Nov. 5, 2009 Grinnell College

Spectrum of High Performance Spectrum of High Performance Medical Imaging InformaticsMedical Imaging Informatics

Jun Ni, Ph.D.Jun Ni, Ph.D.Associate Professor, Dept. of Radiology Associate Professor, Dept. of Radiology

Director, Medical Imaging HPC & Informatics LabDirector, Medical Imaging HPC & Informatics LabCarver College of MedicineCarver College of Medicine

The University of IowaThe University of Iowa

Medical Imaging Digital Radiology

Medical Imaging Software Resources

Medical Imaging Hardware Facility

Medical Imaging Workforce is needed forKnowledge based Computer-aided Diagnostics

MII Domain DefinitionMII Domain Definition

Medical Imaging Informatics (MII)Medical Imaging Informatics (MII) == == Radiology Informatics?Radiology Informatics?

One of medical informatics disciplinesOne of medical informatics disciplinesSubSub--specialty of radiologyspecialty of radiology

Boundary Boundary ------ > medical image data mining?> medical image data mining?Technical driven: Technical driven: teleradiologyteleradiology/telemedicine/telemedicine

Job market: 70,000 on demand, education challengesJob market: 70,000 on demand, education challenges

Iowa Iowa ------ Hawkeye Radiology Informatics (HRI)Hawkeye Radiology Informatics (HRI)

Visions: Radiology InformaticsVisions: Radiology Informatics

Frontier in cancer diagnosticsFrontier in cancer diagnosticsProliferated applications:Proliferated applications:

Oncology, cardiology, dermatology, surgery, Oncology, cardiology, dermatology, surgery, gastroenterology, obstetrics, gynecology and gastroenterology, obstetrics, gynecology and pathology, and other medical fieldspathology, and other medical fields

Strong digital requirement and IT engagementStrong digital requirement and IT engagement

Medical Imaging Informatics (MII) Medical Imaging Informatics (MII) ScopeScope

What is current scope of MII?What is current scope of MII?A subspecialty of radiology that aims to improve A subspecialty of radiology that aims to improve medical medical imaging relatedimaging related discovery and technical services within the discovery and technical services within the healthcare enterprisehealthcare enterprise

Accuracy (methodology)Accuracy (methodology)Efficiency (workflow)Efficiency (workflow)Usability (feasibility or applicability)Usability (feasibility or applicability)Reliability (accessibility)Reliability (accessibility)SustainabilitySustainabilityCost/performanceCost/performance

Its ultimate goal to Its ultimate goal to improve health care systemsimprove health care systems

Previous Paradigm: Data Oriented Previous Paradigm: Data Oriented RoadmapRoadmap

Study how medical images (within radiology and Study how medical images (within radiology and throughout medical enterprise) are processed by throughout medical enterprise) are processed by

AcquisitionAcquisitionArchivingArchivingRetrieving/recoveringRetrieving/recoveringImage ProcessingImage Processing

AnalyzedAnalyzedEnhancedEnhancedVisualizedVisualizedData format conversionData format conversion……

Current Paradigm: CrossingCurrent Paradigm: Crossing

A multidiscipline and interdisciplinaryA multidiscipline and interdisciplinaryIntersection with other fields:Intersection with other fields:

Medical science (radiology, internal medicine, Medical science (radiology, internal medicine, neuroscience, neuroscience, ……))Computer and information scienceComputer and information scienceBiomedical engineeringBiomedical engineeringElectrical engineering (signal and data processing)Electrical engineering (signal and data processing)Biological and physiological sciences Biological and physiological sciences Medical physicsMedical physics……

hospital registration order exam waiting room

exam operation

modality

send to officePaperwork film packageRadiologist previewFetch report to HIS

radiologist review

final report on RIS Workflow

Digitization In Medical SciencesDigitization In Medical Sciences and Data Issueand Data Issue

Source: UC Berkeley, School of Information Management and Systems.

