Monitoring patients in chronic and acute diseases by hybrid data analysis
Predicting the individual risk of disease progression is of high clinical relevance for individualized therapies particularly if the disease switches from a chronic benign to an acute malignant phenotype. Moreover, in intensive care the emergence of acute critical states may lead to lethal situations and requires early diagnosis to enable rapid therapeutic response in time.
Transition states along developmental biological processes have been interpreted recently in terms of bifurcations of complex dynamical systems in physics, chemistry and engineering. As critical states in complex systems are often characterized by unique behavior of observable systems parameters, early identification of occurrence of critical states in clinical settings using the typical patterns of criticality of complex systems is an emerging area of research in computational biomedicine.
At JRC-COMBINE we combine mechanistic or hybrid models with machine learning to extract the relevant patterns indicating criticality of patients in clinics from heterogeneous Big Data resources. Roughly, the models serve as model-based filters for the raw data resulting in a significant reduction of the dimensionality of the data sets but minimal loss of information resulting in a significant increase in reliability of identified patterns in the data. We apply the model-integrated Big Data analysis concept on characterizing critical states in patients both in chronic as well as acute diseases.