SlideShare ist ein Scribd-Unternehmen logo
1 von 25
Two-stage Unsupervised
Learning for 3D TOF-MRA
BISPL - BioImaging, Signal Processing,
and Learning lab.
KAIST, Korea
Chung et al, Medical Image Analysis, 2021
Introduction
• Contrast agent free Imaging method to visualize vessels
• Flow-related enhancement for high contrast in fresh blood
• 3D Acquisition for full volume coverage: Long scan time
• Often coupled with maximum intensity projection (MIP) to visualize only the vessels
Time-of-Flight Magnetic Resonance Angiography
Subsampling
ℱ ℱ−1
Reconstruction
𝓕−1
𝓕
Reconstruction
• Fully acquiring k-space
• Time consuming
• Subsampling k-space
• Accelerated scan time
• Aliasing artifacts
• Compressed-sensing(CS) for reconstruction
• Fill in missing k-space
• Iterative optimization
• Slow reconstruction
Acceleration of MRI
Deep Learning for Accelerated MRI
Supervised Learning
ℱ−1
• Deep Learning (DL) reconstruction
• Very fast computation
• High interest in research and application
• Supervised Learning
• Large amount of matched training data
• Restricted to certain situations
• Unsupervised Learning
• Without matching training data
• Wider application
• Generative Adversarial Networks (GAN)
Unsupervised Learning
Generative
Adversarial Networks
Generative Adversarial Nets (GANs)
https://www.cfml.se/blog/generative_adversarial_networks/
Wasserstein GAN
generator
discriminator
Geometry of GAN
subject to
Lei, Na, arXiv:1710.05488 (2017)
CycleGAN
J.-Y. Zhu, et al, CVPR, 2017
Theory & Methods
Geometry of CycleGAN
Forward physics
(unknown, partially known, known)
inverse solution
3D TOF-MRA
http://mri-q.com/2d-vs-3d-mra.html
• MOTSA multiple-slab 3D acquisition
• Partial 3D volume captured in slab
• Sub-sampling mask in coronal plane of each slab
Conventional cycleGAN for MR reconstruction
• Two sets of generator/discriminator for each mapping
• Forward measurement mapping replaceable
Two-stage cycleGAN: Step I (coronal)
• Physics-informed cycleGAN for stable training
• Complex multi-coil information processed in coronal plane
Replace generator with
MR forward operator
Two-stage cycleGAN: Step II (axial)
• Further enhancement in axial plane
• Multi-headed projection discriminator to capture blood activation
Multi-headed
Projection
Discriminator
Multi-headed Projection Discriminator
• Quality of MIP images equally important with source MRA
• Max-pooling along depth: pseudo-MIP
𝜑𝛾(𝐱) = 𝜑𝛾1
(𝐱) + 𝜑𝛾2
max
(𝐱)
3D information
2D Max-pooled
information
Overall Reconstruction Flow
• Coronal reconstruction in complex multi-coil domain
• Axial reconstruction in SSOS image domain
Step I: coronal recon. Step II: axial recon.
Results
In Vivo Results
In Vivo Results
Reconstruction of pathologic lesions
Quantitative Results
Comparison study
Radiological Evaluation
Summary
• Unsupervised reconstruction method for TOF-MRA proposed
• Two-stage multiplanar learning shown effective
• Projection discriminator
• Capture characteristics of 3D angiogram
• Extensive experiments verify superiority & generalizability
Questions?

Weitere ähnliche Inhalte

Was ist angesagt?

Pleiades - satellite imagery - very high resolution
Pleiades - satellite imagery - very high resolutionPleiades - satellite imagery - very high resolution
Pleiades - satellite imagery - very high resolutionSpot Image
 
07 a70401 remotesensingandgisapplications
07 a70401 remotesensingandgisapplications07 a70401 remotesensingandgisapplications
07 a70401 remotesensingandgisapplicationsimaduddin91
 
GoogleSky Status at Google
GoogleSky Status at GoogleGoogleSky Status at Google
GoogleSky Status at GoogleAlberto Conti
 
Creating A Multi-wavelength Galactic Plane Atlas With Amazon Web Services
Creating A Multi-wavelength Galactic Plane Atlas With Amazon Web ServicesCreating A Multi-wavelength Galactic Plane Atlas With Amazon Web Services
Creating A Multi-wavelength Galactic Plane Atlas With Amazon Web Services G. Bruce Berriman
 
Sonnentag phenocams 2014
Sonnentag phenocams 2014Sonnentag phenocams 2014
Sonnentag phenocams 2014aceas13tern
 
Underground utility survey
Underground utility surveyUnderground utility survey
Underground utility surveyDipak Kadam
 
Using Very High Resolution Satellite Images for Planning Activities in Mining
Using Very High Resolution Satellite Images for Planning Activities in MiningUsing Very High Resolution Satellite Images for Planning Activities in Mining
Using Very High Resolution Satellite Images for Planning Activities in MiningArgongra Gis
 
Milton analyticalconstellation
Milton analyticalconstellationMilton analyticalconstellation
Milton analyticalconstellationgreenstarfish
 
Accuracy of UAV Photogrammetry
Accuracy of UAV PhotogrammetryAccuracy of UAV Photogrammetry
Accuracy of UAV Photogrammetrybaselinesurvey
 
New features presentation: meteodyn WT 4.8 software - Wind Energy
New features presentation: meteodyn WT 4.8 software - Wind EnergyNew features presentation: meteodyn WT 4.8 software - Wind Energy
New features presentation: meteodyn WT 4.8 software - Wind EnergyJean-Claude Meteodyn
 
Lymphedema measurement using kinect volume reconstruction
Lymphedema measurement using kinect volume reconstructionLymphedema measurement using kinect volume reconstruction
Lymphedema measurement using kinect volume reconstructionWonjoongCheon
 
Ilris vs lynx highway surveying and data post processing - munich2008
Ilris vs lynx highway surveying and data post processing - munich2008Ilris vs lynx highway surveying and data post processing - munich2008
Ilris vs lynx highway surveying and data post processing - munich2008Michael Xinogalos
 
TEAM 3: Improving Open Land Use Map by using Satellite Data
TEAM 3: Improving Open Land Use Map by using Satellite DataTEAM 3: Improving Open Land Use Map by using Satellite Data
TEAM 3: Improving Open Land Use Map by using Satellite Dataplan4all
 
