ixi(co): progressing a scientific grid project to an end-to-end solution
TRANSCRIPT
Overview
A brief history of “IXI”From the idea to a business plan Challenges in the first two years – GCP– Maintaining an audit trail and FDA 21 CFR part11– Quality Management System / ISO 9001:2000– Software validation
Conclusion
Why using Grid in Automated Medical Image Analysis?
Huge volume of image data–hospitals typically produce several tera bytes per year–results of automated image analysis can be few times larger
Computationally demanding algorithms- sophisticated image analysis of one single image can take several hours (or even days)
Multiple processing steps (Workflows)- one solution consists often out of many single applications
Collaboration of different insitutions- is often required in clinical trials or research studies
Image database
Grid Demonstrator:A Dynamic Brain Atlas (2001-2002)
registration create atlas
AtlasPatient scan
transformation
Computational expensive !
DB query
Grid-enabled image registration and segmentation for medical research:
– to assess scalability of Dynamic Brain Atlas – to extend to data from outside the brain– to build an application interface
3 year project funded by e-science core programme& GSK, Philips Medical Systems, Dunhill Charitable Trust
IXI components
IXI Portal GT3 Index Service
IXIWorkflow service
IXI insertion Service
IXI Core Service
ComputeResources
Service Provider #1
IXI Core Service
ComputeResources
Service Provider #2
MedicalImage
Database
Available Services Aggregations of MICL SDEs
Workflow submission
Invocations &Notifications
Result management
GridFTP
IXI components
IXI Portal GT3 Index Service
IXIWorkflow service
IXI insertion Service
IXI Core Service
ComputeResources
Service Provider #1
IXI Core Service
ComputeResources
Service Provider #2
MedicalImage
Database
Available Services Aggregations of MICL SDEs
Workflow submission
Invocations &Notifications
Result management
GridFTP
Algorithms wrapped as grid services for medical image analysis, e.g.•image segmentation•image registration
•non-rigid registration:•run-time on single CPU between
10 minutes and 24 hours
IXI components
IXI Portal GT3 Index Service
IXIWorkflow service
IXI insertion Service
IXI Core Service
ComputeResources
Service Provider #1
IXI Core Service
ComputeResources
Service Provider #2
MedicalImage
Database
Available Services Aggregations of MICL SDEs
Workflow submission
Invocations &Notifications
Result management
GridFTP
<IXIWorkflow>
<stage id=”0” serviceRef=”http://fmrib.ox.ac.uk/IXI/BETFactory”><global> <paramValue portID=”4”>0.6</paramValue> <paramValue portID=”5”>true</paramValue></global>
<tasks><task id=”0”> <paramValue portID=”0”> gsiftp://ixi.doc.ic.ac.uk/query/brain001.hdr </paramValue>
<paramValue portID=”1”> segmentedTarget.hdr </paramValue></task>
<task id=”1”> <paramValue portID=”0”> gsiftp://ixi.doc.ic.ac.uk/query/brain002.hdr </paramValue>
<paramValue portID=”1”> segmentedSrc.hdr </paramValue>
</task></tasks></stage>
<stage id=”1” serviceRef=”http://ixi.doc.ic.ac.uk/IXI/RigidRegFactory”><global> <!-- rreg parameters --> <paramValue portID=”5”>NMI</paramValue> <paramValue portID=”6”>100</paramValue></global>
<tasks><task id=”0”> <paramValue portID=”0”>${0.0.1}</paramValue> <paramValue portID=”1”>${0.1.1}</paramValue></task></tasks></stage></IXIWorkflow>
• Workflow engine• Workflow language
to combine image analysis services
•XML workflow file
IXI components
IXI Portal GT3 Index Service
IXIWorkflow service
IXI insertion Service
IXI Core Service
ComputeResources
Service Provider #1
IXI Core Service
ComputeResources
Service Provider #2
MedicalImage
Database
Available Services Aggregations of MICL SDEs
Workflow submission
Invocations &Notifications
Result management
GridFTP
• Web interface to launch workflows
• Grid applications to design workflows and to control the execution
IXI components
IXI Portal GT3 Index Service
IXIWorkflow service
IXI insertion Service
IXI Core Service
ComputeResources
Service Provider #1
IXI Core Service
ComputeResources
Service Provider #2
MedicalImage
Database
Available Services Aggregations of MICL SDEs
Workflow submission
Invocations &Notifications
Result management
GridFTP
• large data set of MRI brains• ~550 normal volunteers• age range 20-80 years• T1, duel echo (proton
density and T2 weighted) • acquired at 3 different sites• publicly available to
research community
Reference image Subject image
Rigid + non-rigid registration
Manual segmentation Computed boundary of calcaneus
Application: Automatic delineation of bones
calcaneus
Elastic Transformation
Apply transformation
Business idea
3 months later 6 months later 9 months later
Automated image analysis for the Drug Development ProcessMeasurement in Rheumatoid Arthritis: eg. grade of bone erosion
Baseline
Start of treatment using drug candidate
From the idea to a business plan ...
