Semi-automatic segmentation using MR images II
The objective of this project is to develop a method to automatically segment the Articular Cartilage Ligament (ACL) of the knee in Magnetic Resonance (MR) images. Even though MR imaging technology provides a non-invasive and accurate diagnosis of ACL injury, the images require interpretation by tra...
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Final Year Project (FYP) |
| Language: | English |
| Published: |
2009
|
| Subjects: | |
| Online Access: | http://hdl.handle.net/10356/16492 |
| _version_ | 1811679095772151808 |
|---|---|
| author | Lung, Wen Zheng. |
| author2 | Poh Chueh Loo |
| author_facet | Poh Chueh Loo Lung, Wen Zheng. |
| author_sort | Lung, Wen Zheng. |
| collection | NTU |
| description | The objective of this project is to develop a method to automatically segment the Articular Cartilage Ligament (ACL) of the knee in Magnetic Resonance (MR) images. Even though MR imaging technology provides a non-invasive and accurate diagnosis of ACL injury, the images require interpretation by trained radiologists. And because the ACL is the most frequently injured structure in the knee, an automated segmentation method is needed to aid radiologists in handling the large volumes of MR images.
The ACL is positioned amongst many other ligamentous structures of the knee; it therefore does not have good contrast with its surrounding tissue in MR images, which makes it difficult for computers to automatically identify it – an algorithm which can achieve unsupervised segmentation of the ACL has not been documented to date.
The algorithm developed in this project is implemented using MATLAB, and uses mathematical morphology to locate the ACL without supervision, allowing the algorithm to take advantage of the ACL’s unique shape and orientation within the image. A series of morphological filters are applied to the image, thereafter an active contour model is employed to delineate the boundary of the ACL. The semi-automatically segmented images were compared against a set of manually segmented images to determine the accuracy of the segmentation. The present algorithm is able to correctly identify the ACL in 92.5% of the images, and a dice coefficient of 80.5% (compared against the manually segmented images) is achieved in MR images where the ACL is physiologically healthy. |
| first_indexed | 2024-10-01T03:03:43Z |
| format | Final Year Project (FYP) |
| id | ntu-10356/16492 |
| institution | Nanyang Technological University |
| language | English |
| last_indexed | 2024-10-01T03:03:43Z |
| publishDate | 2009 |
| record_format | dspace |
| spelling | ntu-10356/164922023-03-03T15:32:56Z Semi-automatic segmentation using MR images II Lung, Wen Zheng. Poh Chueh Loo School of Chemical and Biomedical Engineering DRNTU::Engineering::Bioengineering The objective of this project is to develop a method to automatically segment the Articular Cartilage Ligament (ACL) of the knee in Magnetic Resonance (MR) images. Even though MR imaging technology provides a non-invasive and accurate diagnosis of ACL injury, the images require interpretation by trained radiologists. And because the ACL is the most frequently injured structure in the knee, an automated segmentation method is needed to aid radiologists in handling the large volumes of MR images. The ACL is positioned amongst many other ligamentous structures of the knee; it therefore does not have good contrast with its surrounding tissue in MR images, which makes it difficult for computers to automatically identify it – an algorithm which can achieve unsupervised segmentation of the ACL has not been documented to date. The algorithm developed in this project is implemented using MATLAB, and uses mathematical morphology to locate the ACL without supervision, allowing the algorithm to take advantage of the ACL’s unique shape and orientation within the image. A series of morphological filters are applied to the image, thereafter an active contour model is employed to delineate the boundary of the ACL. The semi-automatically segmented images were compared against a set of manually segmented images to determine the accuracy of the segmentation. The present algorithm is able to correctly identify the ACL in 92.5% of the images, and a dice coefficient of 80.5% (compared against the manually segmented images) is achieved in MR images where the ACL is physiologically healthy. Bachelor of Engineering (Chemical and Biomolecular Engineering) 2009-05-26T08:13:50Z 2009-05-26T08:13:50Z 2009 2009 Final Year Project (FYP) http://hdl.handle.net/10356/16492 en Nanyang Technological University 67 p. application/pdf |
| spellingShingle | DRNTU::Engineering::Bioengineering Lung, Wen Zheng. Semi-automatic segmentation using MR images II |
| title | Semi-automatic segmentation using MR images II |
| title_full | Semi-automatic segmentation using MR images II |
| title_fullStr | Semi-automatic segmentation using MR images II |
| title_full_unstemmed | Semi-automatic segmentation using MR images II |
| title_short | Semi-automatic segmentation using MR images II |
| title_sort | semi automatic segmentation using mr images ii |
| topic | DRNTU::Engineering::Bioengineering |
| url | http://hdl.handle.net/10356/16492 |
| work_keys_str_mv | AT lungwenzheng semiautomaticsegmentationusingmrimagesii |