Digital Elevation Models - WUR - Grontmij
Digital Elevation Models - WUR - GrontmijDigital Elevation Models - WUR - Grontmij
Digital Elevation Models - WUR - GrontmijXander Bakker
 
Geospatial Research At UCL
Geospatial Research At UCLGeospatial Research At UCL
Geospatial Research At UCLJeremy Morley
 

Was ist angesagt? (20)

Pleiades - satellite imagery - very high resolution
Pleiades - satellite imagery - very high resolutionPleiades - satellite imagery - very high resolution
Pleiades - satellite imagery - very high resolution
 
Poster : STM/STS Techniques
Poster : STM/STS TechniquesPoster : STM/STS Techniques
Poster : STM/STS Techniques
 
Gabriele Candela_Image-based 3d reconstruction of a bamboo-steel spatial trus...
Gabriele Candela_Image-based 3d reconstruction of a bamboo-steel spatial trus...Gabriele Candela_Image-based 3d reconstruction of a bamboo-steel spatial trus...
Gabriele Candela_Image-based 3d reconstruction of a bamboo-steel spatial trus...
 
07 a70401 remotesensingandgisapplications
07 a70401 remotesensingandgisapplications07 a70401 remotesensingandgisapplications
07 a70401 remotesensingandgisapplications
 
GoogleSky Status at Google
GoogleSky Status at GoogleGoogleSky Status at Google
GoogleSky Status at Google
 
Icelandic Bathy model
Icelandic Bathy modelIcelandic Bathy model
Icelandic Bathy model
 
Creating A Multi-wavelength Galactic Plane Atlas With Amazon Web Services
Creating A Multi-wavelength Galactic Plane Atlas With Amazon Web ServicesCreating A Multi-wavelength Galactic Plane Atlas With Amazon Web Services
Creating A Multi-wavelength Galactic Plane Atlas With Amazon Web Services
 
Sonnentag phenocams 2014
Sonnentag phenocams 2014Sonnentag phenocams 2014
Sonnentag phenocams 2014
 
Underground utility survey
Underground utility surveyUnderground utility survey
Underground utility survey
 
Using Very High Resolution Satellite Images for Planning Activities in Mining
Using Very High Resolution Satellite Images for Planning Activities in MiningUsing Very High Resolution Satellite Images for Planning Activities in Mining
Using Very High Resolution Satellite Images for Planning Activities in Mining
 
Milton analyticalconstellation
Milton analyticalconstellationMilton analyticalconstellation
Milton analyticalconstellation
 
Accuracy of UAV Photogrammetry
Accuracy of UAV PhotogrammetryAccuracy of UAV Photogrammetry
Accuracy of UAV Photogrammetry
 
Photographic survey
Photographic survey Photographic survey
Photographic survey
 
New features presentation: meteodyn WT 4.8 software - Wind Energy
New features presentation: meteodyn WT 4.8 software - Wind EnergyNew features presentation: meteodyn WT 4.8 software - Wind Energy
New features presentation: meteodyn WT 4.8 software - Wind Energy
 
PanicO
PanicOPanicO
PanicO
 
Lymphedema measurement using kinect volume reconstruction
Lymphedema measurement using kinect volume reconstructionLymphedema measurement using kinect volume reconstruction
Lymphedema measurement using kinect volume reconstruction
 
Ilris vs lynx highway surveying and data post processing - munich2008
Ilris vs lynx highway surveying and data post processing - munich2008Ilris vs lynx highway surveying and data post processing - munich2008
Ilris vs lynx highway surveying and data post processing - munich2008
 
TEAM 3: Improving Open Land Use Map by using Satellite Data
TEAM 3: Improving Open Land Use Map by using Satellite DataTEAM 3: Improving Open Land Use Map by using Satellite Data
TEAM 3: Improving Open Land Use Map by using Satellite Data
 
Digital Elevation Models - WUR - Grontmij
Digital Elevation Models - WUR - GrontmijDigital Elevation Models - WUR - Grontmij
Digital Elevation Models - WUR - Grontmij
 
Geospatial Research At UCL
Geospatial Research At UCLGeospatial Research At UCL
Geospatial Research At UCL
 

Ähnlich wie Two stage deep learning for accelerated 3 d time-of-flight mra without matched training data

CycleMorph: Cycle consistent unsupervised deformable image registration
CycleMorph: Cycle consistent unsupervised deformable image registrationCycleMorph: Cycle consistent unsupervised deformable image registration
CycleMorph: Cycle consistent unsupervised deformable image registrationBoahKim2
 
Deep Learning Tomography
Deep Learning TomographyDeep Learning Tomography
Deep Learning TomographyAmir Adler
 
Zebra - TRIAD-ES Joint Presentation
Zebra - TRIAD-ES Joint PresentationZebra - TRIAD-ES Joint Presentation
Zebra - TRIAD-ES Joint PresentationZEBRA Environmental
 
Real-Time Analysis of Streaming Synchotron Data: SCinet SC19 Technology Chall...
Real-Time Analysis of Streaming Synchotron Data: SCinet SC19 Technology Chall...Real-Time Analysis of Streaming Synchotron Data: SCinet SC19 Technology Chall...
Real-Time Analysis of Streaming Synchotron Data: SCinet SC19 Technology Chall...Globus
 
[Seminar] hyunwook 0624
[Seminar] hyunwook 0624[Seminar] hyunwook 0624
[Seminar] hyunwook 0624ivaderivader
 
Miccai2018 paperlist
Miccai2018 paperlistMiccai2018 paperlist
Miccai2018 paperlistKevin Zhou
 
Improving Physical Parametrizations in Climate Models using Machine Learning
Improving Physical Parametrizations in Climate Models using Machine LearningImproving Physical Parametrizations in Climate Models using Machine Learning
Improving Physical Parametrizations in Climate Models using Machine LearningNoah Brenowitz
 
INTERACTIVE ANALYTICAL TOOL FOR CORNEAL CONFOCAL IMAGING
INTERACTIVE ANALYTICAL TOOL FOR CORNEAL CONFOCAL IMAGINGINTERACTIVE ANALYTICAL TOOL FOR CORNEAL CONFOCAL IMAGING
INTERACTIVE ANALYTICAL TOOL FOR CORNEAL CONFOCAL IMAGINGMadhavi Tippani
 
