscope driven by pressure differential has been developed to autonomously overcome flexures and navigate through the colon using a safe range of pressures.

Imaging Group
The outcomes from the IG are: prototype systems for (i) classification of normal-abnormal frames in colonoscopy videos; (ii) de-noising (spot noise removal) from colonoscopy videos; (iii) depth estimation for de-hazing and navigation; (iv) multi-class image classification with novel feature-based representations.
Currently the group is working on video (as opposed to frame) classification, specialized feature and image descriptor for high-accuracy classification, system interface.

Biological Group
The most significant achievement by the BG, to date, has been the construction of a dedicated experimental set-up, for fundamental bubble cloud cavitation observations. This unique configuration has allowed the identification of the source of a key cavitation cloud acoustic signal, used throughout a range of cavitation-mediated applications, the mechanism for which has eluded cavitation research for some decades. Research underway includes the development of detectors exploiting this knowledge of the ‘sub-harmonic’ signals, and the incorporation of live-bowel specimens in the experimental cavitation chamber for direct cavitation-tissue interaction studies, and downstream bio-effect quantification.
The following work has been presented and or published:
i. New algorithm for classifying normal-abnormal colonoscopy images, published in IEEE ISBI 2013.
ii. New image descriptor for colonoscopy images, published in IEEE ICCV (workshop) 2013.
iii. Novel algorithm to detect and remove suspended particles in water and sensor noise based on joint low-rank and sparse matrix decomposition; published in ICCV2013, journal paper being submitted.
iv. Introduced of a novel method to select pairs for adding inter-cluster features for feature encoding – submitted to MICCAI 2014
v. A novel pooling method to capture intermediate-scale structure features – submitted to MICCAI 2014
vi. A classification system for colonoscopy image classification – soon to be submitted to journal