0 C.E.

2003

40,000 BCE cave paintings

bone tools 3500 writing

paper 1051450

printing1870

electricity, telephonetransistor 1947

computing 1950

1993The Web

Digital Cardiology

Electronic Medical Record

E-Health Initiatives/Linkages

Digital Radiology

1999

Late 1960sInternet

P e t a b y t e s

Digital Pathology

NM (128, 128)

MRI (256, 256)

CT (512, 512)

DSA (1024, 1024)

CR (2048, 2048)

Mammogram (4096, 4096)

PACS ChallengesPACS Challenges

Different regional and industrial interpretation, Different regional and industrial interpretation, configuration, and implementationconfiguration, and implementationDifferent interfaces and prototypesDifferent interfaces and prototypesDifferent standardization Different standardization

DICOM, HL7, Other IT standardsDICOM, HL7, Other IT standards

Different image digitalization of modalitiesDifferent image digitalization of modalitiesDifferent scopesDifferent scopes

PACSPACS--IT Technical ComponentsIT Technical Components

Image acquisition and management technologyImage acquisition and management technologyData visualization or image displayData visualization or image displayNetwork and communicationsNetwork and communicationsComputer application softwareComputer application software

PACS Technical ConcernsPACS Technical Concerns

Data MigrationBack-up archiveFault-toleranceIntegration with legacy systemsFast wide-area networksSecurity

PACS Distributed Computing PACS Distributed Computing ArchitectureArchitecture

Large scale (multiple Large scale (multiple modulemodule--based): based):

Module 1 Module 2 Module 3

Distributed multiple-

modules within multiple services units; but with single health organization

Local networked

PACS ClassificationPACS Classification

Super scale (enterpriseSuper scale (enterprise--, , cyberinfrastructurecyberinfrastructure--, heterogeneous, , heterogeneous, distributed griddistributed grid--based, cross organization, or even globally): based, cross organization, or even globally):

Module 1 of site A Module 2 of site A

Module 1 of Site B Module 2 of site B

High speed network

MII Challenges (1)MII Challenges (1)

Lack generic MII ontology (Lack generic MII ontology (Philological IssuePhilological Issue))Systematic identification and classification of domain Systematic identification and classification of domain entities and existences, and entity relations entities and existences, and entity relations (communications)(communications)No semantic languages for communications or No semantic languages for communications or workflowsworkflowsLooselyLoosely--defined terminologydefined terminologyNo linkage and leverage to knowledge, artificial No linkage and leverage to knowledge, artificial intelligent (AI), decision makingintelligent (AI), decision making

Ontological Data Model in Ontological Data Model in Radiology Informatics?Radiology Informatics?

What is ontology?What is ontology?Existing, Entity and relationshipsExisting, Entity and relationships

Domain ontology?Domain ontology?Domain, domain model, scope, boundary, crossing, Domain, domain model, scope, boundary, crossing, machine, object and service model, class, object, machine, object and service model, class, object, services, process, services, process, ……

Medical Imaging Informatics ontology?Medical Imaging Informatics ontology?DD--EE--RR--Graph and Machine descriptionsGraph and Machine descriptions

Ontology Wisdom

Philology

Knowledge Decision MakingCAD

Cognitive Sciences

Information Science

Data Data Data Data Data

Metadata MetadataMetadata

Information Management

Domain

Entity Relations

Artificial Intelligence

Strategic ChallengesStrategic Challenges

Ontological data model (terminology Ontological data model (terminology classification, entity definition, and relations classification, entity definition, and relations establishing) (Methodology issue)establishing) (Methodology issue)KnowledgeKnowledge--drivendrivenArtificial intelligenceArtificial intelligence--driven driven Unprecedented capacity for handling massive Unprecedented capacity for handling massive datadataSystem integration and interoperation among System integration and interoperation among various hospital/clinic systemsvarious hospital/clinic systemsExpansion of RI domain scopeExpansion of RI domain scope

RI Challenges (3)RI Challenges (3)

No standard protocols (No standard protocols (Technical issuesTechnical issues))To facilitate the interoperation and communication To facilitate the interoperation and communication among globallyamong globally--distributed MII resources distributed MII resources To deploy concurrent hardware and software To deploy concurrent hardware and software solutions solutions To utilize cyberTo utilize cyber--enabled highenabled high--speed networksspeed networks

Short of education/training programs (Short of education/training programs (Business Business issueissue))

To foster the next generation in digital health care To foster the next generation in digital health care systems.systems.

