Decision Support System for the Diagnosis of Chronic Wounds Using Artificial Intelligence Algorithms on Images

Abstract

A solution is proposed that consists of supporting the professional in deciding how to act on the wound by offering a diagnosis proposal. Artificial Intelligence (AI) algorithms have been developed to allow the extraction of the most relevant wound characteristics through an image and providing similar successful wounds from the health center itself. Five pre-trained Convolutional Neural Networks (CNN) have been used to compare the results with images processed in different ways. In this way, the professional would have a diagnostic reference of other wounds similar to the one being evaluated and thus be able to make the right decision. A total of 711 images were processed and analyzed in order to obtain their most identifying morphological and textural characteristics. From each of the images, the five most similar images in terms of characteristics were searched for and clinically validated by comparing them using an objective assessment scale. The results showed an overall accuracy of 71.12%, calculated as the weighting of the scale match of similar images to the original. With this solution, clinicians improve their confidence in clinical practice by having support in decision making, observing favorable outcomes and progression of chronic wounds.