DATA CREATION Towards the Prevention of Psychosis Schizophrenia Studies in Edinburgh

1
DATA CREATIONTowards the Prevention of
Psychosis SchizophreniaStudies in Edinburgh
the Edinburgh High Risk StudyNeuroGrid
PsyGridNeuroPsyGrid Imaging OntologyDatabasing
Images and associated dataPsychosis
SupercomputingPlans for further studies

• Stephen Lawrie
• Edinburgh

2
Schizophrenia

• Characterised by delusions, hallucinations,
  altered affect and a lack of drive
• Surprisingly common 1 lifetime risk
• Highly genetic risk in 10 in those with an
  affected first degree relative
• Age at onset usually 25 years
• Usually chronic, relapsing, incapacitating
• Treatments ameliorative at best
• Associated with abnormalities of brain structure
  and function

3
Imaging and clinical studies of psychosis in
Edinburgh

• Treatment resistant / responsive (40 sMRI)
• EHRS phase I 1995-2000 (300 sMRI)
• Sex Chromosome Aneuploidy (60 sMRI)
• Schizophrenia and/or Learning disability (100
  sMRI)
• Psychosis spectrum (100 sMRI)
• EHRS phase II 2000-5 (200 sMRI and 200 fMRI)
• DTI and MRS in Schizophrenia (100 sMRI etc)
• Edinburgh Study of Co-morbidity (200 sMRI)
• Bipolar I, II, III, IV (400 sMRI, 100 fMRI)
• Bipolar high risk study (200 sMRI and 200 fMRI
  and ongoing)
• Amygdala I, II, III (100 sMRI, 100 fMRI and
  ongoing)

4
Edinburgh High Risk Study

• A prospective study of 200 subjects at high risk
  (HR) of Schizophrenia for genetic reasons i.e.
  initially healthy subjects aged 16-25 who had two
  or more close relatives with schizophrenia.
  Compared to first-episode cases and healthy
  controls on...

• Baseline measures
• genetic liability
• dermatoglyphics
• obstetric complications
• minor physical anomalies / neurological soft
  signs
• CBCL
• SIS
• RISC
• Also took blood for genes at the end of the study

• Repeated measures
• substance use
• life events
• neuropsychology
• structural MRI
• functional MRI
• PSE
• PANSS

5
Neuro-psychology NART IQ, WAIS-R VRs, RBMT
story especially AVLT 1-5 total score
EHRS Baseline predictors
Neuro-anatomy AMYG-HIPP vol Gyrification Index
R PFC
6
EHRS changes towards psychosis
Cannabis use and major life events are associated
with psychotic symptoms and (weakly) with
psychosis 2-4 yrs later. 0-2 yrs pre-diagnosis,
anxiety/depression fall, typical psychotic
symptoms supervene and GM density falls. But, no
apparent changes in neuropsychological test
scores over this time.
Mean scores on the six PSE principal components
on three occasions of 8 HR subjects who fell ill
(relative to NP chronics)
GM density Reduces In Right Uncus, Fusiform
Cerebellum 2.5 yrs on average before Dx
7
       
 
A collaboration between clinical, imaging and
e-scientists to create a Grid-based network of
neuroimaging centres and a neuroimaging tool-kit,
focused on three clinical exemplars dementia,
stroke and psychosis. Sharing data, experience
and expertise will facilitate the archiving,
curation, retrieval and analysis of imaging data
from multiple sites enable large clinical
studies.
 
 
 
 
 
 
 
 
8
The scanner harmonisation problem

• Each scanner presents a unique bias
• Harmonising pre-processing approach
• Developed metrics to measure effects
• Trade off between within scanner accuracy and
  reduction in the effects of the bias
• Ultimately, seeking to show enhanced rather than
  reduced power from combining scanner data

9
       
 
 
 
 
 
 
 
 
 
10
NeuroPsyGrid (NPG)Stephen Lawrie, Carole
Goble, Jano van Hemert, Rob Proctor, Alan Rector
and John Suckling- A Barnes, D Job, J Kola
11
       
 
NeuroPsyGrid towards an ontology and
multi-centre brain imaging in early psychosis

• MRC Collaboration grant application with the
  central aims of
• harmonising and accelerating PsyGrid and
  NeuroGrid towards a common database, a shared
  metadata model and an ontology of terms for
  psychosis relevant clinical and biological data,
  with a particular focus on changes around the
  time of onset
• to extend the PsyGrid sMRI 3-site reliability
  study to include two centres from NeuroGrid, to
  repeat the scans one year later and include a
  functional MRI (fMRI) reliability study, to
  inform future multi-centre imaging studies
• c) facilitate multi-centre clinical studies and
  trials with neuro-imaging e.g. multi-centre
  clinical, genetic and brain imaging studies of
  people at high risk of psychosis.

