MEng Projects
Projects offered for Fall 2013
1. Image registration of time separated 3D CT images and change analysis
The objective of this project is to register in 3D two low-dose chest CT images. That is find the non-linear mapping that matches one scan to the other. This is necessary for a number of applications including (a) locating a lung nodule in a second scan given its location in the first and (b) matching an inspiration scan to an expiration scan to identify lung abnormalities.
There was a grand challeng on this topic http://www.grand-challenge.org/index.php/MICCAI_2010_Workshop see the proceedings at http://www.diagnijmegen.nl/~bram/grandchallenge2010/MIACGC2010.pdf . Tools available at http://stnava.github.io/ANTs/ uses ITK http://www.itk.org/. See also: http://www.ncbi.nlm.nih.gov/pubmed/21632295
2. Modeling of vessels in the lungs to better segment pulmonary nodules.
The objective of this project is to identify pulmonary vessels within 3D low-dose chest images of the lungs. The main normal structures in the lungs are airways and vessels. In this project algorithms will be developed and evaluated for the identification and segmentation of these structures.
3. Lung lobe fissure identification in CT images
The objective of this project is to separate the five lobes of the lungs in low-dose CT images of the chest. This is a challenging task for which considerable previous studies have been reported. For this year the primary objective will be to identify the start of these fissures on the pleural surface.
4. Image quality evaluation of CT images of the chest
The first part of this project will be to identify regions of the image are not suitable for analysis. Some subjects have metal implants such as pacemakers that cause serious artifacts in parts of the image making analysis of those parts impossible. Also in some cases contrast material is injected into the blood that changes features of the image. The project will also estimate the point spread funtion (Frequency response) of the image at different locations.
5. Image separation in 3D space on CT images of the chest
There are a number of tasks in which it is necessary to separate two 3D objects that may be "touching" each other (e.g. two lungs... a very real problem or two bones at a joint another very real problem). The solution will involve finding a surface that separates the two objects involving a minimum cost for that cut.
6. Image visualization using javascript
This project is to develop an advanced web based image viewing and annotation tool in a web system using javascript. This project requires a high level of programming skill and some previous experience with javascript is important.
7. 3D boundary marking Multi-dimensional image slicing and reconstruction
To analyze the shape of lesions or body organs, physicians indicate the boundary of a 3D region of interest by marking boundaries on a set of images slices through one dimension of the 3D image. A problem with this approach is poor precision on the “ends” of the image slice sequences due to excessive partial pixel effects. A second issue is the amount of time required to mark all image slices. This project is to develop a program that will slice a 3D ROI in a number of different directions through the center of the 3D ROI. Since slicing is done through the center of the ROI the effects of partial voxels is minimized. Also the tradeoff between number of slices and volumetric precision and repeatability can be explored allowing the minimum number of slices to be marked by the physician to be optimized. The main effort for this program is on the development and evaluation of a multidirectional 3D slicing and reconstruction algorithm