Automated Motion Detection from space in sea surveillance
Automated Motion Detection from space in sea surveillanceAutomated Motion Detection from space in sea surveillance
Automated Motion Detection from space in sea surveillanceLiza Charalambous
 
HACC: Fitting the Universe Inside a Supercomputer
HACC: Fitting the Universe Inside a SupercomputerHACC: Fitting the Universe Inside a Supercomputer
HACC: Fitting the Universe Inside a Supercomputerinside-BigData.com
 
Chapter 6 image quality in ct
Chapter 6 image quality in ct Chapter 6 image quality in ct
Chapter 6 image quality in ct Muntaser S.Ahmad
 
Temporal Superpixels Based on Proximity-Weighted Patch Matching
Temporal Superpixels Based on Proximity-Weighted Patch MatchingTemporal Superpixels Based on Proximity-Weighted Patch Matching
Temporal Superpixels Based on Proximity-Weighted Patch MatchingNAVER Engineering
 
Rrecent advances in linear accelerators [MR linac]
Rrecent advances in linear accelerators [MR linac]Rrecent advances in linear accelerators [MR linac]
Rrecent advances in linear accelerators [MR linac]Upasna Saxena
 
Distributed Data Processing using Spark by Panos Labropoulos_and Sarod Yataw...
Distributed Data Processing using Spark by  Panos Labropoulos_and Sarod Yataw...Distributed Data Processing using Spark by  Panos Labropoulos_and Sarod Yataw...
Distributed Data Processing using Spark by Panos Labropoulos_and Sarod Yataw...Spark Summit
 

Ähnlich wie Two stage deep learning for accelerated 3 d time-of-flight mra without matched training data (20)

CycleMorph: Cycle consistent unsupervised deformable image registration
CycleMorph: Cycle consistent unsupervised deformable image registrationCycleMorph: Cycle consistent unsupervised deformable image registration
CycleMorph: Cycle consistent unsupervised deformable image registration
 
Deep Learning Tomography
Deep Learning TomographyDeep Learning Tomography
Deep Learning Tomography
 
Zebra - TRIAD-ES Joint Presentation
Zebra - TRIAD-ES Joint PresentationZebra - TRIAD-ES Joint Presentation
Zebra - TRIAD-ES Joint Presentation
 
Real-Time Analysis of Streaming Synchotron Data: SCinet SC19 Technology Chall...
Real-Time Analysis of Streaming Synchotron Data: SCinet SC19 Technology Chall...Real-Time Analysis of Streaming Synchotron Data: SCinet SC19 Technology Chall...
Real-Time Analysis of Streaming Synchotron Data: SCinet SC19 Technology Chall...
 
Inverse problems in medical imaging
Inverse problems in medical imagingInverse problems in medical imaging
Inverse problems in medical imaging
 
Kintinuous review
Kintinuous reviewKintinuous review
Kintinuous review
 
[Seminar] hyunwook 0624
[Seminar] hyunwook 0624[Seminar] hyunwook 0624
[Seminar] hyunwook 0624
 
Cbct
CbctCbct
Cbct
 
Miccai2018 paperlist
Miccai2018 paperlistMiccai2018 paperlist
Miccai2018 paperlist
 
Improving Physical Parametrizations in Climate Models using Machine Learning
Improving Physical Parametrizations in Climate Models using Machine LearningImproving Physical Parametrizations in Climate Models using Machine Learning
Improving Physical Parametrizations in Climate Models using Machine Learning
 
INTERACTIVE ANALYTICAL TOOL FOR CORNEAL CONFOCAL IMAGING
INTERACTIVE ANALYTICAL TOOL FOR CORNEAL CONFOCAL IMAGINGINTERACTIVE ANALYTICAL TOOL FOR CORNEAL CONFOCAL IMAGING
INTERACTIVE ANALYTICAL TOOL FOR CORNEAL CONFOCAL IMAGING
 
CLIM: Transition Workshop - Optimization Methods in Remote Sensing - Jessica...
CLIM: Transition Workshop - Optimization Methods in Remote Sensing  - Jessica...CLIM: Transition Workshop - Optimization Methods in Remote Sensing  - Jessica...
CLIM: Transition Workshop - Optimization Methods in Remote Sensing - Jessica...
 
Automated Motion Detection from space in sea surveillance
Automated Motion Detection from space in sea surveillanceAutomated Motion Detection from space in sea surveillance
Automated Motion Detection from space in sea surveillance
 
HACC: Fitting the Universe Inside a Supercomputer
HACC: Fitting the Universe Inside a SupercomputerHACC: Fitting the Universe Inside a Supercomputer
HACC: Fitting the Universe Inside a Supercomputer
 
Chapter 6 image quality in ct
Chapter 6 image quality in ct Chapter 6 image quality in ct
Chapter 6 image quality in ct
 
Temporal Superpixels Based on Proximity-Weighted Patch Matching
Temporal Superpixels Based on Proximity-Weighted Patch MatchingTemporal Superpixels Based on Proximity-Weighted Patch Matching
Temporal Superpixels Based on Proximity-Weighted Patch Matching
 
Rrecent advances in linear accelerators [MR linac]
Rrecent advances in linear accelerators [MR linac]Rrecent advances in linear accelerators [MR linac]
Rrecent advances in linear accelerators [MR linac]
 
Distributed Data Processing using Spark by Panos Labropoulos_and Sarod Yataw...
Distributed Data Processing using Spark by  Panos Labropoulos_and Sarod Yataw...Distributed Data Processing using Spark by  Panos Labropoulos_and Sarod Yataw...
Distributed Data Processing using Spark by Panos Labropoulos_and Sarod Yataw...
 
