Prediction of Evolution in First Episode Psychosis by Means of Multi-Input Deep Learning Algorithms
Objectives: Deep learning (DL) algorithms lay at the core of most successful artificial intelligence applications developed in recent years. In this project we will take advantage of their capacity to easily and smoothly combine data of very different nature and to exploit the predictive power of previously fitted DL models on similar data to make predictions on the evolution of individuals with a first episode of psychosis (FEP). Specifically, we will generate DL algorithms combining data of different types (see next paragraph) sampled at the onset of a first episode with previously fitted DL models on schizophrenia and bipolar disorder patients with the aim of predicting diagnosis, levels of clinical severity and degree of functionality at six months and at one year of the onset.
Methodology: Functional, structural and diffusion brain MRI images, fingerprint images, genetic data on single nucleotide polymorphisms (SNPs) previously related to schizophrenia and bipolar disorder, and scores from clinical, functionality and cognition scales will be obtained from a sample of N = 200 FEP patients. This data will be combined with a large database of historical data of similar nature previously sampled at our institutions in order to generate DL predictive algorithms of FEP evolution.
Instituto de Salud Carlos III, con la cofinanciación de Fondo Europeo de Desarrollo Regional "Una manera de hacer Europa"