Segmentation of granulation in diabetic foot ulcer by using improved fuzzy c-means method / Syarafana Seri Buana, Nurul Nadiah Rosli and Anis Dayana Rusli
Fuzzy C-Means (FCM) is a method that has been broadly used in image segmentation of medical imaging. However, up until now it is hard to find a suitable method that can be applied in segmenting the granulation in a diabetic foot ulcer (DFU) images. This is because the complexity of a true colour ima...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Student Project |
| Language: | en |
| Published: |
2018
|
| Subjects: | |
| Online Access: | https://ir.uitm.edu.my/id/eprint/49317/1/49317.pdf https://ir.uitm.edu.my/id/eprint/49317/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Fuzzy C-Means (FCM) is a method that has been broadly used in image segmentation of medical imaging. However, up until now it is hard to find a suitable method that can be applied in segmenting the granulation in a diabetic foot ulcer (DFU) images. This is because the complexity of a true colour images which included noise, blur and intensity non-uniformity always cause difficulties in segmentation process. FCM is also related to the repetitive computation of the distance between pixels within clustering centers and local spatial neighbours that makes the images having a complexity in computation. In order to overcome this issue, an Improved Fuzzy C-Means (IFCM) method based on Morphological Reconstruction (MR) and Median Filtering (MR) are used in this study. MR utilize the local spatial information of images in order to enhance the segmentation results. Meanwhile, the employment of MF is to modify the membership partition of the distance computation between pixels within local spatial neighbours in order to improve an image detail preservation and the noise-immunity of the images. The IFCM was tested on 30 DFU images that were taken by using smartphone where the segmentation is focused on the red part of the wound which called as granulation tissue. The performance of the segmentation results is measured in terms of accuracy and sensitivity based on percentage relative error obtained by comparing the area of image segmentation results using IFCM and expert findings. The results show that the percentage average relative error is 6% which implies the accuracy and sensitivity of the method of 94% and 93% respectively. From the results achieved, it is proven that the method used is applicable in segmenting the granulation of DFU images efficiently and effectively. In future study, the use IFCM can be extended to other types of medical imaging, predominantly on open cut wound. |
|---|