CBCT.pptx
CBCT.pptxCBCT.pptx
CBCT.pptx
 
mr angiography.pptx
mr angiography.pptxmr angiography.pptx
mr angiography.pptx
 

Mehr von Chung Hyung Jin

DiffusionMBIR_presentation_slide.pptx
DiffusionMBIR_presentation_slide.pptxDiffusionMBIR_presentation_slide.pptx
DiffusionMBIR_presentation_slide.pptxChung Hyung Jin
 
BlindDPS_presentation_silde.pptx
BlindDPS_presentation_silde.pptxBlindDPS_presentation_silde.pptx
BlindDPS_presentation_silde.pptxChung Hyung Jin
 
diffusion_posterior_sampling_for_general_noisy_inverse_problems_slideshare.pdf
diffusion_posterior_sampling_for_general_noisy_inverse_problems_slideshare.pdfdiffusion_posterior_sampling_for_general_noisy_inverse_problems_slideshare.pdf
diffusion_posterior_sampling_for_general_noisy_inverse_problems_slideshare.pdfChung Hyung Jin
 
mr-denoising-and-super-resolution-using-regularized-reverse-diffusion.pptx
mr-denoising-and-super-resolution-using-regularized-reverse-diffusion.pptxmr-denoising-and-super-resolution-using-regularized-reverse-diffusion.pptx
mr-denoising-and-super-resolution-using-regularized-reverse-diffusion.pptxChung Hyung Jin
 
Low-dose sparse-view HAADF-STEM-EDX tomography of nanocrystals using unsuperv...
Low-dose sparse-view HAADF-STEM-EDX tomography of nanocrystals using unsuperv...Low-dose sparse-view HAADF-STEM-EDX tomography of nanocrystals using unsuperv...
Low-dose sparse-view HAADF-STEM-EDX tomography of nanocrystals using unsuperv...Chung Hyung Jin
 
Score-based diffusion models for accelerated MRI.pptx
Score-based diffusion models for accelerated MRI.pptxScore-based diffusion models for accelerated MRI.pptx
Score-based diffusion models for accelerated MRI.pptxChung Hyung Jin
 
Come-Closer-Diffuse-Faster Accelerating Conditional Diffusion Models for Inve...
Come-Closer-Diffuse-Faster Accelerating Conditional Diffusion Models for Inve...Come-Closer-Diffuse-Faster Accelerating Conditional Diffusion Models for Inve...
Come-Closer-Diffuse-Faster Accelerating Conditional Diffusion Models for Inve...Chung Hyung Jin
 
Missing cone artifact removal in odt using unsupervised deep learning in the ...
Missing cone artifact removal in odt using unsupervised deep learning in the ...Missing cone artifact removal in odt using unsupervised deep learning in the ...
Missing cone artifact removal in odt using unsupervised deep learning in the ...Chung Hyung Jin
 
A deep learning model for diagnosing gastric mucosal lesions using endoscopic...
A deep learning model for diagnosing gastric mucosal lesions using endoscopic...A deep learning model for diagnosing gastric mucosal lesions using endoscopic...
A deep learning model for diagnosing gastric mucosal lesions using endoscopic...Chung Hyung Jin
 
Feature disentanglement in generating a three dimensional structure from a tw...
Feature disentanglement in generating a three dimensional structure from a tw...Feature disentanglement in generating a three dimensional structure from a tw...
Feature disentanglement in generating a three dimensional structure from a tw...Chung Hyung Jin
 
Unsupervised deep learning methods for biological image reconstruction and en...
Unsupervised deep learning methods for biological image reconstruction and en...Unsupervised deep learning methods for biological image reconstruction and en...
Unsupervised deep learning methods for biological image reconstruction and en...Chung Hyung Jin
 

Mehr von Chung Hyung Jin (11)

DiffusionMBIR_presentation_slide.pptx
DiffusionMBIR_presentation_slide.pptxDiffusionMBIR_presentation_slide.pptx
DiffusionMBIR_presentation_slide.pptx
 
BlindDPS_presentation_silde.pptx
BlindDPS_presentation_silde.pptxBlindDPS_presentation_silde.pptx
BlindDPS_presentation_silde.pptx
 
diffusion_posterior_sampling_for_general_noisy_inverse_problems_slideshare.pdf
diffusion_posterior_sampling_for_general_noisy_inverse_problems_slideshare.pdfdiffusion_posterior_sampling_for_general_noisy_inverse_problems_slideshare.pdf
diffusion_posterior_sampling_for_general_noisy_inverse_problems_slideshare.pdf
 
mr-denoising-and-super-resolution-using-regularized-reverse-diffusion.pptx
mr-denoising-and-super-resolution-using-regularized-reverse-diffusion.pptxmr-denoising-and-super-resolution-using-regularized-reverse-diffusion.pptx
mr-denoising-and-super-resolution-using-regularized-reverse-diffusion.pptx
 
Low-dose sparse-view HAADF-STEM-EDX tomography of nanocrystals using unsuperv...
Low-dose sparse-view HAADF-STEM-EDX tomography of nanocrystals using unsuperv...Low-dose sparse-view HAADF-STEM-EDX tomography of nanocrystals using unsuperv...
Low-dose sparse-view HAADF-STEM-EDX tomography of nanocrystals using unsuperv...
 
Score-based diffusion models for accelerated MRI.pptx
Score-based diffusion models for accelerated MRI.pptxScore-based diffusion models for accelerated MRI.pptx
Score-based diffusion models for accelerated MRI.pptx
 
Come-Closer-Diffuse-Faster Accelerating Conditional Diffusion Models for Inve...
Come-Closer-Diffuse-Faster Accelerating Conditional Diffusion Models for Inve...Come-Closer-Diffuse-Faster Accelerating Conditional Diffusion Models for Inve...
Come-Closer-Diffuse-Faster Accelerating Conditional Diffusion Models for Inve...
 
Missing cone artifact removal in odt using unsupervised deep learning in the ...
Missing cone artifact removal in odt using unsupervised deep learning in the ...Missing cone artifact removal in odt using unsupervised deep learning in the ...
Missing cone artifact removal in odt using unsupervised deep learning in the ...
 
A deep learning model for diagnosing gastric mucosal lesions using endoscopic...
A deep learning model for diagnosing gastric mucosal lesions using endoscopic...A deep learning model for diagnosing gastric mucosal lesions using endoscopic...
A deep learning model for diagnosing gastric mucosal lesions using endoscopic...
 
Feature disentanglement in generating a three dimensional structure from a tw...
Feature disentanglement in generating a three dimensional structure from a tw...Feature disentanglement in generating a three dimensional structure from a tw...
Feature disentanglement in generating a three dimensional structure from a tw...
 