Prof. David Hawkes
Prof. Jo Hajnal
Prof. Derek Hill
Prof. Daniel Rueckert
Founders of IXICO:
Ixico Presentation
Business plan
The Market
The industry
Product & Unique Selling Preposition
Sustainable advantage
The team
Intellectual Property
Finance (forecast, funding requirements)
Exit strategy
Ixico Presentation
A crisis in drug discovery & development
Imaging can help…
“Imaging is a key technology for assessing, accelerating the development of, and guiding the use of new therapeutic options”George Mills, FDA
“It is estimated that between $10m
and $30m could be saved per drug
program through the use of advanced
imaging technologies which could
guide go/no go decisions.”
Business Insights Journal, May 2006
Ixico Presentation
The Market
The customers are the pharmaceutical and biotech companies
What does an “imaging biomarker” product look like?
Ixico Presentation
Imaging supplier
Image Analysis in clinical trials
Results for databaseupload
Data transfermetadata
Image scoringand analysis
Procedures
Study set-up, regulatory charter,Site management(training, SOPs…)
Ixico Presentation
Potential revenue growth
Potential Market compared to projected sales
Total Market£161.9m
IXICO£??m
CNS & Oncology£124.6m
0
20
40
60
80
100
120
140
160
180
2006 2007 2008 2009 2010
Years
£m
Ixico Presentation
Image analysis suppliers
Dominated by small number of companies that specialise in traditional “radiological reading”
Fastest growth is in more sophisticated methods
– Regulator push for methods that are less subjective
– Industry need for more precise measures for better, faster, cheaper decision making
Ixico Presentation
Product
•Advice on image acquisition parameters•Automated image analysis service•Quality control of analysis results•Quality Management System•Electronic audit trail •Site Management •Data logistics • Imaging charter for regulators (FDA)
Ixico Presentation
Product
•Advice on image acquisition parameters•Automated image analysis service•Quality control of analysis results•Quality Management System•Electronic audit trail •Site Management •Data logistics • Imaging charter for regulators (FDA)
Ixico Presentation
USP
Image acquisition Image analysis+
Due diligence conducted by the Capital Fund identified unique combination of:
Technology Know-how Quality system+ +
The team
Dr. Thomas Hartkens
Dr. Kate McLeish
James Cameron
Richard Brooman
Prof. Derek Hill
Ian Alexander
IXICO's non executives•Peter Morgan – Chairman– Executive career with IBM and experience on several company boards– Formerly Director General of the Institute of Directors in London.
•Prof. Sir Michael Brady–Track record of building successful entrepreneurial companies using academic technology (most recently Mirada, acquired by CTI)
•Prof. Sir Colin Dollery–Clinical pharmacologist and leading authority on Clinical Trials
–Senior consultant in GlaxoSmithKline (GSK): adviser to Dr Yamada the Chairman of Research and Development and member of the committees that review all protocols for human research, on the Development Investment Board and on the Global Safety Board
•Peter Ellis–Executive Director of People in Health and previously senior partner in Ernst & Young's Global Consulting life sciences practice.
