In the scope of the ITI Healthtech call for PhD theses, we are looking for a student who will complete his/her Master degree in the coming months and who wish to enroll in a PhD thesis in the field of robotics and computer vision for computer-guided OCT (Optical Coherence Tomography) biopsies in Fall 2022.
PhD Project : Geometrically correct volumetric OCT scans for digestive applications
Abstract: Optical biopsies in the digestive tract using OCT (Optical Coherence Tomography) can be used to detect pathological tissues on site. However, diagnosis is made difficult by geometrical distortions due to imperfect motions of the imaging probe and motions of tissues. The objective of the project it to combine robotic motions, OCT images and white-light endoscopic images analysis to create geometrically correct OCT volumes, in order to facilitate image interpretation by gastroenterologists.
Context: Optical Coherence Tomography is a tool for diagnosing pathologies, which provides images of tissues up to few millimeters under their surface at the scale of a few microns. It is especially used in ophthalmology and for cardiovascular imaging. The application to the digestive tract has been demonstrated for diagnosing Barrett’s esophagus [Gora17] by using rotating OCT catheters or capsules [Gora13], but its use remains challenging. Optical biopsies should ideally entirely cover suspicious areas and images should be presented in an easily interpretable manner to the physician in order to make diagnosis accurate.
Several problems are currently faced: scanning large irregular surfaces of tissues is difficult and acquired images are deformed because of scanning imperfections and tissues movements.
The AVR and IPP teams of ICube have developed new tools, combining biomedical optics, robotics and image processing for improving the use of OCT images, by increasing the possible scanning volumes [CaravacaMora20] and correcting rotational distortions [Liao21, Liao22]. Such corrections can be applied to helical scans used in capsule systems and to low-profile catheters in contact with locally flat areas. However, there are still several limitations for providing corrected 3D volumes for off-line analysis by a gastroenterologist or a pathologist.
Objectives: The objectives of the PhD thesis are to develop advanced tools for scanning large areas and providing geometrically accurate OCT volumes for visualization.
Linear scanning is not possible on uneven surfaces (for example around a colorectal polyp). Moreover, even in the case of a linear scan, the translational imperfections are not corrected (non-uniformity of the movement of the capsule, peristaltic movements of the tissues). This can result in 3D images which are difficult to interpret and impair longitudinal monitoring of disease progression. Moreover, mapping these images to the real tissues for posterior treatments is a real challenge.
We propose to rely on steerable robotic catheters to realize complex movements with the probe. These movements will have to ensure complete scanning of the area of interest. For reconstructing 3D images, we will rely on different information:
- The (approximate) knowledge of the movement made by the catheter
- The OCT images themselves: Translation non-uniformities can create visible artefacts on the reconstructed 3D images (repetition in the event of a stop, backtracking). Non-parallel B-scans will locally intersect and reconstructed 3D images should incorporate these intersections.
- The white light images provided by the camera embedded in the endoscope carrying the OCT probe. This sensor can allow to measure the position of the catheter [Sestini21] and detect the motions of tissues [Penza18].
The work will combine image processing techniques, machine learning and robotics.
To apply send a CV, cover letter, master program and master grades (M1 and first semester of M2) to : [email protected]
PhD starting dates: between September and November 2022