Kidney abnormality detection and classification using ultrasound vector graphic image analysis
Ultrasound imaging has been widely used in kidney diagnosis, especially to estimate kidney size, shape and position, and to provide information about kidney function, and to help in diagnosis of structural abnormalities like cysts, stone, and infection. However, the use of ultrasound in kidney diagn...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English |
| Published: |
2013
|
| Subjects: | |
| Online Access: | http://eprints.utm.my/id/eprint/43970/5/WanMahaniHafizahPFPSK2013.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Ultrasound imaging has been widely used in kidney diagnosis, especially to estimate kidney size, shape and position, and to provide information about kidney function, and to help in diagnosis of structural abnormalities like cysts, stone, and infection. However, the use of ultrasound in kidney diagnosis is operator dependent where the images may be interpreted differently depending on operators’ skills and experiences, variations in human perceptions of the images, and differences in features used in diagnosis. Current kidney diagnosis may be improved by implementing automated techniques and computer aided diagnosis systems, but have not been widely explored. Therefore, this study proposed a vector graphic image formation method which enables the ultrasound images to be manipulated for various applications including region of interest (ROI) generation, cysts detection and segmentation and abnormality classification. Automatic kidney ROI generation algorithm able to achieve 89.6% true ROI when tested with 125 kidney images. Besides that, the vector graphic formation helps in detection and segmentation of cysts automatically with high accuracy (true positive area ratio = 0.9584, similarity index = 0.9439, Hausdorff distance = 11.4018) and less execution time (11.4 seconds). Performance evaluation to 50 single cyst images, and 25 multiple cysts images gave accuracy of 92%, and 86.89% respectively. This vector graphic formation also helps in extracting better features that successfully classify kidney ultrasound images into three different groups namely normal, infectious and cystic with testing and validation accuracy of 93.33% and 91.67% respectively (p<0.05). Overall, this study has shown promising results and implementation of these proposed algorithms into current kidney diagnosis technique may help in improving current diagnosis accuracy while reducing human intervention and operator dependency. |
|---|