The Health-e-Child Project & PlatformData Integration - Semantic and Syntactic

Interoperability

David Manset – MAAT-G

March 5th, 2009

EGEE-UF/OGF25 Catania, Sicily

2 Health-e-Child

• Establish Horizontal and Vertical integration of data, information and knowledge for Paediatrics

• Develop a grid-based biomedical information platform, supported by sophisticated and robust search, optimisation, and matching

techniques for heterogeneous information,

• Build enabling tools and services that improve the quality of care and reduce its cost by increasing efficiency

• Integrated disease models exploiting all available information levels

• Database-guided decision support systems

• Large-scale, cross-modality information fusion and data mining for knowledge discovery

• A Knowledge Repository for Paediatrics

Cardiology•Tetralogy of Fallot (ToF)•Cardiomyopathy (HCM, DCM)

Rheumatology•Juvenile Idiopathic Arthritis (JIA)

NeuroOncology•Brain Tumors – Gliomas

Data IntegrationEnabling Tools

Knowledge DiscoveryDecision Support

Data Mining …

3 Health-e-Child

The Grid

World is moving from supercomputers to grid computing that for a fraction of the cost are able to deliver the same services…

Several regular computers+ Powerful+ Cheap+ DeCentralized+ UnLimited Scalability

One big computer+ Powerful- Expensive- Centralized- Limited ScalabilityS

uper

Com

putin

gG

rid C

ompu

ting

HealthgridHealthgrid +

Storage Capacity• Storing Millions of Medical

Images and Clinical Records• Storing newly generated

Knowledge • Caching intermediary image

processing data

Computing Power• Data Mining, e.g. Image

Processing• Knowledge Discovery,

similarity searches over large populations of patients

Connectivity, Security• Sharing Data, Knowledge and Applications

cross-institution, cross-country• Securing Infrastructure and ensuring

patient privacy• Gluing heterogeneous Information

Systems• Abstracting from legacy technologies

4 Health-e-Child

Three peadiatric hospitals Gaslini, Genoa, Italy

GOSH, London, UK

Necker, Paris, France

OPBG, Rome, Italy

Strong interdisciplinary team across Countries and languages Technical and clinical fields

Research on three peadiatric disease areas:

Arthritis Cardiac Disorders Brain Tumours

Health-e-Child Europe-wide Information Platform for Pediatrics

5 Health-e-Child

Research Focus in Rheumatology

Wrist Hip

163 patients enrolled (Target – 300)

Improve current classification of JIA subtypes• Identify homogeneous groups of clinical features• Find early predictors of poor outcome• Identify sensitive markers of joint damage

progressionDevelop MRI and US paediatric scoring system

• Joint space width varies with age – studies performed on adult are not applicable on children.

Robust Information Fusion• Pattern discovery in multimodal data, correlation

between genomic, clinical and image data

Rely on the collaboration with PRINTO: Pediatric Rheumatology INternational Trials Organization

6 Health-e-Child

Research Focus in Cardiology • Concentrating on Right Ventricular Overload

and Cardiomyopathies• Computational electromechanical models of

the heart• RVO monitoring and decision support based

on similar cases – similarity search on complex, multimodal data

• Decision Support based on semi-automatic feature extraction from cardiac MR

• Health-e-Child CaseReasoner

• Visualizing integrated biomedical data for patient cohorts using treemaps and neighborhood graphs

257(RVO)+39(CMP) patients enrolled (Target – 300)

Short AxisLong Axis

7 Health-e-Child

Research Focus in Neuro-oncology:Glioma growth model:• Interpolating growth between two time instances • Using proliferation and diffusion of tumor cells • Including high speed of tumor invasion in white vs. grey matter

Knowledge Discovery, Finding Prognostic Markers:• Classification of low vs. high grade• Sub-typing of pilocytic astrocytomas (e.g. regarding tumour site, age)• Regression analysis of factors (clinical, imaging, genetics) that affect treatment

outcome• Prediction of prognosis (survival rate and quality of life)

49 Studies Collected (Target – 77)

8 Health-e-Child

Vertical Data Integration

9 Health-e-Child

De-Identified Electronic Patient Record

• Siemens web based data collection tool• Adjusted for Health-e-Child

10 Health-e-Child 10

Patient Study, Diagnosis, Therapy

Patient Information

Pedigree

Medical History

ICD

Data Import into HeC

11 Health-e-Child

Data Import into HeC• Migration tool imports XML forms

created by Siemens data collection tool

• Tool semi-automatically analyses forms and suggests name and type according to HeC meta data model and UMLS

• Tool instantiates HeC data model and migrates patient data using gateway API

• no need to know underlying data base management system

After once establishing the mapping, patient data can be migrated to the HeC grid fully automatically

12 Health-e-Child

DistributionDistribution

transaction

transaction

transaction

IGG GOSH NECKER

AccessPoint

HeC Gateway

+ + +

ICD

Integrated Case Database(ICD)

-Grid Database of Patient Data- From clinical records to files- Distributed (1 per Hospital)

- Multi-centre (federation)

-Fine-grained Access Controls- Synced with VO • new VO AMGA sync daemon- ACLs until records

Data Overview

-Database Backend Abstraction (AMGA Layer)-Transactional insertion and updates-Replication of portions of the data for ISD and ICD v1

-Multi-level Integrated Data Model (IDM)- From Organs, to Cells, to Genes…- Medical Images along with clinical records

-Multi-centre Case Database (ICD)- ICDs are federated and seen as a single one

-Patient privacy is ensured from the beginning- Anonymisation client-side- UUIDs for all patient folders

-Peer-To-Peer Patient Privacy for storing mappings- Useful for retrieving concerned sets of patients

13 Health-e-Child

Exploiting Integrated DataCaseReasoner Application

Cardiac Example

14 Health-e-Child

Step 1: Anatomical Model from Cardiac MR• Anatomical model of right ventricle

(RV) created from HeC data (based on 30 isotropic volumes from Gosh)

• Semi-automatic initialisation of model based on detection library from Siemens Corporate Research

• Multi-sequence view for model editing

Fast, accurate 4D quantification of RV volumes (ES, ED) from which RV ejection fraction and further measurements can be easily derived

Manual annotations in diastole and sysole

HeC application for semi-automatic annotations

15 Health-e-Child

DistributionDistribution

Similarity DistanceCalculation

IGG GOSH NECKER

AccessPoint

HeC Gateway

+ + +

Process:1. Query for RV Meshes in ICD2. Process Similarity Distance Measurement « where data is »3. Aggregate results in a WEKA dataset4. Display result using Treemaps, NG graph or Heatmapper

16 Health-e-Child

Visualization of Result Set

• 3 specific non-traditional visualisation techniques• Treemaps [Shneiderman, 1992] (integration in progress)• Neighbourhood graphs [Toussaint, 1980]• Combined correlation plots/heatmaps [Verhaak, 2006]

17 Health-e-Child

Step2: Electromechanical Model and Simulation

Volumetric mesh at time 0 Simulated fibres (+60° on the endocardium to

-60° on the epicardium)

Visual adjustment of simulation(Segmentation / Simulation)

Simulated beating heart + fibresColors: strain anisotropy

Simulated beating heart + fibresColors: contraction

18 Health-e-Child

Virtual Volume Reduction Surgery

19 Health-e-Child

Cross-Project InteroperabilityHealth-e-LINK Application

Data Mining Example

20 Health-e-Child

Health-e-Child 3D Knowledge Browser

21 Health-e-Child

Integration of @neurLINK from @neurIST

Thank you for your attention!