In recent years, the ability of computers to extract information from images has exponentially increased. One of the responsible for this advance is the Convolutional Neural Networks (CNN). CNNs have consistently outperformed classic Machine Learning techniques since 2012, when AlexNet won the ImageNet Large Scale Visual Recognition Competition. Most of the design work of a classical Machine Learning algorithm is to choose the appropriate characteristics. In contrast, a Deep Neural network as the CNN takes raw information and automatically learns the characteristics through training. We aim to apply these state-of-the-art techniques to the field of medical image to help diagnosis. For this, at the Instituto de Fisica de Cantabria, we have at our disposal the PAFIP (Assistance Program of the Initial Phases of Psychosis) cohort imaging data. This cohort is a well-characterized sample of patients with a first episode of non-affective psychosis scanned in the first 3 months of entering the program, a follow-up (1, 3, 5 and 10 years) and a comparable sample of healthy volunteers. This dataset contains not only anatomical and diffusion imaging data but also biochemistry, sociodemographic, clinical, genetics and cognition information. The project final goal is to develop a virtual assistant for the health experts (mostly radiologists) helping to improve significantly the diagnosis time and thus the quality of the medical service.

This job requires computing and software skills, since the candidate will be involved to software development using state-of-the-art techniques. The candidate will work with a large group of physicist and software engineers (from senior to students) that will provide the necessary training. All resources needed are available in the group and the institute.