RI Challenges (4)RI Challenges (4)

Software DevelopmentSoftware DevelopmentComputerComputer--Aided Detection and Diagnosis (CAD)Aided Detection and Diagnosis (CAD)ComputerComputer--aided interventional radiologyaided interventional radiologyMetrics and computing performanceMetrics and computing performanceMedical imaging facility and infrastructure Medical imaging facility and infrastructure developmentdevelopmentFundamental research and developmentFundamental research and development

Medical Imaging Service Pack (MISP)Medical Imaging Service Pack (MISP)Medical Imaging Informatics Knowledge Integration Medical Imaging Informatics Knowledge Integration Toolkit (M2KIT)Toolkit (M2KIT)

Lab MissionLab Mission

Establishment of a nationally and globallyEstablishment of a nationally and globally--recognized research lab in medical imaging recognized research lab in medical imaging informatics or radiology informaticsinformatics or radiology informatics

Short Term Action TasksShort Term Action Tasks

Learning any subjects and shaping knowledgeLearning any subjects and shaping knowledgeDevelop infrastructure of unprecedented computing Develop infrastructure of unprecedented computing facility in medical imaging informaticsfacility in medical imaging informaticsCollaborating with enterprise IT and health care Collaborating with enterprise IT and health care industrialsindustrialsWorking with external and internal professionalsWorking with external and internal professionalsSeeking for fundsSeeking for fundsDevelop software solutions for future health care Develop software solutions for future health care systemssystemsAttract more people including you.Attract more people including you.

LongLong--Term GoalTerm Goal

Computation (future projects)Computation (future projects)Infrastructure and algorithm developments for data Infrastructure and algorithm developments for data mining in medical imagemining in medical imageArtificial Intelligence in medical imagingArtificial Intelligence in medical imagingLargeLarge--scale image processing and associated scale image processing and associated modeling and simulationsmodeling and simulationsDigitalization of human body (massDigitalization of human body (mass--phantom phantom system)system)Computational radiologyComputational radiologySystem radiologySystem radiology

MIHI Lab ProjectsMIHI Lab Projects

Medical Imaging & Radiology Informatics (MIRI) Medical Imaging & Radiology Informatics (MIRI) Hawkeye Radiology Informatics (HRI) Hawkeye Radiology Informatics (HRI)

http://http://www.uiowa.edu/~hriwww.uiowa.edu/~hri//Radiology Informatics Domain Ontology (RIDO)Radiology Informatics Domain Ontology (RIDO)Medical Imaging Informatics Ontology (MIIO)Medical Imaging Informatics Ontology (MIIO)Medical Imaging Informatics Terminology (MIIT)Medical Imaging Informatics Terminology (MIIT)CyberinfrastructureCyberinfrastructure--enabled Radiology Informatics (CIRI)enabled Radiology Informatics (CIRI)

Medical Imaging Information System (MIIS)Medical Imaging Information System (MIIS)http://http://www.uiowa.edu/mihpclab/projects_miis.htmlwww.uiowa.edu/mihpclab/projects_miis.html

Radiology Informatics Education and Training (RIET)Radiology Informatics Education and Training (RIET)http://http://www.uiowa.edu/~hri/education.htmlwww.uiowa.edu/~hri/education.html

ProjectsProjects

Parallel Computing in Medical Imaging (PCMI)Parallel Computing in Medical Imaging (PCMI)http://http://www.uiowa.edu/mihpclab/projects_pcmi.htmlwww.uiowa.edu/mihpclab/projects_pcmi.html

