Personalized protective ventilation of intensive care patients using a digital twin
This is a VINNOVA supported project together with the MedTech company Getinge AB.
A ventilator is the machine in the hospital intensive care unit which helps you to breathe when you are not able to breathe by yourself. It is thereby a life-saving medical device. The aim of this project is to construct and evaluate a computational system which can use data from the patient, and from this data make adjustments to the operation of the ventilator, in essence make a digital "twin". We use machine learning to adapt parameters of the equations controlling in the ventilator. The machine learning methods are time-series regression methods including symbolic regression.
This project was started as a part of dBrain, a consortium supported by DigitalFutures, KTH.
The aim is to provide health care professionals with tools supporting diagnostic and prognostic assessments. We use machine learning methods to extract features from eye-tracking and MEG data obtained from Parkinson’s patients and healthy volunteers. Parkinson’s disease is characterized by problems pertaining to movements, however cognitive and sleep perturbations often occurr too. Eye-tracking has become a very successful tool when assessting reading deficits (dyslexia) and is currently explored with relation to a number of neurological and psychiatrical disorders. Parkinson’s patients may show several characteristical differences in eye movements compared to healty volunteers. Furthermore, one aspect of Parkinson’s patients motor dysfunction is tremor, shaking of the limbs. This has an origin in the brain where particular brain-waves are altered. MEG is a method to measure brain oscillations to find alterations in location or oscillation frequency and hence analysis of this data also provides important information.
This project is conducted in collaboration with Seth Grant, Center for Clinical Brain Sciences, Edinburgh University and it is supoprted by the Swedish VR grant agency.
The recipient side of excitatory neuronal communication, synaptic spines, are characterized by a postsynaptic density consisting of an assembly of proteins. These take part in transmission of the signal evoked by receptor activation and subsequent generation of biochemical and electrical activity. Changes to the efficacy of this transmission is generally assumed to be of central importance to learning and memory. We study by using computational modeling the amplitude (roughly the efficacy of this transmission) and in particular duration of changes to this efficacy. We are interested in the relationship between molecular events in the postsynaptic density to changes and their duration. We are also studying more principal functions of the molecular assembly as an information-processing device.
This project was conducted in collaboration with Martin Schmelz, Translational pain research, Heidelberg University.
Acute pain is relatively well treated by today's pain killers or compounds like lidocaine. However, for treatment of chronic pain, there is a large unmet need for new drugs. Research is in part unsuccessful due to the lack of an understanding of what changes in ion channels underlie the changes in excitability of peripheral nerves. One of the key problems is that the intracellular membrane potential of these nerves is not experimentally accessible. The present project uses computational neuroscience to construct a model of a peripheral nerve, a C-fiber. We are thereby able to provide a causal link between ion channel function and function of the axon, as well as between changes in ion channels and pathological changes in disease.
This project was conducted in collaboration with AstraZeneca R&D Södertälje.
In this project, we develop a computational search method to design ion channels so that they can achieve optimal physiological/terapeutical effect on cell or network function. The project currently evaluates direct search strategies to find optimal characteristics. In each cycle of the procedure, new channel parameters are set, next biophysical simulations include the channel in the cells of the network/system at study. From the simulations, resulting physiological function is measured/evaluated. Based on this evaluation, new parameters are computed using the search method.
Synchronous activity is an integral part of brain function. At the single neuron level, there may under normal conditions be mechanisms that maximizes processing while proving sufficient safety margins to undesirable hypersynchronous states such as epilepsy. In this project using quantitative modeling we are studying the possibility of controlling a neurons bias to respond to synchronous synaptic input by adding a novel potassium current. Experimentally, pharmacological manipulation of endogenous ion channel types, or genetic knock-in of new channels, might provide possible ways of implementing our results.
In learning and memory it has recently become clear that in
addition to synaptic plasticity there are cellular changes in
excitability due to changes in ion channels. The project focuses on
cationic (TRP) currents which are known from in vitro studies to
produce long-lasting depolarizing plateau potentials. Further, these
currents are activated by group I metabotropic glutamate receptors as
well as muscarinic type 1 receptors, and blocking of these receptors
have been shown to produce behavioral deficits in long-term and
working memory experiments. In the project, we combine
pharmacological, electrophysiological and modeling techniques.