Clustering and classification of local image of wound blotting for assessment of pressure ulcer

This paper describes applying image recognition techniques to the stained image captured by wound blotting. The wound blotting adsorbs the proteins on the wound surface and visualizes protein distribution as a stained image. The local patterns of the stained image may indicate wound healing. For investigation of relationship between pressure ulcer healing process and protein distribution, the categorization and classification by image recognition technique are required because manual classification and annotation are time-consuming and troublesome. In order to apply clustering and classification to the stained image, three features (GLCM, wavelet, and LBP) were compared. As for the clustering, three features achieved the similar performance, however, the clustering results were slightly different from human labeling. As for the classification, wavelet and LBP features achieved good performance. However, particular texture pattern, which is defined as texture whose intensity was stable or changed on direction, was difficult to classify. These results demonstrated the feasibility of applying image recognition technique to the stained images for wound assessment.