Unsupervised deep learning methods for biological image reconstruction and en...
Unsupervised deep learning methods for biological image reconstruction and en...Unsupervised deep learning methods for biological image reconstruction and en...
Unsupervised deep learning methods for biological image reconstruction and en...
 

Kürzlich hochgeladen

CCS355 Neural Networks & Deep Learning Unit 1 PDF notes with Question bank .pdf
CCS355 Neural Networks & Deep Learning Unit 1 PDF notes with Question bank .pdfCCS355 Neural Networks & Deep Learning Unit 1 PDF notes with Question bank .pdf
CCS355 Neural Networks & Deep Learning Unit 1 PDF notes with Question bank .pdfAsst.prof M.Gokilavani
 
Call Girls Delhi {Jodhpur} 9711199012 high profile service
Call Girls Delhi {Jodhpur} 9711199012 high profile serviceCall Girls Delhi {Jodhpur} 9711199012 high profile service
Call Girls Delhi {Jodhpur} 9711199012 high profile servicerehmti665
 
INFLUENCE OF NANOSILICA ON THE PROPERTIES OF CONCRETE
INFLUENCE OF NANOSILICA ON THE PROPERTIES OF CONCRETEINFLUENCE OF NANOSILICA ON THE PROPERTIES OF CONCRETE
INFLUENCE OF NANOSILICA ON THE PROPERTIES OF CONCRETEroselinkalist12
 
Electronically Controlled suspensions system .pdf
Electronically Controlled suspensions system .pdfElectronically Controlled suspensions system .pdf
Electronically Controlled suspensions system .pdfme23b1001
 
An experimental study in using natural admixture as an alternative for chemic...
An experimental study in using natural admixture as an alternative for chemic...An experimental study in using natural admixture as an alternative for chemic...
An experimental study in using natural admixture as an alternative for chemic...Chandu841456
 
CCS355 Neural Network & Deep Learning Unit II Notes with Question bank .pdf
CCS355 Neural Network & Deep Learning Unit II Notes with Question bank .pdfCCS355 Neural Network & Deep Learning Unit II Notes with Question bank .pdf
CCS355 Neural Network & Deep Learning Unit II Notes with Question bank .pdfAsst.prof M.Gokilavani
 
computer application and construction management
computer application and construction managementcomputer application and construction management
computer application and construction managementMariconPadriquez1
 
Concrete Mix Design - IS 10262-2019 - .pptx
Concrete Mix Design - IS 10262-2019 - .pptxConcrete Mix Design - IS 10262-2019 - .pptx
Concrete Mix Design - IS 10262-2019 - .pptxKartikeyaDwivedi3
 
Introduction-To-Agricultural-Surveillance-Rover.pptx
Introduction-To-Agricultural-Surveillance-Rover.pptxIntroduction-To-Agricultural-Surveillance-Rover.pptx
Introduction-To-Agricultural-Surveillance-Rover.pptxk795866
 
complete construction, environmental and economics information of biomass com...
complete construction, environmental and economics information of biomass com...complete construction, environmental and economics information of biomass com...
complete construction, environmental and economics information of biomass com...asadnawaz62
 
Sachpazis Costas: Geotechnical Engineering: A student's Perspective Introduction
Sachpazis Costas: Geotechnical Engineering: A student's Perspective IntroductionSachpazis Costas: Geotechnical Engineering: A student's Perspective Introduction
Sachpazis Costas: Geotechnical Engineering: A student's Perspective IntroductionDr.Costas Sachpazis
 
Piping Basic stress analysis by engineering
Piping Basic stress analysis by engineeringPiping Basic stress analysis by engineering
Piping Basic stress analysis by engineeringJuanCarlosMorales19600
 
Oxy acetylene welding presentation note.
Oxy acetylene welding presentation note.Oxy acetylene welding presentation note.
Oxy acetylene welding presentation note.eptoze12
 
Application of Residue Theorem to evaluate real integrations.pptx
Application of Residue Theorem to evaluate real integrations.pptxApplication of Residue Theorem to evaluate real integrations.pptx
Application of Residue Theorem to evaluate real integrations.pptx959SahilShah
 
TechTAC® CFD Report Summary: A Comparison of Two Types of Tubing Anchor Catchers
TechTAC® CFD Report Summary: A Comparison of Two Types of Tubing Anchor CatchersTechTAC® CFD Report Summary: A Comparison of Two Types of Tubing Anchor Catchers
TechTAC® CFD Report Summary: A Comparison of Two Types of Tubing Anchor Catcherssdickerson1
 
Class 1 | NFPA 72 | Overview Fire Alarm System
Class 1 | NFPA 72 | Overview Fire Alarm SystemClass 1 | NFPA 72 | Overview Fire Alarm System
Class 1 | NFPA 72 | Overview Fire Alarm Systemirfanmechengr
 
Past, Present and Future of Generative AI
Past, Present and Future of Generative AIPast, Present and Future of Generative AI
Past, Present and Future of Generative AIabhishek36461
 
Vishratwadi & Ghorpadi Bridge Tender documents
Vishratwadi & Ghorpadi Bridge Tender documentsVishratwadi & Ghorpadi Bridge Tender documents
Vishratwadi & Ghorpadi Bridge Tender documentsSachinPawar510423
 
lifi-technology with integration of IOT.pptx
lifi-technology with integration of IOT.pptxlifi-technology with integration of IOT.pptx
lifi-technology with integration of IOT.pptxsomshekarkn64
 

Kürzlich hochgeladen (20)

CCS355 Neural Networks & Deep Learning Unit 1 PDF notes with Question bank .pdf
CCS355 Neural Networks & Deep Learning Unit 1 PDF notes with Question bank .pdfCCS355 Neural Networks & Deep Learning Unit 1 PDF notes with Question bank .pdf
CCS355 Neural Networks & Deep Learning Unit 1 PDF notes with Question bank .pdf
 
Call Girls Delhi {Jodhpur} 9711199012 high profile service
Call Girls Delhi {Jodhpur} 9711199012 high profile serviceCall Girls Delhi {Jodhpur} 9711199012 high profile service
Call Girls Delhi {Jodhpur} 9711199012 high profile service
 
INFLUENCE OF NANOSILICA ON THE PROPERTIES OF CONCRETE
INFLUENCE OF NANOSILICA ON THE PROPERTIES OF CONCRETEINFLUENCE OF NANOSILICA ON THE PROPERTIES OF CONCRETE
INFLUENCE OF NANOSILICA ON THE PROPERTIES OF CONCRETE
 
Electronically Controlled suspensions system .pdf
Electronically Controlled suspensions system .pdfElectronically Controlled suspensions system .pdf
Electronically Controlled suspensions system .pdf
 
An experimental study in using natural admixture as an alternative for chemic...
An experimental study in using natural admixture as an alternative for chemic...An experimental study in using natural admixture as an alternative for chemic...
An experimental study in using natural admixture as an alternative for chemic...
 
