Lung anatomy segmentation using hybrid global and local methods for high resolution computed tomography images
Also available in printed version: RC78.7.D53 T43 2015 raf
| 第一著者: | |
|---|---|
| その他の著者: | |
| フォーマット: | Master's thesis |
| 言語: | 英語 |
| 出版事項: |
Universiti Teknologi Malaysia
2025
|
| 主題: | |
| オンライン･アクセス: | https://utmik.utm.my/handle/123456789/55644 |
| Abstract | Abstract here |
| _version_ | 1854975054707687424 |
|---|---|
| author | Than, Joel Chia Ming |
| author2 | Norliza Mohd. Noor, supervisor |
| author_facet | Norliza Mohd. Noor, supervisor Than, Joel Chia Ming |
| author_sort | Than, Joel Chia Ming |
| description | Also available in printed version: RC78.7.D53 T43 2015 raf |
| format | Master's thesis |
| id | utm-123456789-55644 |
| institution | Universiti Teknologi Malaysia |
| language | English |
| publishDate | 2025 |
| publisher | Universiti Teknologi Malaysia |
| record_format | dspace |
| record_pdf | Abstract |
| spelling | utm-123456789-556442025-08-20T20:22:42Z Lung anatomy segmentation using hybrid global and local methods for high resolution computed tomography images Than, Joel Chia Ming Norliza Mohd. Noor, supervisor Diagnostic imaging Diagnostic imaging -- Digital techniques Also available in printed version: RC78.7.D53 T43 2015 raf Segmentation of structures in medical images is challenging due to anatomical differences, abnormalities caused by the presence of disease, image noise, and differences in acquisition parameters. Segmentation is one of the initial components of Computer Aided Diagnosis (CAD) system. Radiologists deal with heavy workload of having to examine a large number of high resolution computed tomography (HRCT) images. CAD-based systems can lighten their load and also aid them in diagnostic evaluations. The development of CAD and hence automatic segmentation is aggressively pursued by researchers worldwide. Automatic segmentation of the lungs may encounter three difficulties which are (i) connected lungs where the right and left lungs are close in proximity, (ii) heavily diseased lungs with too much damaged tissue where the lung tissue contrast is different than healthy lung tissue, and (iii) where the right lung is too close to the airway tree. Two segmentation techniques are proposed in this study, which are Automatic Lung Segmentation System (ALSS) with Graph Cut and ALSS with Texture method. ALSS uses Otsu thresholding, empirical thresholding, morphological operations and Radon transform. The gold standard used for evaluating lung segmentation is the delineation of the lungs done by a human expert. The smaller the deviation of the segmentation compared to the human expert, the higher the performance or the quality of the segmentation. The performance evaluation for the segmentation is divided into quantitative and qualitative methods. For the quantitative method, five segmentation measures were used, namely, Dice Similarity Coefficient (DSC), Jaccard Index, Relative Volume Different (RVD), Volume Overlap Error (VOE) and polyline distance Metric (PDM). For the qualitative method, Bland-Altman and scatter plots were used to compare manual tracing from three observers in gauging the performance of the proposed segmentation system. The segmentation performance for ALSS with Texture has a DSC score of 98.4% whereas ALSS with Graph Cut has a slightly lower DSC score of 98.2% for both lungs. When three different observers are introduced to test the accuracy of the ALSS with Texture method, the variation of the segmentation performance for all three is small and high accuracy is achieved. Thus, ALSS with Texture performed better than ALSS with Graph Cut. sof UTM 120 p. Thesis (Sarjana (Falsafah)) - Universiti Teknologi Malaysia, 2015 2025-03-17T03:59:54Z 2025-03-17T03:59:54Z 2015 Master's thesis https://utmik.utm.my/handle/123456789/55644 valet-20170206-092145 vital:95554 ENG Closed Access UTM Complete Unpublished application/pdf Universiti Teknologi Malaysia |
| spellingShingle | Diagnostic imaging Diagnostic imaging -- Digital techniques Than, Joel Chia Ming Lung anatomy segmentation using hybrid global and local methods for high resolution computed tomography images |
| thesis_level | Master |
| title | Lung anatomy segmentation using hybrid global and local methods for high resolution computed tomography images |
| title_full | Lung anatomy segmentation using hybrid global and local methods for high resolution computed tomography images |
| title_fullStr | Lung anatomy segmentation using hybrid global and local methods for high resolution computed tomography images |
| title_full_unstemmed | Lung anatomy segmentation using hybrid global and local methods for high resolution computed tomography images |
| title_short | Lung anatomy segmentation using hybrid global and local methods for high resolution computed tomography images |
| title_sort | lung anatomy segmentation using hybrid global and local methods for high resolution computed tomography images |
| topic | Diagnostic imaging Diagnostic imaging -- Digital techniques |
| url | https://utmik.utm.my/handle/123456789/55644 |
| work_keys_str_mv | AT thanjoelchiaming lunganatomysegmentationusinghybridglobalandlocalmethodsforhighresolutioncomputedtomographyimages |