Sepsis prediction: modelling using neural networks and its robustness

• Grant number: 17/11272-2
• Support type: Scholarships in Brazil - Scientific Initiation
• Effective date (Start): August 01, 2017
• Effective date (End): November 02, 2018
• Field of knowledge: Engineering - Biomedical Engineering - Bioengineering
• Principal researcher: Karl Heinz Kienitz
• Grantee: José Lucas de Alencar Saraiva
• Home Institution: Divisão de Engenharia Eletrônica (IEE). Instituto Tecnológico de Aeronáutica (ITA). Ministério da Defesa (Brasil). São José dos Campos , SP, Brazil
• Associated scholarship(s): 17/25497-6 - Artificial Intelligence for personalized sepsis patient outcome prediction, BE.EP.IC

Abstract

This proposal deals with modeling techniques for sepsis, a clinical condition that arises when the body's response to infection causes injury to its own tissues and organs. Since millions of patients are annually affected by sepsis, advances in sepsis-related research have a high potential of impact. It is expected that the modeling of the influence of the relevant variables of the survival cases will point to evidence of successful measures, and will allow the construction of decision support systems for the dynamization of treatment and training of the professionals involved. This work will continue a research already carried out in 2014/2015 by a student of ITA's scientific initiation. The student will be guided by Prof. Karl Heinz Kienitz and will have the weekly or biweekly monitoring of Dr. Otávio Monteiro Becker Jr., of the São Paulo State Transplant Hospital. Given a neural network produced in the previous work, trained to prognosticate sepsis outcome, the initial objective of this work will be to validate and / or improve this neural network. Next, we intend to investigate the robustness of the model used in the training of the network in relation to the quality of the data used. For this, the technique developed in the design of the first network will be used in the design of a second neural network from a new database, whose records are more reliable than those of the database used in the previous work. The networks will be compared with a comparison approach that will also be developed during the research. (AU)