DIADEM: Data Information and Analysis for clinical DEcision Making

Lead Research Organisation: Queen Mary, University of London
Department Name: Computer Science

Abstract

Our vision of healthcare in the digital economy is that it will be based on intelligent modelling and real-time analysis of medical data, assessable directly in the doctor's surgery, or even online, to drive individual diagnosis and treatment decisions. Much research towards this vision will inevitably address the data generation problem using technologies such as medical imaging, patient information systems, bio-sensors and genomics. However, these sources of data will not be useful without dramatic improvements in data analysis techniques. Current techniques used in the health sector, which are typically based on classical statistics, cannot adequately contribute to causal reasoning and learning, pattern recognition and feature extraction. Nor can they adequately address the issue of integrating data to arrive at intelligent decision-making (a limitation of previous rule-based AI techniques).The first step is to handle the integration of new data sources to improve the process of data analysis at all levels, ranging from major statistical studies (including international controlled trials) through to the very widespread smaller scale clinical trials and audits driving the development of healthcare. However, ultimately the vision requires the extension of these techniques to individual clinical consultation and decision-makingKey to progress towards our vision is bringing together diverse communities in medicine, statistics and machine learning. Before new techniques are developed, existing ones need to be understood, applied and refined in the light of experience. Thus, the focus of the cluster's work is on the use of ICT in Healthcare, and in particular the analysis of health data:1. Exploiting techniques for data analysis and understanding, using technique not currently applied in the medical discipline.2. Analysis of novel data. The analysis techniques relevant to the cluster will include Bayesian nets and causal models; fusion, feature extraction, and interpretation of the data provided by each analytical instrument; multivariate data analysis for clinical decision support.This interdisciplinary cluster involves over a dozen world-renowned Centres of excellence (medical, computer science/risk assessment, maths and statistics), end-users and ICT providers. The team is made up of the most experienced members of these Centres who will help ensure the widest participation and dissemination of results.

Publications

10 25 50
publication icon
Fenton N (2010) Comparing risks of alternative medical diagnosis using Bayesian arguments in Journal of Biomedical Informatics