 
 
 
 
 
 
 
 
12
Harmonisation - Histogram matching
Grey matter White matter
13
Towards a Psychosis Ontology

• Dealing with different assessment scales and
  structured interviews - PSE, PANSS, SCID, OPCRIT,
  CATEGO - and two classification systems of
  disorders (ICD-10, DSM - IV)
• Aiming to map different ways of eliciting the
  same information so as to be able to use as much
  information as possible in clinical prediction
• 1st Draft of Psychosis Ontology
• Protégé, OWL, intermediate and binding ontologies
• Imports the Basic Formal Ontology (BFO)
• Imports an extract of the BirnLex annotations
• PSE to SCID symptoms and signs
• To form logical hierarchies where possible
• Mapping and synonyms as annotations on concepts
• Modular to allow extensions
• Anatomy and Psychotic Disorders

14
Databasing Brain Images and associated data

• Working with Angus Whyte Peter Buneman at
  Digitial Curation Centre
• Completed a SCARP project
• - To consider our unusually large and rich
  datasets (2000 sMRI and 800 fMRI scans and
  associated data)
• - To apply DRAMBORA - a risk assessment approach
  for digital repositories - along with the OAIS
  functional model for archival information
  systems, to consider recommended activities for a
  data archive
• - To prepare for future imaging studies where
  disease risk or diagnosis could be assessed with
  reference to a large image resource

15
SCARP recommendations(see http//www.dcc.ac.uk/sc
arp/)

• Recommendation 1 The neuroimaging community
  requires investment in tools to support wider
  trading and collaborative re-use of
  neuroimaging data.
• Recommendation 2 Digital Curation Centre should
  be supported to help develop and provide
  guidelines for data access policies, using
  neuroimaging as an exemplar.
• Recommendation 3 DCC should further investigate
  information management requirements in
  (neuro)imaging, to provide better advice on tools
  and methods.
• Recommendation 4 The neuroimaging community
  requires further support to assess the viability
  and utility of combining existing MRI data sets.
• Recommendation 5 DCC should map factors that
  affect the value of reusing imaging datasets, to
  support better advice on appraising and valuing
  datasets.
• Recommendation 6 DCC should support curation in
  neuroimaging and related fields by providing
  lifecycle management training targeted at
  students and briefing materials for supervisors.
• Recommendation 7 DCC should adapt the DRAMBORA
  risk assessment tool to be used by data
  custodians at the department or research team
  level.

16
Parallel computing

• Parallel computing experience
• White matter tractography
• (TBSS) on 150 subjects in total
• time estimates for the analysis
• serial time of 469 days
• 8 cores 40 days
• Eddie 28 hours
• Parallel computing possibilities
• TBM and super sampling for anatomical change
• TBSS like approaches for other modalities e.g.
• - shape analyses,
• - automated parcellation,
• - multi-dimensional pattern classification
• fMRI connectivity analyses

An FA skeleton image overlaid on an average T1
template showing reduced FA in subjects at high
risk of schizophrenia, compared to controls, in
the anterior limb of the internal capsule. (S.
Muñoz Maniega 2007)
17
Putting it all together examples
The psychosis ontology (NPG), retrospective
harmonisation of existing MRI datasets (NG) and
parallel computing of 1000s of pooled MRI scans
could identify differential disease, symptom and
risk effects from existing (expensively acquired)
data for future hypothesis testing at little
additional cost. NPG prospective sMRI and fMRI
harmonisation routines will inform power
calculations for a planned multi-centre high risk
study in several UK centres, where data could be
shared using the NG toolkit and the most
powerful clinical prediction will likely come
from combining genetic, imaging and clinical data
in an individual against a database of
individuals in risk strata for a variety of
psychiatric disorders.
18
Biobanks and Image-banks

• Biospecimens collected by clinical and academic
  centers are primitively categorized, with
  insufficient linkage to clinical data.
• Realising the potential of stored images and
  genomic technologies requires standards that
  allow interoperability of imaging with other
  data.
• We need secure, ethical, automated systems to
  store and retrieve biospecimens.
• Academic institutions and health systems must
  assign higher priority to biobanking.
• Most images are not stored or indexed in ways
  that facilitate research.
• Nevertheless, medical (brain) imaging has the
  advantage of
• - some standardisation of image acquisition
  procedures
• - DICOM (Digital Imaging and Communications in
  Medicine) standards, which provide uniform
  nomenclature and processes for handling images.
• - some standardisation of image analysis
  routines
• As image acquisition becomes better co-ordinated
  and image analysis technologies develop, the
  harmonisation issue will recede standardised,
  secure, ethical, data storage and access to
  sufficient computing power will become the top
  priorities