Development of Acute Stroke Lesion Segmentation Algorithm in Brain MRI using Pseudo-colour with K-means Clustering
Manual segmentation infarct core of acute ischemic stroke from medical imaging currently faces a few challenges and causes a high intra- and inter-observer difference. Besides the present standard is tedious and time taking task performed by the radiologists, and the outcome is depending on their ex...
Saved in:
| 主要作者: | |
|---|---|
| 格式: | Thesis |
| 語言: | English |
| 出版: |
2021
|
| 主題: | |
| 在線閱讀: | http://ir.unimas.my/id/eprint/36627/3/Abang%20Mohd%20Arif.pdf |
| 標簽: |
添加標簽
沒有標簽, 成為第一個標記此記錄!
|
| 總結: | Manual segmentation infarct core of acute ischemic stroke from medical imaging currently faces a few challenges and causes a high intra- and inter-observer difference. Besides the present standard is tedious and time taking task performed by the radiologists, and the outcome is depending on their experience. Research has shown that an automated segmentation from the Magnetic Resonance Image (MRI) is potentially giving more effective and accurate results. This study aims to develop an automatic segmentation by utilizing clustering algorithm for acute ischemic stroke lesion identification. Developing an automatic segmentation, the question remains: To what extent does an automatic segmentation give accurate results in segmenting the acute ischemic stroke region from the medical images, particularly in MRI image? Based on the thorough review of the literature on the automated segmentation of acute ischemic stroke, it can conclude that automatic segmentation consists of image acquisition, pre-processing, segmentation, and validation of the segmented image from the MRI. The result in this work shows the potential of the automated segmentation can distinguish between the healthy and affected brain tissue by high as 90.08% in accuracy and 0.89 in the dice coefficient. The development of an automatic segmentation algorithm was successfully achieved by entirely depending on the computer without human interaction. Further research is needed to identify other factors that could increase the effectiveness of automated segmentation. |
|---|