CCS355 Neural Network & Deep Learning Unit II Notes with Question bank .pdf
CCS355 Neural Network & Deep Learning Unit II Notes with Question bank .pdfCCS355 Neural Network & Deep Learning Unit II Notes with Question bank .pdf
CCS355 Neural Network & Deep Learning Unit II Notes with Question bank .pdf
 
computer application and construction management
computer application and construction managementcomputer application and construction management
computer application and construction management
 
Concrete Mix Design - IS 10262-2019 - .pptx
Concrete Mix Design - IS 10262-2019 - .pptxConcrete Mix Design - IS 10262-2019 - .pptx
Concrete Mix Design - IS 10262-2019 - .pptx
 
Introduction-To-Agricultural-Surveillance-Rover.pptx
Introduction-To-Agricultural-Surveillance-Rover.pptxIntroduction-To-Agricultural-Surveillance-Rover.pptx
Introduction-To-Agricultural-Surveillance-Rover.pptx
 
complete construction, environmental and economics information of biomass com...
complete construction, environmental and economics information of biomass com...complete construction, environmental and economics information of biomass com...
complete construction, environmental and economics information of biomass com...
 
Exploring_Network_Security_with_JA3_by_Rakesh Seal.pptx
Exploring_Network_Security_with_JA3_by_Rakesh Seal.pptxExploring_Network_Security_with_JA3_by_Rakesh Seal.pptx
Exploring_Network_Security_with_JA3_by_Rakesh Seal.pptx
 
Sachpazis Costas: Geotechnical Engineering: A student's Perspective Introduction
Sachpazis Costas: Geotechnical Engineering: A student's Perspective IntroductionSachpazis Costas: Geotechnical Engineering: A student's Perspective Introduction
Sachpazis Costas: Geotechnical Engineering: A student's Perspective Introduction
 
Piping Basic stress analysis by engineering
Piping Basic stress analysis by engineeringPiping Basic stress analysis by engineering
Piping Basic stress analysis by engineering
 
Oxy acetylene welding presentation note.
Oxy acetylene welding presentation note.Oxy acetylene welding presentation note.
Oxy acetylene welding presentation note.
 
Application of Residue Theorem to evaluate real integrations.pptx
Application of Residue Theorem to evaluate real integrations.pptxApplication of Residue Theorem to evaluate real integrations.pptx
Application of Residue Theorem to evaluate real integrations.pptx
 
TechTAC® CFD Report Summary: A Comparison of Two Types of Tubing Anchor Catchers
TechTAC® CFD Report Summary: A Comparison of Two Types of Tubing Anchor CatchersTechTAC® CFD Report Summary: A Comparison of Two Types of Tubing Anchor Catchers
TechTAC® CFD Report Summary: A Comparison of Two Types of Tubing Anchor Catchers
 
Class 1 | NFPA 72 | Overview Fire Alarm System
Class 1 | NFPA 72 | Overview Fire Alarm SystemClass 1 | NFPA 72 | Overview Fire Alarm System
Class 1 | NFPA 72 | Overview Fire Alarm System
 
Past, Present and Future of Generative AI
Past, Present and Future of Generative AIPast, Present and Future of Generative AI
Past, Present and Future of Generative AI
 
Vishratwadi & Ghorpadi Bridge Tender documents
Vishratwadi & Ghorpadi Bridge Tender documentsVishratwadi & Ghorpadi Bridge Tender documents
Vishratwadi & Ghorpadi Bridge Tender documents
 
lifi-technology with integration of IOT.pptx
lifi-technology with integration of IOT.pptxlifi-technology with integration of IOT.pptx
lifi-technology with integration of IOT.pptx
 

Two stage deep learning for accelerated 3 d time-of-flight mra without matched training data

