|
3D Reconstruction and Tracking of
Anatomical Structures from Endoscopic Images |
THESIS COMMITTEE
Branislav Jaramaz, Srinivasa Narasimhan, Yanxi Liu, Ko Nishino ( |
ABSTRACT
Endoscopes are widely used in minimally
invasive surgery. As a key tool, endoscopy is attracting increasing attention
for its potential role in computer assisted surgery. It is important also
necessary to understand the 3D environment of the operating field during
surgeries. However, 3D reconstruction from endoscopic images is very
challenging due to the small field of view, big distortion, featureless
tissues and messy blood etc. Previous works either used shape from shading or
shape from motion to recover the 3D shape of tissues. In this thesis a novel
methodology is developed to fulfill 3D bone reconstruction based on images, camera
motions and statistical models from populations. An efficient and robust
approach is also proposed to track soft tissues and fulfill the 3D view of
interested features. For
hard tissues, a solution is presented to reconstruct the Lambertian
surface of bones using a sequence of overlapped endoscopic images, with
partial boundaries in each image. We extended the shape-from-shading problem
to deal with perspective projection and two near point light sources that are
not co-located with the camera center. The framework can be easily extended
to multiple near point light sources. And then multi-image framework is
proposed to recover complete occluding boundaries of the bone, by aligning
partial shapes obtained from different images in the world coordinates
tracked by the endoscope. Finally a complete and consistent shape is obtained
by simultaneously re-growing the surface normals
and depths in all views. To fulfill the algorithm, a complete calibration
scheme is carefully designed to estimate both geometric and photometric
parameters including the rotation angle, light intensity and light sources
spatial distribution. In order to deal with the over-smoothness and
occlusions, we employed a statistical atlas to constraint and refine the mutli-view shape from shading. A two-level framework is
also developed for efficient atlas construction.
For
soft tissue, with richer texture features in soft tissues, providing 3D view
of soft tissues and Telestration markers is
possible. We developed a real-time 3D Telestration
system to locate and track the correspondences between a pair of stereo
images. We implemented a hierarchical feature descriptor to represent the Telestration markers which are selected by surgeons at
the very beginning. Our feature descriptor is robust to the environmental
noise. A global geometric constraint is discovered to deal with occlusions
caused by smoke and tool interactions. Finally, the correspondences are used
to provide 3D view of Telestration by fusing two
images in the da Vinci’s surgical system. MAIN CONTENTS
|
|
|
|
Last Update: September 2009 |
