Master Thesis: Radiation-Free Drill Guidance for the Interlocking of Intramedullary Nails using a Video-Augmented-Fluoroscopic Device. (2012)
The interlocking of intramedullary nails is a technically demanding procedure
which involves an excessive amount of X-ray acquisitions; one study lists as
many as 48 to successfully complete the procedure and fix screws into 5 mm holes
in the nail. In this work, a novel low cost radiation-free drilling guide is
designed to assist surgeons in completing the distal locking procedure without
any X-ray acquisitions. Using an augmented reality fluoroscope that
coregisters optical and X-ray images, we exploit solely the optical images
to detect the drilling guide in order to estimate the tip position in the
real-time in X-ray. To evaluate the accuracy of the tip estimation,
we tested over 200 random drill guide poses showing a mean tip-estimation
error of 1.72 +/- 0.7 mm which is significantly robust for the interlocking.
Also, three expert surgeons participated in a preclinical study on dry bone
phantom and successfully completed the distal locking 56 out of 60 trials with
no X-ray acquisition for guidance and an average time of 2 minutes.
These promising results have shown the robustness and accuracy of our
method thereby offering a low cost and efficient solution in decreasing the
complexity of interlocking of intramedullary nails procedure.
Supervision: Pascal Fallavollita, Lejing Wang, Nassir Navab
A first draft of a CamC simulation based on blender's game engine.
Digitally reconstructed radiographs (DRR) are produced over calls to our C++ DRR generation component over the blender python API.
The user moves the C-Arm over the keyboard to a desire pose and takes radiographs.
Those radiographs are overlaid over the video stream shown on screen.
IDP: Markerless Motion Capture in the Operating Room. (2011)
Goal of this IDP was to extend the patch-based surface tracking algorithm to a skeleton
based solution. Such a solution could be used to track surgeons in an operating room and enhance
human-machine interaction paradigms in this context. We implemented a model-based motion capture
algorithm based on 3D point cloud data. An Expectation-Maximisation algorithm computes the point
cloud/body parts assignment and the pose estimation is based on an inverse kinematic framework.
Supervision: Cédric Cagniart
Technical Report: IDP Project Markerless Motion Capture in the Operating Room.
HIWI: 3D Reconstruction Multi-Camera Studio. (2010)
I supported documentation and follow-up works on the Real-Time 3D Reconstruction Studio of the chair,
as well as the implementation of the skeleton based motion capture component.