Segmentation of Diabetic Wound by Multidimensional Clustering for Quantitative Assessment of Healing Process

Background: Periodically assessment of the healing process of diabetic wound is a very important task for clinicians. The quantitative analysis of wound is useful to monitor and evaluate the healing process with the intervention of drug therapies. Currently, the evaluation process focuses mainly on visual inspection, which is not accurate enough to accomplish the task. Methods: Segmentation and detection of region of the wound in tissues is an emerging field for assessing the healing process. This paper presents the supervised learning methods to detect and classify the type of wound tissues. Fast Fuzzy C-means and K-means clustering algorithms has been implemented for classifying and detecting the boundary of wounds and also the size of the wound has been measured from images. Results: The intensity of the images has been normalized, and wounds were segmented in five images collected from the open source database. The output result shows that the boundary of the wound has been extracted accurately by both the methods. The enhanced Fast Fuzzy C-means provides better visual segmented output compared K-means process in terms of high PSNR and low MSE values. The high PSNR of 55.57 is measured in EnFuzzy C-means compared to K-means algorithm with a value of 54.74 and the effectiveness of the process has also been calculated by integrating both the process. Conclusion: These methods may be useful in accessing and analyzing the wound healing process in qualitative and quantitative manner in diabetic cases.