Main Page

Public health expenditure accounts for almost 6% of GDP in Spain, making the health sector one of the most important of our socio-economic system. The unprecedented ability to gather higher volumes of data represents an opportunity to extract useful knowledge for the clinical practice as well as a difficulty due to the challenging nature of data.

Two applications of particular socio-economic interest are:

1. Predicting the evolution of the health status of chronically ill patients. A significant amount of resources must be devoted to the care of chronic patients due to the increase in life expectancy, together with the prevalence of chronic diseases in an ageing population. Predicting the evolution of the health status of these patients would slow down the progression of chronic disease, and allow for an efficient allocation of health resources.

2. Predicting healthcare associated infections in Intensive Care Unit (ICU). About 30% of ICU patients are affected by at least one episode of infection during his hospital stay. Predicting the evolution of bacteria and their multi-resistance would improve prognosis, reduce patient stay and resource consumption

The aim in both cases is to improve not only the health evolution of the patients, but also the efficiency of the allocation of health resources.

From the point of view of data analytics both problems exhibit common challenges: a) data are scattered, heterogeneous, of high dimensionality, from clinical sources of different nature (free text, diagnostic, drugs or laboratory tests, among others), and b) extracting knowledge useful for clinical practices is difficult. These challenges call for the adaptation of the traditional tools in data analysis and statistical inference in order to obtain more effective and interpretable solutions. KLINILYCS proposes designing and adapting methods from data analytics and graph signal processing in order to model, process and identify patterns in data bases of complex nature, together with its particularization to the aforementioned clinical applications. The project will address the development of new feature selection methods and similarity measures adapted to the heterogeneous and irregular data structure; algorithms to find patterns in trajectories using probabilistic networks; path modeling using signals defined in the nodes or edges of a graph; and inference methods applied to data defined on graphs.

To ensure the success of the project, the work team includes multidisciplinary researchers and international collaborations with prestigious groups. Part of the group has a technical profile with wide experience in data analytics and signal processing, while the rest are clinical experts with experience in the health applications addressed in this project. KLINILYCS project development will generate scientific results in a promising research topic as well as technological solutions to current socio-economic problems.

KEYWORDS

Tools

Information Processing, Graph Signal Processing, Probabilistic Models, Topology Inference, Tensor Data, Statistical Inference, Machine Learning, Data Mining

Applications

Health analytics: Chronically ill, ICU, Diabetes

The KLINILICS project has been supported by the Spanish Government: Ministerio de Ecomomía y Competividad - Plan Estatal de Investigación Científica y Técnica y de Innovación 2013-2016 - Programa Estatal de I+D+i Orientada a los Retos de la Sociedad - Grant number: TEC2016-75361-R.