Parallelism of Medical Imaging ProcessingParallelism of Medical Imaging ProcessingCT ReconstructionCT ReconstructionSegregationSegregationRegistrationRegistrationTexturing and classificationTexturing and classificationEnhancementEnhancementImage compressionImage compressionImage data miningImage data mining……

Parallel CT Medical Image Parallel CT Medical Image ReconstructionReconstruction

LargeLarge--scale Parallel CT Medical Image scale Parallel CT Medical Image ReconstructionReconstruction

CT Technology CT Technology Invented by British Engineer G. Invented by British Engineer G. HounsfieldHounsfield in 1971in 1971Principle: utilizes XPrinciple: utilizes X--ray technology and computers to ray technology and computers to create images of crosscreate images of cross--section section ““slicesslices”” through the through the bodybody

LargeLarge--scale Parallel CT Medical Image scale Parallel CT Medical Image ReconstructionReconstruction

TodayToday’’s CT Technology s CT Technology Advanced in technology, software applications and Advanced in technology, software applications and clinical performanceclinical performanceCT scanners are fast and patient friendlyCT scanners are fast and patient friendlyExpand the role of CT in both diagnosis and Expand the role of CT in both diagnosis and treatment. treatment.

CT Technology BasisCT Technology Basis

XX--ray CT technologies ray CT technologies Classification: XClassification: X--ray beamray beam’’s s geometry, motion of Xgeometry, motion of X--ray locus ray locus (source), and design of (source), and design of corresponding detectors which corresponding detectors which measure the decay of Xmeasure the decay of X--ray ray intensity. intensity.

Classified by beam geometryClassified by beam geometryParallel BeamParallel BeamFan BeamFan BeamCone BeamCone Beam

Classified by the motion of XClassified by the motion of X--ray locusray locusCircleCircleSpiral Spiral

Classified by detectorClassified by detectorOne rowOne rowMultiple rowsMultiple rows

Tube (X-ray source)

CT Technology BasisCT Technology Basis

Generations:Generations:First Generation: First Generation:

ParallelParallel--beam, in which a beam, in which a single Xsingle X--ray tube generates a ray tube generates a beam passed through the beam passed through the object in parallel and a single object in parallel and a single detector obtains an optical detector obtains an optical signal correspondinglysignal correspondinglyThe whole system is in a The whole system is in a translationtranslation--thenthen--rotation rotation manner time consumingmanner time consuming

X-raydetector

source

Object or patient

Parallel bean with multiple X-ray sources

CT Technology BasisCT Technology Basis

Second Generation: Second Generation: Fan beam of XFan beam of X--rays and a rays and a linear detector array linear detector array (multiple detectors on the (multiple detectors on the plane). plane). The XThe X--ray source emits ray source emits radiation over a large angle, radiation over a large angle, while every detector in the while every detector in the group receives the signals group receives the signals (which are called (which are called projection data). projection data). Improved efficiencyImproved efficiencyEmploys a translateEmploys a translate--rotate rotate scanning motionscanning motionCorresponding Corresponding reconstruction algorithm is reconstruction algorithm is a little more complexa little more complex

Fan Beam with one single X-ray source

CT Technology BasisCT Technology Basis

Third GenerationThird GenerationCT scanners uses coneCT scanners uses cone--beam. beam. The detector array in these The detector array in these scanners remains stationary scanners remains stationary while Xwhile X--rays are produced rays are produced by a highby a high--energy electron energy electron beam, rotating around a beam, rotating around a patient without moving the patient without moving the CT scanner. CT scanner. This kind of CT scanner This kind of CT scanner system is sometimes system is sometimes referred to as a rotatereferred to as a rotate--stationary or rotatestationary or rotate--only only geometrical system. geometrical system.

z

x

y

Cone Beam with single X-ray source and multiple row’s detector

CT Technology BasisCT Technology Basis

Fourth generationFourth generationThe Helical (Spiral) CT scanner, The Helical (Spiral) CT scanner, first invented in 1989, used an first invented in 1989, used an innovative scanning mechanism in innovative scanning mechanism in which the gantry rotates which the gantry rotates continuously with the continuously with the simultaneous translation of the simultaneous translation of the patient table. patient table. With the motion of the patient With the motion of the patient table, the scanner can reconstruct table, the scanner can reconstruct a large number of slices and a large number of slices and produce a 3D image of the whole produce a 3D image of the whole object. object.

