University of Twente Student Theses

Login

Segmenting knee bone structures from 3dimensional MRI data

Vogelzang, A.R. (2015) Segmenting knee bone structures from 3dimensional MRI data.

[img] PDF
1MB
Abstract:The knee bone is a very important joint which supports the body when upright. In older people however wear and tear of the joint cause pain, assessing the damage is very difficult. For assessing damage MRI is a valuable tool, however interpreting data has to be done manually and is very time consuming. The Medical Imaging group from the university has developed an algorithm to interpret, segment, the image data from an MRI scan. The current model however is not yet accurate enough, the current assignment is to improve the accuracy of the bone model. The algorithm is built using an Active Shape Model (ASM). The ASM requires datasets for input, these datasets all need to have points at approximately the same location on the surface, landmarks. For this the sets are pre-processed before building the model. Next to the pre-processed datasets it has several variables important for processing these datasets into a mean shape with the most common variations. This is done using principal component analysis. The resulting model is then used to segment new data sets. This part has its own variables needed to correctly apply the model to new data. With this algorithm tests are run. The first test is a generalised cross validation using a homogenous set selected from the available training sets from MICCAI. All variables are kept constant while running this test. The second test is to optimise the variables in order to increase the accuracy. This is done with a less homogenous dataset and the variables are varied one by one to test their effect on the algorithm. The generalised cross validation gives results with an average AvgD of 3,5 times higher than the difference between trained experts. Next to this the results have a standard deviation 2 times higher than the difference between trained experts. The results from optimising the variables shows that improvements can be made but that the algorithm is far from being accurate enough. The algorithm works as it is now, it however has room for improvement. In order to increase the accuracy recommendations are made, including rebuilding large parts of the algorithm.
Item Type:Essay (Bachelor)
Faculty:EEMCS: Electrical Engineering, Mathematics and Computer Science
Programme:Electrical Engineering BSc (56953)
Link to this item:https://purl.utwente.nl/essays/93916
Export this item as:BibTeX
EndNote
HTML Citation
Reference Manager

 

Repository Staff Only: item control page