Automatic Construction of Dynamic and Setting-Specific Prediction Rules in Radiology (11862)

Project nummer: 11862

Omschrijving van het onderzoek

In this project, we will investigate the automatic construction and adaptation of clinical prediction rules in radiology, in order to determine the performance of such an approach and to assess the feasibility of its future extension into a broader domain. Since this is rather unexplored territory, we will explore this research area via the development of a prototype. The research will be conducted in four stages. First, we will collect setting-specific clinical data, which will be acquired via text mining from various data collections within the hospital information system (ZIS), the patients' electronic medical records (Elpado), and the picture archiving and communication system (PACS). The second stage comprises the development of a prototype module for the automatic construction of prediction rules, based on a logistic regression model. Third, we will develop a prototype module to adapt prediction rules automatically via Bayesian updating. Finally, the fourth step concerns the evaluation of the modules, the resulting prediction rules, and the full prototype in various respects.
The research will result in a fully functional prototype, capable of constructing and adapting prediction rules automatically. Furthermore, the research will produce insights into the feasibility of the future extension of such a system into a broader domain.

Gebruikers

Three companies are involved in this project.

Projectleider

Prof.dr. M.G.M. Hunink
Erasmus MC
Radiodiagnostiek
Postbus 1738
3000 DR ROTTERDAM

Status van het project

Starting date:	This project had not started, yet.
End date:

Print | Over deze site |

Sitemap | Voorbehoud | Gewijzigd 10-1-2012