Spiral Cone Beam with single X-ray source and multiple row’s detector

CT Technology BasisCT Technology Basis

Fourth generationFourth generationHelical CT scanners are often cataloged into singleHelical CT scanners are often cataloged into single-- (single(single--slice and single detector row), dualslice and single detector row), dual-- (dual(dual--slice and dual slice and dual detector row), or multidetector row), or multi-- (multi(multi--slice and multislice and multi--detector, or detector, or multimulti--row) sections according to the row number of detector row) sections according to the row number of detector elements. elements. A MultiA Multi--Section CT (MSCT) scanner deploys a cone beam Section CT (MSCT) scanner deploys a cone beam projection (single Xprojection (single X--ray source and planner arrays of ray source and planner arrays of detectors); thus further speeding up data collections or detectors); thus further speeding up data collections or acquisitions. acquisitions.

PhilipsSiemensGE GE

Image Reconstruction BasisImage Reconstruction Basis

Image reconstruction algorithmsImage reconstruction algorithmsConstruct images based on projection data from scannersConstruct images based on projection data from scannersAssociated with the evolution of fourth generations of CT Associated with the evolution of fourth generations of CT systems (geometrical design and spatial motion)systems (geometrical design and spatial motion)

Image Reconstruction BasisImage Reconstruction Basis

A twoA two--step processstep processthe transmission of an Xthe transmission of an X--ray beam is ray beam is measured through all possible straightmeasured through all possible straight--line paths in a plane of an object line paths in a plane of an object the attenuation of an xthe attenuation of an x--ray beam is ray beam is estimated at points in the object estimated at points in the object

Attenuation coefficientAttenuation coefficientu is the x-ray linear attenuation coefficientP(xP(x) is a projection function) is a projection functionAn attenuation function f(x,y) for 2D object; To evaluate f(x), while p(x) is given

eIIL

dxxu

t∫= −0

)(

0

00

( ) ln( ) ( )L tIf x dx P x

I= − =∫

Image Reconstruction BasisImage Reconstruction Basis

Theorem 1Theorem 1The value of a 2D function at an The value of a 2D function at an

arbitrary point is uniquely obtained arbitrary point is uniquely obtained by integrals along the lines of all by integrals along the lines of all directions passing the point. directions passing the point.

Mathematically inversed Mathematically inversed problemproblem

P: observed data.X: unknown original imageA: non-zero M by N matrix

1 2

1 2

( , ,..., )

( , ,..., )

TM

TN

Ax PP p p p

x x x x

=

=

=

Image Reconstruction BasisImage Reconstruction Basis

Analytic Algorithms (Filtered BackAnalytic Algorithms (Filtered Back--Projection, FBP)Projection, FBP)

Efficient computationEfficient computationPredominant in commercial marketPredominant in commercial marketSensitive to noise, inaccurate projection dataSensitive to noise, inaccurate projection data

Iterative Algorithms (ART, EM)Iterative Algorithms (ART, EM)Tremendous computation and easy implementationTremendous computation and easy implementationHighHigh--quality reconstructed image from noisy or lowquality reconstructed image from noisy or low--dose and incomplete projection datadose and incomplete projection dataWeight or penalty functions to redeem the loss of Weight or penalty functions to redeem the loss of project dataproject data

Image Reconstruction BasisImage Reconstruction Basis

Analytic Algorithms Analytic Algorithms (Filter Back(Filter Back--Projection)Projection)

