Desarrollo de aplicación móvil y web para la detección de enfermedades faciales mediante el uso de inteligencia artificial

Abstract

Facial diseases represent a significant diagnostic challenge in the medical field, and their early detection is crucial for effective treatment. This project addresses the need for a reliable and accessible detection of these diseases, so it was defined as a general objective to develop a mobile and web application using artificial intelligence for the recognition of facial diseases. A quantitative approach was adopted, which is particularly appropriate for testing the hypothesis and handling the large volume of data that will be used to train the model. Through a systematic review, we first sought to understand deep learning technologies and convolutional neural networks. Then, from the set of training images, main features of facial diseases were extracted. The research had a quasi-experimental design, appropriate to subsequently evaluate the functionality and diagnostic efficacy of the developed application. For the development of the model, the CRISP-DM methodology was applied, which offers a solid and well-defined methodological structure. For the development of the user interfaces, the EXTREME PROGRAMMING (XP) methodology was adopted, which consists of four well-defined phases in which the requirements and user stories for the web and mobile application were defined. Under the same methodology, the application was evaluated by means of acceptance tests that demonstrated the correct operation and compliance with the established requirements. On the other hand, the neural network model was evaluated through the use of a confusion matrix with an evaluation data that was outside the training data resulting in an 88% accuracy rate in detection; then it was concluded that the hypothesis where it was stated that the minimum accuracy rate would be 70% has been successfully corroborated by exceeding the defined threshold