A Vision for e-Research Prof. Sir Mike Brady : A Vision for e-Research Prof. Sir Mike Brady


Slide2: Oxford e-Research Centre a vision and some examples Professor Sir Michael Brady FRS FREng Department of Engineering Science Oxford University


Changing research scene: Changing research scene Research on a larger scale, scope and ambition than the norm for 3 year grants Globally, research funding bodies are encouraging projects of greater ambition and across institutions/disciplines Such projects mostly do not sit comfortably in a single Department, Division, … Interviews with research leaders across the University reveals identical issues and aspirations The emerging IT infrastructure facilitates such developments. e-Science = data driven science + computational science


Overview: Overview Databases and archives Integrated analysis of complex systems Architecture of complex systems Rapid execution of computationally intensive systems Effective inter-disciplinary collaboration On-demand (adaptive) creation of exemplars Toolkits to support state-of-the-art research


Modelling the shape of breast tumours: Modelling the shape of breast tumours Regions defined by dense attenuation and significant changes in local phase Have associated descriptor of the shape of the region (left, here spiculated) Have associated texture descriptors learned from training set (textons from filter response  hidden MRF (right)


Slide6: Need for statistical power Screening results in about 6 cancers per 1000 cases A typical centre sees 10,000-15,000 screening cases annually, that is, 60-90 cancers: this is far too few to learn all the tumour models There are 93 screening centres in the UK, and hundreds more in Europe. Imagine pooling all their resources, providing the required statistical power, while respecting local data curation imposed by law!! The Grid potentially enables this at acceptable bandwidth and with guarantees on secure image/data transmission This assumes image transmission can be done sufficiently rapidly, and securely This motivated the eDiamond and now GIMI projects


Exploiting eDiamond: Image data mining: FindOneLikeIt: Exploiting eDiamond: Image data mining: FindOneLikeIt Search features include: boundary, shape, texture inside & outside, … Major developments since at Oxford: Gottlob’ semantic db, Zisserman’s visual words


Integrated analysis of complex systems: Integrated analysis of complex systems Improved CT and MR imaging (Tasks 1-5) Add PET and Optical as appropriate Genomic Pathology (Tasks 7 and 9) Pharmacogenomic therapy /toxicity selection (Task 8) Data Integration (Task 6) Prospective Trial (Task 10) Oxford University/GE Healthcare colorectal/liver cancer project: PIs Mike Brady & David Kerr (Clin Pharm). Depts: Eng, Clin Pharm, Biochem, IMM, Surgery, Radiology, Pathology DNA through drug trials through image analysis through patient decision making


Colorectal cancer: Colorectal cancer After removal of “bias field” Original image for multidisciplinary team 3D reconstruction of colorectum and mesorectal region Detected sentinel lymph nodes Bond, Brady, et. al, CARS 2005


Slide10: Modelling tumour development System of interacting reaction diffusion processes: image analysis & modelling (Eng Sc), confirmation by resection/histology (Surgery, Pathology, IMM), and mathematical modelling (Mathematics)


Simulated Dynamic PET Dataset: Simulated Dynamic PET Dataset Ali, Schottlander and Brady, Am. Molec. Im., March 2006 Positron Emission Tomography is rapidly improving, and promises to deliver time activity curves at regions of interest, sufficient to model the pharmacokinetics and dynamics of drug activity. But there are few machines available, so algorithm development is difficult! Solution: simulate on realistic systems, eg SORTEO. This is a major DTI/Siemens project with Eng Science and Chemistry


Validation of Simulated Data: Validation of Simulated Data Rigid registration performed with RevealMVS1 Simulated Data Patient Data Registered Composite Image 1 Siemens Molecular Imaging


PET Simulations: PET Simulations NCAT Human PET Phantom Real FHBG Patient PET Image Experimental TAC From PET Images Ali and Brady, Am. Molec. Im., March 2006 Schottlander, Brady, Reilhac, et. al. ISBI 2006 We have acquired a powerful 20 node computer cluster to reduce SORTEO simulations from a day to a few minutes, typically 500,000 positron decay events. We’d like 10*events on 2000 computers…


Effective inter-disciplinary collaboration: Effective inter-disciplinary collaboration GP Emergency Other Colonoscopy Diagnosis Staging Chemotherapy Radiotherapy Surgery Palliative care Imaging Pathology Detection Work up Therapy Follow-up Recurrence Surveillance Metastatic site referral Patient journeys in colorectal cancer


“MDT Suite”: Oxford + GE: “MDT Suite”: Oxford + GE Information integration Clinical Image analysis, image fusion (and signals) Histopathology images Efficient access to that information Integration with genetics information (eg microarray) More informed decision making Patient information, drug information/counter indications/… Remembering what was decided, and why What decisions were contingent on information to be gathered, and how previous tentative decisions may change Data analysis: Structured record of patient journey provides excellent opportunity for data mining MDT Decision Support Software


Architecture: Architecture MDT Suite Server Database PROforma Decision Support Images Text Data Hospital Patient Data Corporate Memory Patient Data Session Data MDT Suite Information Integration Projector Clinicians This is being developed in 5 Departments at Oxford University and with 2 groups in GE Research in the USA


Patient Summary: Patient Summary


Slide20: Used to create an atlas – the “average” brain, so that differences between this brain and the average can be noted The database of brains may comprise: young/old; male vs female; normal vs (many) diseases; left vs right handed; … The iXi project at Imperial/UCL/Oxford asked: can we dynamically create an atlas that is relevant for this patient? Ixi is now a start-up company. On-demand (adaptive) creation of exemplars


Slide21: Oxford University King’s College London (Guy’s Campus) IMPERIAL COLLEGE KING’S COLLEGE LONDON Create atlas


Toolkits to support research: Toolkits to support research Beyond Matlab, ITK, … Image analysis Medical image analysis MRI Cancer Colorectal/liver Breast … Brain images CT PET … Visual images Algorithms are available, eg over the web; but typically in several different programming languages, and deployed on different platforms WebServices emergent standards potentially offer a solution Microsoft project awarded to Anne Trefethen & Mike Brady to develop this idea


http://www.oerc.ox.ac.uk: http://www.oerc.ox.ac.uk