By matching children with rare or life-threatening diseases and modelling potential disease progression, researchers hope to find new routes forward.
Software tools are being developed that can search and compare patient data at hospitals across Europe to find children with closely matched conditions. The doctors can then study how the matched patients at other hospitals were treated and whether that treatment was successful. The information will greatly improve doctors' ability to choose the right path for their own patient.
The tools being developed within the Health-e-Child project can compare a vast range of structured and unstructured data, including genetic and clinical data, as well as images from CAT and MRI scans and other records.
The Health-e-Child system, protected by high security, links anonymised databases of patient information at hospitals in Paris, Genoa, Rome and London. There are plans to extend the network to 25 hospitals.
Health-e-Child researchers are working on tools for three complex paediatric diseases with at least partly unknown causes: heart diseases resulting from an overload of the right ventricle, juvenile idiopathic arthritis, and brain tumours (gliomas).
The EU-funded project has tackled fundamental data sharing infrastructural problems as well as ethical and data protection questions, data analysis and data mining issues. Both disease-specific and cross-disease tools have been developed.
For unstructured data such as images, the Health-e-Child project has created tools that translate visual information into machine-readable (and therefore machine-comparable) language. The project's 3D registration tool for MRI scans, and its MRI 'erosion scoring' system for juvenile idiomatic arthritis have been recognised as important advances in their fields.
Matching and modelling two in a million
Health-e-Child's CaseReasoner tool enables clinicians to search thousands of disease diagnoses, treatments and outcomes to find a child similar to their own patients. The clinicians set the search parameters themselves. In the case of heart patients, clinicians could include factors they consider important, such as genetic markers, the age of the child, the heart rate - even the amount of exercise the child takes.
The results can be displayed as a 'network' with cohorts of patients with similar diagnoses clustered together and colour-graded accorded to the level of similarity. Clinicians can then dive into the detailed data on any of the patients or clusters to better understand their diagnoses and the success of the procedures the patients have been through.
The CaseReasoner could also be used to search out the procedures that have been most successful, giving the clinician insights into the optimal path forward.