The EurValve Retrospective Study will acquire data from as large a patient group as possible, across EurValve’s three clinical centres, to facilitate the development of the mechanism to infer missing data, and to provide evidence for the generation of the rule sets that will drive the detailed decision support process. This study will therefore gather data to inform a ‘machine learning’ process.
Machine learning and Rule Sets
An important component of EurValve is a machine learning module to infer data that is not available but is required for execution of computational, mechanistic, physiological models to provide information relevant for decision support. In EurValve this process is associated with work to access and represent publication and population data, and activities to develop rule sets from accumulated data.
Multiple machine learning paradigms are being assessed in order to find the best possibilities for inferring data required for execution of models. This task involves, more specifically, selection and preparation of data for learning, development and selection of promising features for learning, development of machine learning algorithms, and derivation of indications of performance of these algorithms by using for example cross validation techniques.
To assist with the machine learning activities outlined above, data will be acquired from as large a patient group as possible, to facilitate the development of the mechanism to infer missing data, and to provide evidence for the generation of the rule sets that will drive the detailed decision support process.
The primary objective of this study is to populate the machine learning system maximally with anonymised data.