Artificial Intelligence for personalized sepsis patient outcome prediction

• Grant number: 17/25497-6
• Support Opportunities: Scholarships abroad - Research Internship - Scientific Initiation
• Effective date (Start): February 12, 2018
• Effective date (End): May 11, 2018
• Field of knowledge: Engineering - Biomedical Engineering - Bioengineering
• Principal Investigator: Karl Heinz Kienitz
• Grantee: José Lucas de Alencar Saraiva
• Supervisor: Visakan Kadirkamanathan
• Host Institution: Divisão de Engenharia Eletrônica (IEE). Instituto Tecnológico de Aeronáutica (ITA). Ministério da Defesa (Brasil). São José dos Campos , SP, Brazil
• Research place: University of Sheffield, England
• Associated to the scholarship: 17/11272-2 - Sepsis prediction: modelling using neural networks and its robustness, BP.IC

Abstract

The research developed by the student in Brazil considers methods and tools for sepsis modeling. Sepsis is a clinical condition usually known as "generalized infection". The research planned for the internship at the University of Sheffield complements scientific initiation research done at ITA, Brazil, which, in turn continues a work done in 2014/2015 in an undergraduate research project at ITA. Given a neural network designed in the previous work and trained for prognosis of patients affected by sepsis based on a specific dataset, the initial goal of the ongoing research is the validation and the improvement of this neural network by integrating information from a new dataset. The investigation will address thechallenge of achieving robustness with heterogeneous datasets, which have only partialoverlap in the data fields and differ in data quality, towards high quality predictionperformance. The goal of the research to take place at the University of Sheffield, UK, is to investigate the robustness of the networkthrough artificially introduced perturbations/errors in the used datasets in response to the data quality challenge. Additionally, under the guidance of Prof. Visakan Kadirkamanathan, thesupervisor at the host institution, it is intended to evaluate alternative topologies and training methods for the neural network in order to achieve more robustness and efficiency in the prognosis of sepsis. In particular, the focus will be on recent advances in artificial intelligence such as deep learning networks. (AU)