Radon Transformation Radon Transformation (geometrical)(geometrical)Fourier transformation Fourier transformation (project(project--data preprocessing)data preprocessing)FilteringFilteringFourier inverse transformationFourier inverse transformation

( , ) ( cos sin , sin cos )

cos sin cos sin,

sin cos sin cos

g s f s u s u du

s x x su y y u

θ θ θ θ θ

θ θ θ θθ θ θ θ

∞

−∞= − +

−⎡ ⎤ ⎡ ⎤⎡ ⎤ ⎡ ⎤ ⎡ ⎤⎡ ⎤= =⎢ ⎥ ⎢ ⎥⎢ ⎥ ⎢ ⎥ ⎢ ⎥⎢ ⎥−⎣ ⎦ ⎣ ⎦⎣ ⎦ ⎣ ⎦ ⎣ ⎦⎣ ⎦

∫

Radon Transformation

Image Reconstruction BasisImage Reconstruction Basis

Projection TheoremProjection TheoremThe One-dimensional Fourier transform of the Radon transform g(θ,s) for s denoted Gθ

(ξ)

variable, and the cross-section of the two-

dimensional Fourier transform of the object f(x,y), sliced by the plane at θ

with the fx

-axis and perpendicular to the (fx

, fy

)-plane, denoted F(fx

, fy

), are identical to Gθ

(ξ)=F(ξcosθ,ξsinθ) )sin,cos()( θξθξξθ FG =

Image Reconstruction BasisImage Reconstruction Basis

Filter BackFilter BackProjection (FBP)Projection (FBP)

)sin,cos()( θξθξξθ FG =

θξθπ

dsgFFyxf }||)},({{),(0

1∫ −=

Parallel Parallel KatsevichKatsevich

AlgorithmAlgorithm

Deng, Yu, Ni, et al. The Journal of Supercomputing, 38, 35–47, 2006

KasevitchKasevitch

AlgorithmAlgorithm

Hilbert Filtering, intermediate Hilbert Filtering, intermediate funtionfuntion

Weighted Weighted BackprojectionBackprojection

( )bs x

PI-segment

x

( )sy

gantry

h R

Ulocus

( )ts x

detector

1d2d( , , )g s u v

β

3d

r

Geometrical illustration of the helical cone-beam CT system

2

2( ) 0

1 1( ) ( ( ), ( , , ))

2 ( ) sin( )PI

f q sI x

df x D y q s x ds

x y s q

π γγ

π λ=

∂= − Θ

− ∂∫ ∫

2

0

( , ) ( ) ,fD y f y t dt Sβ β β∞

= + ∈∫

2 2 2

2 2 2( , , ) ( , , )

( )g

D u vs u v D s u v duD u v u u

ψ∞

−∞

+ +=

+ + −∫ % % %

% % %

2 2

( , , ) ( , , )gD u uvD s u v g s u v

s D u D v⎛ ⎞∂ + ∂ ∂

= + +⎜ ⎟∂ ∂ ∂⎝ ⎠

( )( )

( )( )

1 2

3 3

( )

2 ( ) ( ) ( )* , *

( ) ( )

1 1( ) ( , *, *)2 ( )

