Testing an Image Mining Approach to Obtain Pressure Ulcers Stage and Texture

摘要

Improvement of pressure ulcers (PU) images analysis through computerized techniques is advantageous both to medical assistance institutions and to patients' life quality. The scientific challenge is to improve assistance to patients with PU by means of reliable image analysis procedures. Diagnosis of stage and predominant texture in a PU is essentially an image colour classification problem that can use existing knowledge. This study performs a classification of pressure ulcers images through an algorithm based on ID3 to construct a decision tree that has RGB statistics as input features and PU stage and texture as target features. A decision tree is constructed first by classification of 18 images of a training set. Then this tree is tested in a set of 45 PU images. Acceptable classification accuracy for training sets was not confirmed in test set.