Good Clinical Practise (GCP)
Audit Trail Electronic audit trail compliant with FDA 21 CFR part 11
Quality Management SystemISO 9001:2000 certification
Software Validatione.g. V-model
31
Automatic task
Image analysisapplication
IWD
1 Coarse Registr. Date:11.12.2006 13:45 Result:Transformation Server:234.232.43.22 Bias Correction3 QC interface
Results & Audit trail
information
32
Task processing
IWD
1 Coarse Registr. Date:11.12.2006 13:45 Result:Transformation Server:234.232.43.22 Bias Correction3 QC interface
33
Task processing
IWD
1 Coarse Registr. Date:11.12.2006 13:45 Result:Transformation Server:234.232.43.22 Bias Correction Date:13.12.2006 3:59
Result:Image Server:234.256.43.83 QC interface
34
Manual task
Custom-made interface
IWD
1 Coarse Registr. Date:11.12.2006 13:45 Result:Transformation Server:234.232.43.22 Bias Correction Date:13.12.2006 3:59
Result:Image Server:234.256.43.83 QC interface
Date:15.12.2006 9:45 Result:Comment
Person:Kate McLeish
Kate McLeish
Results & Audit trail
information &Digital Signature
35
IXICO's Electronic Audit Trail
Result images
Electronic Audit TrailElectronic Audit TrailElectronic Audit Trail IWD 1 Coarse Registr. Date:11.12.2006 13:45 Result:Transformation Server:234.232.43.2
2 Bias Correction Date:13.12.2006 3:59
Result:Image
Server:234.256.43.83 QC interface
Date:15.12.2006 9:45 Result:Comment
Person:Kate McLeish
Kate McLeish
21 CFR Part 11Electronic records; electronic signatures
Designed to provide FDA assurance that electronic records / signatures are– Trustworthy– Reliable
Equivalent to paper records and handwritten signatures
Applies to all FDA regulated industries
Requirements
Electronic recordsControls aimed at security and integrity of recordsWhat information needs to be included with a signed electronic record How the signature must be linked to the record
Electronic signaturesUnique to an identified individualComponents of an electronic signature (in general, 2)Security and integrity of ID codes and passwords
Software validation
IXICO's Workflow System
IXICO's Electronic Paper Trail
Automatic Task
IWD
1 Coarse Registr.2 Bias Correction3 QC interface
IWD
1 Coarse Registr. Date:11.12.2006 13:45 Result:Transformation Server:234.232.43.22 Bias Correction3 QC interface
IWD
1 Coarse Registr. Date:11.12.2006 13:45 Result:Transformation Server:234.232.43.22 Bias Correction Date:13.12.2006 3:59
Result:Image Server:234.256.43.83 QC interface
IWD
1 Coarse Registr. Date:11.12.2006 13:45 Result:Transformation Server:234.232.43.22 Bias Correction Date:13.12.2006 3:59
Result:Image Server:234.256.43.83 QC interface
Date:15.12.2006 9:45 Result:Comment
Person:Kate McLeish
Kate McLeish
IWD
1 Coarse Registr. Date:11.12.2006 13:45 Result:Transformation Server:234.232.43.22 Bias Correction Date:13.12.2006 3:59
Result:Image Server:234.256.43.83 QC interface
Date:15.12.2006 9:45 Result:Comment
Person:Kate McLeish
Kate McLeish
Derek Hill
Manual Task
Automatic Task
ReportAnalysis has shown that XY% of all cases ....
ISO 9001:2000
Does NOT say anything about the quality of the product.
BUT guarantees a constant product quality or an improvement of quality over time.
“Idea” of QMS
1.Write down what you are doing.2.Follow what you have written down.3.Improve documentation in a “controlled” way.
Preserve and improve knowledge base
ISO 9001:2000 certification
Does NOT certify a product Butcertifies that the company works according to the ISO 9001:2000 standard
ISO 9001:2000 based QMS4. Quality Management System4. Quality Management System
5. Management responsibilities5. Management responsibilities
6. Resource management6. Resource management
7. Product realization7. Product realization
8. Measurement, analysis and improvement8. Measurement, analysis and improvement
Written procedures System deployment procedure
Contract Agreement
Contract Amendment
Contract Processing Overview
Site Management
Site TrainingProtocol Optimization
Adverse Findings
Data Transfer
Data Admission
Data Export
Data Delivery Software Design
Workflow System Deployment
Workflow design
Software Transfer
Control of documents
Control of recordsInternal audit
Issue tracking
Control of Nonconformity, Corrective and Preventive Action
Customer Feedback
Document Review
Document Version ControlTraining
Recruitment
Purchasing
Risk Assessment
IP Transfer
Management responsibility
To define measurable objectives
Involvement of people: creates an environment where people become fully involved in achieving the organization's objectives
Leadership: is needed to provide unity of purpose and direction
Review
Product realization
Customer-related processes–Determination / review of product requirements –Customer communication
Design and development–Definition of inputs and outputs of design–Planning–Verification and Validation
PurchasingProduction and service provisionControl of service provision; Validation of service Identification and traceability of product
FDA: Software validation
Diagram by Alford Taylor, Director, Devision of Electrical and Software Engineering, CDRH / FDA
Conclusion
•Innovations in the technology might not be “selling point”
•Research implementation <-> production system
•Electronic company cannot be implemented as a business without quality standards.
•Exciting journey