t

b

s

s D s D su v

s s

f s u v dssψ

π − −= =

− −

= −−∫

x

x x y d x y dx y d x y d

xx y

Parallel Parallel KatsevichKatsevich

AlgorithmAlgorithm

Parallel implementationsParallel implementations

parallel reconstruction process

PE 1 PE 1

PE nPE n

Root PE

Projection data

Filtereddata

Reconstructed data

Collectedreconstructe d data

Filtering stage Backprojection stage

PE’s Initialization

Projection Data Generation/Distribution

Projection Data Filtration

Projection Data Gathering and Distribution

Backprojection

Gathering Reconstructed Data on Root PE

PE’s Finalization

0 8 16 24 320

8

16

24

32Case ICase IICase IIICase IVIdeal Speedup

x

y

0 10 16 24 32101

102

103

104

Case ICase IICase IIICase IV

y

0 8 16 24 320

0.2

0.4

0.6

0.8

1

1.2

Case ICase IICase IIICase IVIdeal Speedup

Data: 3-D Shepp-Logan phantom : 1283, 2563, 3843, 5123

time speedup

efficiency

0 50 100 150 200 250 300 350 4000

50

100

150

200

250

300

350

400

2563

5123

Ideal speedupTeraGrid/NCSA cluster

ProjectsProjectsModeling Biotransport in Biophysical System (MBBS)Modeling Biotransport in Biophysical System (MBBS)

http://http://www.uiowa.edu/mihpclab/projects_mbbs.htmlwww.uiowa.edu/mihpclab/projects_mbbs.html

NanothermotheropyNanothermotheropy ((nanoHyperthmiananoHyperthmia))http://http://www.uiowa.edu/mihpclab/projects_nmni.htmlwww.uiowa.edu/mihpclab/projects_nmni.html

Tumor growth and dynamics (computational oncology)Tumor growth and dynamics (computational oncology)

Optical Imaging Tomography and Applications (OITA)Optical Imaging Tomography and Applications (OITA)http://http://www.uiowa.edu/mihpclab/projects_oita.htmlwww.uiowa.edu/mihpclab/projects_oita.html

ProjectsProjects

Stereological Analysis and Tumor Volume Stereological Analysis and Tumor Volume Metrics (SATVM)Metrics (SATVM)

voluMeasurevoluMeasure Software Development Project (RSNA'05) Software Development Project (RSNA'05)

The Effect of the Shape and Orientation of a Mass on the AccuracThe Effect of the Shape and Orientation of a Mass on the Accuracy y Estimating Its Size Using RECIST (RSNA'09)Estimating Its Size Using RECIST (RSNA'09)

Tumor volume measurement in MRI breast imagingTumor volume measurement in MRI breast imaging

http://http://www.uiowa.edu/mihpclab/projects_isca.htmlwww.uiowa.edu/mihpclab/projects_isca.html

StereotacticStereotactic Atlas for the Anatomic Topology Atlas for the Anatomic Topology (SAAT) (SAAT)

http://http://www.uiowa.edu/mihpclab/projects_saat.htmlwww.uiowa.edu/mihpclab/projects_saat.html

Couple Diffusions for Image Enhancement (DDIE) Couple Diffusions for Image Enhancement (DDIE) http://http://www.uiowa.edu/mihpclab/projects_cdie.htmlwww.uiowa.edu/mihpclab/projects_cdie.html

ProjectsProjectsKnowledgeKnowledge--based CAD for Breast Imaging based CAD for Breast Imaging (KCBI) (KCBI)

architectural distortionarchitectural distortioncalcification calcification DeformationDeformation

http://http://www.uiowa.edu/mihpclab/projects_kcbi.htmlwww.uiowa.edu/mihpclab/projects_kcbi.html

New projectsNew projectsTomosynthesisTomosynthesis and Molecular Breast Imagingand Molecular Breast ImagingUS Medical ImagingUS Medical Imaging3D Volume Rendering3D Volume Rendering

Sponsorships and CollaborationsSponsorships and Collaborations

Current sponsorsCurrent sponsorsNIH (HPC medical imaging)NIH (HPC medical imaging)NSF (HPC computations in nanotechnology)NSF (HPC computations in nanotechnology)Intel (HPC)Intel (HPC)MicrosoftMicrosoft

Collaborators:Collaborators:Siemens (medical modality, MII software resources)Siemens (medical modality, MII software resources)IBM (Cell/BE)IBM (Cell/BE)NavidaNavida (GPU/CUDA)(GPU/CUDA)Mayo Clinic (projects)Mayo Clinic (projects)You who love to support this missionYou who love to support this mission

Go Hawkeye!Go Hawkeye!

Thanks!Thanks!

Q & AQ & A