Contact:
R.G. Belleman, PhD
Science Park 904
1098 XH Amsterdam
The Netherlands
Tel: +31 20 525 7272
Fax: +31 20 525 7419

Platform for Development and Evaluation of Strategies for Medical Image Registration

Summary

The goal is to design and implement a software platform to facilitate the development and evaluation of strategies for medical image registration by integrating (public) software, interactive visualization tools and assistance for choosing methods.

Description

Medical images enable the observation of inner structures of the (human) body in a non-invasive manner. A large variety of image acquisition equipments are currently available, such as X-ray, ultrassound, computed tomography (CT), magnetic ressonance imaging (MRI). Each type of equipment (or "scanner") reveals a particular aspect of the human body, providing complementary information that can be used to support diagnosis, clinical research, and treatment planning, guidance and follow-up. Images often have to be combined, for example, to fuse information obtained with different equipments, to compare the images of the same person acquired before and after a surgery, to compare the image of a person with some "atlas", or to report results for a population.

The combination of medical images starts with their geometrical alignment via an operation called "image registration". In the simplest case, the acquired images can be aligned with an affine transformation using "rigid registration" methods (e.g., to register bones in CT images).
In more complex situations, the deformation from one image into another requires a non-linear transformation with a "nonrigid registration" method (e.g., to align MRI scans of the brain of different persons). Image registration is an active area of research that strives for the development of new and improved registration methods.
The solutions are largely problem-dependent, that is, the adequate (sequence of) registration methods typically have to be (developed and) optimized for each new application.

A large amount of software is publicly available (e.g., FSL, AIR, ITK) and can be used for the implementation of custom image registration strategies. However, even when existing software is adopted, image registration cannot be taken for granted mainly because the development and evaluation of custom strategies requires much technical knowledge and manual effort. Choosing the adequate method and parameter settings is difficult, and depends on the type of application. Moreover, the combination of methods available at different software packages typically requires adaptation, which is typically done by hand. Finally, visualization tools to inspect registration are usually limiting.

The development and evaluation of registration strategies could be facilitated by using a platform with the following features:

it integrates a large repertoire of (existing) registration methods into a common environment, in which they can be easily combined to each other and tested interactively
it provides interactive and intuitive visualization tools to inspect the results of registration and evaluate their quality
it provides aids to help the user choose the adequate methods and parameter settings depending on the problem, reducing the need for technical knowledge to design a registration strategy.

The goal of this project is to develop a platform to facilitate the development and evaluation of strategies for medical image registration with the above mentioned properties. The student is expected to

perform a survey on free software for medical image registration, visualization and component-based application development
propose the platform requirements
define platform architecture
implement a prototype in which the basic functionality can be demonstrated (integration, visualization, assistance)
evaluate the platform by implementing at least one case.

At all levels the focus will be on reusing exiting software and brain imaging applications.

This work will be developed under the scope of the VL-e Project (http://www.vl-e.nl), SP 1.3 (Medical Diagnosis and Imaging).

Subjects involved:

Component-based software development
Medical image processing (basics)
Visualization of medical images
Programming

Additional material

VTK (Visualization Tool Kit):
http://www.vtk.org/
FSL (Functional Magnetic Resonance Imaging of the Brain Software Library):
http://www.fmrib.ox.ac.uk/fsl/
AIR (Automated Image Registration):
http://bishopw.loni.ucla.edu/AIR5
ITK (Insight Segmentation and Registration Toolkit):
http://www.itk.org/
J. B. A. Maintz and M. A. Viergever. A survey of medical image registration. Medical Image Analysis, 2(1):1-36, 1998. http://www.cs.uu.nl/people/twan/personal/media97.pdf

Student profile

We are looking for a student who is interested in multidisciplinary projects, can program, has good communicative skills and lots of initiative.

Supervisors

Dr. Silvia D. Olabarriaga, silvia@science.uva.nl
Dr. Robert G. Belleman, robbel@science.uva.nl
Kruislaan 403, Instituut voor Informatica, UvA.