Hinweis der Redaktion

  1. 안녕하세요. 카이스트의 예종철 교수입니다. 2021년 한국뇌공학회 심포지움에 저를 초청해주셔서 대단히 감사합니다. 오늘 발표할 내용은 적대적 생성모델로 알려져 있는 GAN의 원리와 그를 이용한 비지도 학습 기법이 어떻게 의료영상 복원에 이용되는지에 대하여 저희 연구실에서 진행하는 연구를 중심으로 말씀드리겠습니다.
  2. 저희랩에서 개발한 이러한 general한 새로운 형태의 cycleGAN의 구조를 좀더 다양한 의료영상 복원 문제에 적용한것을 지금부터 보여드리겠습니다.
  3. 저희랩에서 개발한 이러한 general한 새로운 형태의 cycleGAN의 구조를 좀더 다양한 의료영상 복원 문제에 적용한것을 지금부터 보여드리겠습니다.
  4. 이러한 GAN이 어떻게 사실적인 얼굴을 만드냐고 물어 볼때 흔히 보여주는 비유의 그림입니다. 즉, 위지 지폐범이 진짜 돈과 구별이 안되는 돈을 만들기 위해 열심히 노력을 하면 경찰은 더 정확하게 구별하는 방식을 만들어내기 위해 연구를 합니다. 그러면 위조지폐범은 더욱 정밀은 위조 지폐를 만들고요. 이런것이 반복되다 보면 진짜 구별이 어려운 위조 지폐가 나타나는것이지요.
  5. 앞에서 말씀드린 위조지폐와 경찰의 비유는 실제 GAN을 설명하는 수식으로 확인할수 있는데요, 여기서 보이는 phi는 경찰에 해당하는 discriminator이고 G는 위조지폐범에 해당하는 generator입니다. 이때 discriminator의 목적은 두개의 거리를 최대한 구별이 가도록 벌리는 것이 목적이고요, generator 는 두개의 거리를 가깝게 해서 구별이 어렵게 만들려고 하는것이지요.
  6. 더욱이나 수학적으로 놀라운 사실은 이러한 GAN의 원리가 다음의 그림에 보이는 것과 같은 확률 분포간의 거리를 최소하는 문제의 다른 표현이라는 사실입니다. 즉 y 공간에 있는 가우신안 노이즈를 x 공간에 있는 점으로 표시되는 얼굴들로 변화시키는 수학적인 변환인 G를 찾는데 있어서 변환후이 확률 분포과 원래 얼굴들이 분포되어 있는 확률공간과의 거리를 최소하하는 문제를 풀면 이것이 GAN으로 나타날수 있다는 것이지요.
  7. 하지만 얼룩말을 말로 바꾼다던지 사진을 모네, 반고호, 세잔, 유키코의 그림으로 바꾸는 문제의 경우는 매칭이 되는 페어가 없는 비지도 학습의 문제인데 이 문제를 해결하기 위해 cycleGAN이라는 방법이 제안되었습니다. 이방법의 기본적인 아이디어는 얼룩말을 말로 만드는 generator와 말을 얼룩말로 만드는 generator를 동시에 학습을 하고 인푼 영상이 두개의 제너레이터를 통과를 하면 원상 복귀되는 조건을 넘으로서 pair가 필요한 지도학습의 단점을 극복한것이지요.
  8. 저희랩에서 개발한 이러한 general한 새로운 형태의 cycleGAN의 구조를 좀더 다양한 의료영상 복원 문제에 적용한것을 지금부터 보여드리겠습니다.
  9. 최근의 본 연구실의 연구에서도 이러한 cycleGAN의 수학적인 구조가 앞에 설명한 GAN의 수학적인 구조에서 확장될수 있다는것을 밝혔습니다. 즉, Y라는 측정 값의 모여있는 공간과 X라는 영상이 있는 도메인이 있고 각각의 도에인데 매칭이 되지 않는 데이타가 있을때, 하나의 저너레이터는 측정치에셔 영상으로 가는 변환을 학습하고 다른 제너레이터는 영상에서 측정치로 가는 변환을 학습하는데, 아깐 말한것 같이 변환된 확률과 실제 확룰분포와의 거리를 동시에 최소화 하면 cycleGAN이 나오단느 것입니다. 더욱 놀라운 사실은 영상에서 측정치로 가는 함수는 많은 경우 알려져 있는데 이 경우 훨씬 단순한 cycleGAN구조가 나온다는 것입니다.
  10. 최근의 본 연구실의 연구에서도 이러한 cycleGAN의 수학적인 구조가 앞에 설명한 GAN의 수학적인 구조에서 확장될수 있다는것을 밝혔습니다. 즉, Y라는 측정 값의 모여있는 공간과 X라는 영상이 있는 도메인이 있고 각각의 도에인데 매칭이 되지 않는 데이타가 있을때, 하나의 저너레이터는 측정치에셔 영상으로 가는 변환을 학습하고 다른 제너레이터는 영상에서 측정치로 가는 변환을 학습하는데, 아깐 말한것 같이 변환된 확률과 실제 확룰분포와의 거리를 동시에 최소화 하면 cycleGAN이 나오단느 것입니다. 더욱 놀라운 사실은 영상에서 측정치로 가는 함수는 많은 경우 알려져 있는데 이 경우 훨씬 단순한 cycleGAN구조가 나온다는 것입니다.
  11. 최근의 본 연구실의 연구에서도 이러한 cycleGAN의 수학적인 구조가 앞에 설명한 GAN의 수학적인 구조에서 확장될수 있다는것을 밝혔습니다. 즉, Y라는 측정 값의 모여있는 공간과 X라는 영상이 있는 도메인이 있고 각각의 도에인데 매칭이 되지 않는 데이타가 있을때, 하나의 저너레이터는 측정치에셔 영상으로 가는 변환을 학습하고 다른 제너레이터는 영상에서 측정치로 가는 변환을 학습하는데, 아깐 말한것 같이 변환된 확률과 실제 확룰분포와의 거리를 동시에 최소화 하면 cycleGAN이 나오단느 것입니다. 더욱 놀라운 사실은 영상에서 측정치로 가는 함수는 많은 경우 알려져 있는데 이 경우 훨씬 단순한 cycleGAN구조가 나온다는 것입니다.
  12. 최근의 본 연구실의 연구에서도 이러한 cycleGAN의 수학적인 구조가 앞에 설명한 GAN의 수학적인 구조에서 확장될수 있다는것을 밝혔습니다. 즉, Y라는 측정 값의 모여있는 공간과 X라는 영상이 있는 도메인이 있고 각각의 도에인데 매칭이 되지 않는 데이타가 있을때, 하나의 저너레이터는 측정치에셔 영상으로 가는 변환을 학습하고 다른 제너레이터는 영상에서 측정치로 가는 변환을 학습하는데, 아깐 말한것 같이 변환된 확률과 실제 확룰분포와의 거리를 동시에 최소화 하면 cycleGAN이 나오단느 것입니다. 더욱 놀라운 사실은 영상에서 측정치로 가는 함수는 많은 경우 알려져 있는데 이 경우 훨씬 단순한 cycleGAN구조가 나온다는 것입니다.
