Service Communautaire d'Information sur la Recherche et le Développement - CORDIS

Final Activity Report Summary - AIMED (Advanced intelligent medication - personalised drug dosage concepts to improve quality of life)

The project worked on two different subjects, i.e. the classification of strokes, either hemorrhagic or ischemic, by means of magnetic impedance tomography, as well as decision support systems for heart failure management.

About 15 million people worldwide suffer a stroke every year. Five million people die as a consequence of the stroke and another five million become permanently disabled. Magnetic induction tomography (MIT) is one of the promising technologies for a fast classification of strokes and a reliable diagnosis to provide the patients with an adequate therapy. The non invasive and non ionising magnetic induction tomography allows for the reconstruction of conductivity distribution images for the diagnosis of ischemia and hemorrhagic stroke in neurology.

A 16 channel MIT system for the analysis of conductivity distributions in human tissue was developed and its applicability was demonstrated in phantom objects. Improvements of data acquisition and imaging processing were achieved by parallel readout and parallel computing of the high-frequency signals. Advanced digital signal processing algorithms were developed to enable an image post-processing on standard desktop computers. Moreover, a dynamic imaging technique was developed for MIT with a linearised Kalman filter to improve the visualisation of the distributions of brain tissue with different permittivity and conductivity/permeability, aiming to reveal electrical and magnetic characteristics of a brain bleeding. Finally, drift and noise as major system limitations were studied and algorithms to eliminate and reduce noise were developed.

Chronic heart failure (CHF) is the major cause for hospitalisation in adults in western societies, mainly due to decompensation of patients. Active prevention and diagnosis and personalised treatment contribute to the stabilisation of chronic patients and the reduction of events. The approach that was taken in the aimed Marie Curie project in collaboration with the European Union's Sixth Framework Project 'MyHeart' was to collect daily vital body signs on CHF patients in an easy and comfortable way. The data was processed via a decision support system (DSS) and the platform provided instant recommendations to the user. The system also sent the information to the professionals for a better follow up.

The designed DSS was based on Bayesian networks (BN) and combined the accepted standardised clinical guidelines with the most advanced monitoring data in daily routine. It consisted of a loop around the patient, giving support both to the professional and to the patient in a more frequent follow up. In this project prototypes of a static BN for treatment guidance and of a dynamic BN (DBN) for daily management and decompensation prediction in chronic heart failure (HF) patients were developed. A validation with cardiologists for the selection and quantification of the input and output variables was completed. The system did not try to substitute the clinical guidelines or the role of the professionals, but, with the help of new technologies, to foster the implementation and follow up of evidence based therapy outside of the hospital.

BNs were interesting for knowledge representation because they allowed both top-down and bottom-up inference, they facilitated decision under uncertainty, they easily captured the experts reasoning in cause-effect terms, they could learn from data and they were easily updatable and personalisable. The availability of statistical tools and software solutions even for devices with low computational capacity rendered them an interesting approach for home medical DSS. The limited availability of daily patient monitoring data for the training and validation of the BN was the main difficulty for these models. This highlighted the necessity of clinical trials to collect unbiased information about the sensing technologies at home and to validate DSS models for home monitoring. The project contributed to the preparation of such trials which were planned to be carried out in the framework of the MyHeart project.

Reported by

See on map