Full metadata record
| DC Field | Value | Language |
|---|---|---|
| dc.contributor | Department of Computing | en_US |
| dc.contributor.advisor | You, Jane (COMP) | en_US |
| dc.creator | Lin, Yi | - |
| dc.identifier.uri | https://theses.lib.polyu.edu.hk/handle/200/11335 | - |
| dc.language | English | en_US |
| dc.publisher | Hong Kong Polytechnic University | en_US |
| dc.rights | All rights reserved | en_US |
| dc.title | Multilayer perceptron based liver vessel segmentation | en_US |
| dcterms.abstract | Liver malignant tumors is one of the diseases with the highest mortality rates at present. The treatment of it is a significant and urgent research topic all over the world. While radiation therapy, especially radiation seeds implantation technology, is considered as one of the core treatment technologies for modern malignant tumors. However, radiation from the implantation seeds will cause damage to the surgeon. Therefore, researchers is going to invent a medical robot or surgical navigation system to help surgeons to implant the radiation seeds. The three-dimensional visualization and quantitative analysis of liver blood vessel is primary function of the designed medical robot. In the experiment, the author concentrated on developing the liver vessel segmentation methods. There are main three developed method including PM equation based multilayer perceptron model (MLP), Hessian matrix based multilayer perceptron model and 3D PM equation and Hessian matrix based multilayer perceptron. The core idea of these methods is extracting spatial and effective features of voxels and training MLP model to classify voxels into background or vessel class so that ultimately realizing volume segmentation. In the experiment, logical operator is made use of to retrieve the liver region of interest from abdomen volume. Resampling will reduce the influence of label quality and improve the final segmentation result. PM based MLP model is more sensitive to label quality and resampling implementation than Hessian matrix based MLP model. Gaussian Mixture Model will find the liver vessel Gaussian distribution which supports the window center and level adjustment. Adjusting window center and level will not only reduce the amount of noise but also enhance the liver vessel to some extent. PM equation is an anisotropic diffusion which can works quickly and rudely on the region but slowly and gently at the edge. The selection of location equation of PM filter is also discussed. 3D PM equation extracts the information of vessels' edge of the liver CT volume as well as enhance the vessel. And the designed PM equation based model works fast but get an over-segmentation result. Hessian matrix based model assign the three eigenvalues of Hessian matrix of each voxel as input features to train the MLP model. The sigma parameter of Hessian matrix in the experiment should be 1. The designed Hessian matrix based model works fast since each training epoch only needs 2 seconds. It only segmented the coarse vessel and ignored the slim blood vessel and there are some holes on the segmented vessel causing the discontinuity problems. Through combining the result of 3D PM equation and three eigenvalues of Hessian matrix to train a more advanced MLP model, over-segmentation and discontinuity problems are relieved. The segmented result of 3D PM equation and Hessian matrix based MLP model is harmonious and decent. | en_US |
| dcterms.extent | viii, 75 pages : color illustrations | en_US |
| dcterms.isPartOf | PolyU Electronic Theses | en_US |
| dcterms.issued | 2020 | en_US |
| dcterms.educationalLevel | M.Sc. | en_US |
| dcterms.educationalLevel | All Master | en_US |
| dcterms.LCSH | Image segmentation | en_US |
| dcterms.LCSH | Radiography, Medical -- Digital techniques | en_US |
| dcterms.LCSH | Liver | en_US |
| dcterms.LCSH | Hong Kong Polytechnic University -- Dissertations | en_US |
| dcterms.accessRights | restricted access | en_US |
Files in This Item:
| File | Description | Size | Format | |
|---|---|---|---|---|
| 5807.pdf | For All Users (off-campus access for PolyU Staff & Students only) | 3.5 MB | Adobe PDF | View/Open |
Copyright Undertaking
As a bona fide Library user, I declare that:
- I will abide by the rules and legal ordinances governing copyright regarding the use of the Database.
- I will use the Database for the purpose of my research or private study only and not for circulation or further reproduction or any other purpose.
- I agree to indemnify and hold the University harmless from and against any loss, damage, cost, liability or expenses arising from copyright infringement or unauthorized usage.
By downloading any item(s) listed above, you acknowledge that you have read and understood the copyright undertaking as stated above, and agree to be bound by all of its terms.
Please use this identifier to cite or link to this item:
https://theses.lib.polyu.edu.hk/handle/200/11335