  13. 최근의 본 연구실의 연구에서도 이러한 cycleGAN의 수학적인 구조가 앞에 설명한 GAN의 수학적인 구조에서 확장될수 있다는것을 밝혔습니다. 즉, Y라는 측정 값의 모여있는 공간과 X라는 영상이 있는 도메인이 있고 각각의 도에인데 매칭이 되지 않는 데이타가 있을때, 하나의 저너레이터는 측정치에셔 영상으로 가는 변환을 학습하고 다른 제너레이터는 영상에서 측정치로 가는 변환을 학습하는데, 아깐 말한것 같이 변환된 확률과 실제 확룰분포와의 거리를 동시에 최소화 하면 cycleGAN이 나오단느 것입니다. 더욱 놀라운 사실은 영상에서 측정치로 가는 함수는 많은 경우 알려져 있는데 이 경우 훨씬 단순한 cycleGAN구조가 나온다는 것입니다.
  14. 최근의 본 연구실의 연구에서도 이러한 cycleGAN의 수학적인 구조가 앞에 설명한 GAN의 수학적인 구조에서 확장될수 있다는것을 밝혔습니다. 즉, Y라는 측정 값의 모여있는 공간과 X라는 영상이 있는 도메인이 있고 각각의 도에인데 매칭이 되지 않는 데이타가 있을때, 하나의 저너레이터는 측정치에셔 영상으로 가는 변환을 학습하고 다른 제너레이터는 영상에서 측정치로 가는 변환을 학습하는데, 아깐 말한것 같이 변환된 확률과 실제 확룰분포와의 거리를 동시에 최소화 하면 cycleGAN이 나오단느 것입니다. 더욱 놀라운 사실은 영상에서 측정치로 가는 함수는 많은 경우 알려져 있는데 이 경우 훨씬 단순한 cycleGAN구조가 나온다는 것입니다.
  15. 최근의 본 연구실의 연구에서도 이러한 cycleGAN의 수학적인 구조가 앞에 설명한 GAN의 수학적인 구조에서 확장될수 있다는것을 밝혔습니다. 즉, Y라는 측정 값의 모여있는 공간과 X라는 영상이 있는 도메인이 있고 각각의 도에인데 매칭이 되지 않는 데이타가 있을때, 하나의 저너레이터는 측정치에셔 영상으로 가는 변환을 학습하고 다른 제너레이터는 영상에서 측정치로 가는 변환을 학습하는데, 아깐 말한것 같이 변환된 확률과 실제 확룰분포와의 거리를 동시에 최소화 하면 cycleGAN이 나오단느 것입니다. 더욱 놀라운 사실은 영상에서 측정치로 가는 함수는 많은 경우 알려져 있는데 이 경우 훨씬 단순한 cycleGAN구조가 나온다는 것입니다.
  16. 저희랩에서 개발한 이러한 general한 새로운 형태의 cycleGAN의 구조를 좀더 다양한 의료영상 복원 문제에 적용한것을 지금부터 보여드리겠습니다.
  17. 최근의 본 연구실의 연구에서도 이러한 cycleGAN의 수학적인 구조가 앞에 설명한 GAN의 수학적인 구조에서 확장될수 있다는것을 밝혔습니다. 즉, Y라는 측정 값의 모여있는 공간과 X라는 영상이 있는 도메인이 있고 각각의 도에인데 매칭이 되지 않는 데이타가 있을때, 하나의 저너레이터는 측정치에셔 영상으로 가는 변환을 학습하고 다른 제너레이터는 영상에서 측정치로 가는 변환을 학습하는데, 아깐 말한것 같이 변환된 확률과 실제 확룰분포와의 거리를 동시에 최소화 하면 cycleGAN이 나오단느 것입니다. 더욱 놀라운 사실은 영상에서 측정치로 가는 함수는 많은 경우 알려져 있는데 이 경우 훨씬 단순한 cycleGAN구조가 나온다는 것입니다.
  18. 최근의 본 연구실의 연구에서도 이러한 cycleGAN의 수학적인 구조가 앞에 설명한 GAN의 수학적인 구조에서 확장될수 있다는것을 밝혔습니다. 즉, Y라는 측정 값의 모여있는 공간과 X라는 영상이 있는 도메인이 있고 각각의 도에인데 매칭이 되지 않는 데이타가 있을때, 하나의 저너레이터는 측정치에셔 영상으로 가는 변환을 학습하고 다른 제너레이터는 영상에서 측정치로 가는 변환을 학습하는데, 아깐 말한것 같이 변환된 확률과 실제 확룰분포와의 거리를 동시에 최소화 하면 cycleGAN이 나오단느 것입니다. 더욱 놀라운 사실은 영상에서 측정치로 가는 함수는 많은 경우 알려져 있는데 이 경우 훨씬 단순한 cycleGAN구조가 나온다는 것입니다.
  19. 최근의 본 연구실의 연구에서도 이러한 cycleGAN의 수학적인 구조가 앞에 설명한 GAN의 수학적인 구조에서 확장될수 있다는것을 밝혔습니다. 즉, Y라는 측정 값의 모여있는 공간과 X라는 영상이 있는 도메인이 있고 각각의 도에인데 매칭이 되지 않는 데이타가 있을때, 하나의 저너레이터는 측정치에셔 영상으로 가는 변환을 학습하고 다른 제너레이터는 영상에서 측정치로 가는 변환을 학습하는데, 아깐 말한것 같이 변환된 확률과 실제 확룰분포와의 거리를 동시에 최소화 하면 cycleGAN이 나오단느 것입니다. 더욱 놀라운 사실은 영상에서 측정치로 가는 함수는 많은 경우 알려져 있는데 이 경우 훨씬 단순한 cycleGAN구조가 나온다는 것입니다.
  20. 최근의 본 연구실의 연구에서도 이러한 cycleGAN의 수학적인 구조가 앞에 설명한 GAN의 수학적인 구조에서 확장될수 있다는것을 밝혔습니다. 즉, Y라는 측정 값의 모여있는 공간과 X라는 영상이 있는 도메인이 있고 각각의 도에인데 매칭이 되지 않는 데이타가 있을때, 하나의 저너레이터는 측정치에셔 영상으로 가는 변환을 학습하고 다른 제너레이터는 영상에서 측정치로 가는 변환을 학습하는데, 아깐 말한것 같이 변환된 확률과 실제 확룰분포와의 거리를 동시에 최소화 하면 cycleGAN이 나오단느 것입니다. 더욱 놀라운 사실은 영상에서 측정치로 가는 함수는 많은 경우 알려져 있는데 이 경우 훨씬 단순한 cycleGAN구조가 나온다는 것입니다.
  21. 지금까지 본 발표에서는 GAN이 의료영상 복원에서 비지도 학습기법으로 점점도 중요한 주제가 되고 있다는것을 보였고, 특히 collaGAN은 MR contrast 의 연구에 사용이 가능하다는것을 보였습니다. 경청해주셔서 감사합닏.