Resumen de Una aproximación basada en interfaces de usuario naturales y aprendizaje profundo para monitorizar pacientes con la enfermedad de Alzheimer

Santos Bringas Tejero

• español

  Esta tesis explora la utilización de sistemas software basados en interfaces de usuario naturales para monitorizar a pacientes diagnosticados con la enfermedad de Alzheimer. Concretamente la utilización de interfaces de usuario naturales ha permitido recopilar datos relativos a la movilidad de estos pacientes y a las apraxias, síntomas que pueden llegar a manifestar. Finalmente, la tesis propone un conjunto de técnicas del ámbito de la Inteligencia Artificial para analizar los datos recopilados por estas interfaces con el objetivo de generar modelos cuyas capacidades de predicción pueden ser de utilidad a los profesionales sanitarios en su labor para controlar la evolución de la enfermedad en los pacientes diagnosticados.

• English

  Nowadays, the applications of Artificial Intelligence are innumerable and increasingly enter more and more different fields, being found in practically all areas. Originally, they were successfully applied to data analysis for the economy, improving industrial processes or medicine; although we can now find them in applications available to anyone, such as embedded in smart devices like smartphones or televisions; and they are even used to generate texts and images that resemble those created by a human. Leaving aside the ethical implications that this entails (which despite being a very important part, will not be discussed in this thesis), it is worth delving into some of the problems solved by Artificial Intelligence in order to study them and try to take their success to other environments.

  One field of great interest to society at large is medicine, as it affects the whole population considerably. A small contribution can help improve the lives of a large number of people. As already mentioned, the application of Artificial Intelligence models in medicine is one of the first to appear, and has been used to address very different problems: from ``simple" analyses of numerical data obtained from statistical studies of the population to complex models that process images obtained from diagnostic tests, such as scans or X-rays.

  This thesis attempts to address some of these problems in medicine by taking advantage of modern software technologies such as AI, aiming to make a small contribution to this field in order to try to support the work of medical teams and, with it, the lives of patients. One condition that is interesting to study due to the increasing ageing of the population is dementia, especially Alzheimer's disease as it is the main cause of dementia today. Currently, there is no cure for this disease, but there are treatments that can attenuate the symptoms, which is why detecting the disease early and assessing its degree of affection can be of great help.

  This thesis hypothesis that it is feasible to use modern software systems, based on natural user interfaces, to monitor and collect data from patients with Alzheimer's disease. This data can be used to design and train Artificial Intelligence models that can be used to create diagnostic tools. To validate this hypothesis, the main objective has been to analyse patients with this disease in order to obtain different data. With this, various Artificial Intelligence models are proposed, each one directed by the type of data and the problem to be addressed, in order to help to evaluate and diagnose the disease. Within the framework of each of the analyses performed, various tests have been carried out to check the effectiveness and efficiency of these models based on different validation metrics. Finally, the user-oriented Artificial Intelligence paradigm is briefly explored, with a view to future steps in the research, in order to try to bring these technologies closer to end users and improve their experience with them.