SlideShare a Scribd company logo
1 of 15
A Probabilistic U-Net for Segmentation of
Ambiguous Images
Hwang seung hyun
Yonsei University Severance Hospital CCIDS
DeepMind, Division of Medical Image Computing, German
Cancer Research Center, Germany | NIPS 2018
2020.04.19
Introduction Related Work Methods and
Experiments
01 02 03
Conclusion
04
Yonsei Unversity Severance Hospital CCIDS
Contents
Probabilistic Unet
Introduction – Limitations of prior methods
• There exist ambiguities in segmentation task, especially in medical imaging applications
• A lesion might be clearly visible, but ground truth labels can vary depending on
radiologists.
• Most existing segmentation algorithms either provide only consistent hypothesis of a
pixel-wise probability(e.g. “each pixel is 50% cat, 50% dog)
• Pixel wise probabilities ignores all co-variances between the pixels.
• Existing methods are Ensemble Unet, dropout Unet, M heads model, etc.
Introduction / Related Work / Methods and Experiments / Conclusion
Probabilistic Unet
Introduction – Probabilistic Unet Architecture
• Probabilistic Unet provides multiple segmentation hypotheses for ambiguous images.
• Combines conditional variational auto encoder(CVAE), and U-Net
• First extract latent space and encodes the possible segmentation variants
• Random sample from the space is injected into the Unet to produce segmentation map.
Introduction / Related Work / Methods and Experiments / Conclusion
Probabilistic Unet
Introduction – Contributions
• Provides consistent segmentation maps instead of pixel-wise probabilities,
providing joint likelihood of modes.
• Able to learn calibrated probabilities of segmentation modes.
• Can produce diverse outputs for single image
Introduction / Related Work / Methods and Experiments / Conclusion
Related Work
CVAE (Conditional Variational Auto Encoder)
Introduction / Related Work / Methods and Experiments / Conclusion
• Encoder를 통해 도출된 latent coding Z를 가우시
안 분포로 나타내기 위해 분산과 평균을 이용함
• Label 정보를 추가로 넣어준다
Related Work
U-Net
Introduction / Related Work / Methods and Experiments / Conclusion
• Encoding Phase
Methods and Experiments
Network Architecture
Introduction / Related Work / Methods and Experiments / Conclusion
• Sampling Process • Training Process
Methods and Experiments
Sampling Process
Introduction / Related Work / Methods and Experiments / Conclusion
• Prior Net (Unet’s encoding phase + global average
pooling) produces Latent Space
• Each position in this space encodes a
segmentation variant
• Broadcast the sample to feature map with the
same shape as the segmentation map, and
concatenate this map to the las activation map of
U-Net
* P : prior probability distribution
* fcomb = three subsequent 1x1 convolutions
* S: segmentation map corresponding to point z in latent space
Methods and Experiments
Training Process
Introduction / Related Work / Methods and Experiments / Conclusion
• Introduce Posterior Net that learns to recognize a
useful segmentation variant
• Posterior Net and Prior Net are updated through the
standard training procedure for CVAE, by minimizing
variational lower bound
(Kullback-Leibler divergence)
• Cross-entropy loss penalizes differences between S
and Y
• KL loss pulls the posterior distribution and prior
distribution towards each other
• Eventually covers the space of all useful segmentation
variants for input image
21
Methods and Experiments
Sampling Process
Introduction / Related Work / Methods and Experiments / Conclusion
Output Samples
Visualization of the Latent Space
Methods and Experiments
Introduction / Related Work / Methods and Experiments / Conclusion
Performance Measures
• Generalized Energy Distance Matrix
• Not only compare deterministic prediction, but also compares
distributions of segmentations
* d: distance measure
* Y, Y’ : Independent samples from the ground truth distribution
* S, S’: independent samples from the predicted distribution
* d(x,y) = 1 - IOU(x,y)
Methods and Experiments
Introduction / Related Work / Methods and Experiments / Conclusion
Results
Methods and Experiments
Introduction / Related Work / Methods and Experiments / Conclusion
Results
• Energy Distance decreases as more samples are drawn indicating an improved
matching of the GT distribution, as well as enhanced sample diversity.
Conclusion
Introduction / Related Work / Methods and Experiments / Conclusion
• Each sample produced by probabilistic Unet is consistent segmentation
result that closely match the multi-modal GT distributions
• Employed energy distance matrix measures whether the model’s
individual samples are both coherent, and whether they are produced
with expected frequencies.
• Can be used to assess annotations with model
• Probabilistic U Net can replace the currently applied deterministic U
Nets in large field of studies, especially in the medical domain
• Guide steps to resolve ambiguities

More Related Content

Similar to A Probabilistic U-Net for Segmentation of Ambiguous Images

End-to-End Object Detection with Transformers
End-to-End Object Detection with TransformersEnd-to-End Object Detection with Transformers
End-to-End Object Detection with TransformersSeunghyun Hwang
 
How useful is self-supervised pretraining for Visual tasks?
How useful is self-supervised pretraining for Visual tasks?How useful is self-supervised pretraining for Visual tasks?
How useful is self-supervised pretraining for Visual tasks?Seunghyun Hwang
 
DeepStrip: High Resolution Boundary Refinement
DeepStrip: High Resolution Boundary RefinementDeepStrip: High Resolution Boundary Refinement
DeepStrip: High Resolution Boundary RefinementSeunghyun Hwang
 
FickleNet: Weakly and Semi-supervised Semantic Image Segmentation using Stoch...
FickleNet: Weakly and Semi-supervised Semantic Image Segmentation using Stoch...FickleNet: Weakly and Semi-supervised Semantic Image Segmentation using Stoch...
FickleNet: Weakly and Semi-supervised Semantic Image Segmentation using Stoch...Seunghyun Hwang
 
Your Classifier is Secretly an Energy based model and you should treat it lik...
Your Classifier is Secretly an Energy based model and you should treat it lik...Your Classifier is Secretly an Energy based model and you should treat it lik...
Your Classifier is Secretly an Energy based model and you should treat it lik...Seunghyun Hwang
 
Deep Generative model-based quality control for cardiac MRI segmentation
Deep Generative model-based quality control for cardiac MRI segmentation Deep Generative model-based quality control for cardiac MRI segmentation
Deep Generative model-based quality control for cardiac MRI segmentation Seunghyun Hwang
 
Lec16: Medical Image Registration (Advanced): Deformable Registration
Lec16: Medical Image Registration (Advanced): Deformable RegistrationLec16: Medical Image Registration (Advanced): Deformable Registration
Lec16: Medical Image Registration (Advanced): Deformable RegistrationUlaş Bağcı
 
ResNeSt: Split-Attention Networks
ResNeSt: Split-Attention NetworksResNeSt: Split-Attention Networks
ResNeSt: Split-Attention NetworksSeunghyun Hwang
 
MEME – An Integrated Tool For Advanced Computational Experiments
MEME – An Integrated Tool For Advanced Computational ExperimentsMEME – An Integrated Tool For Advanced Computational Experiments
MEME – An Integrated Tool For Advanced Computational ExperimentsGIScRG
 
Iwsm2014 cosmic approximate sizing using a fuzzy logic approach (alain abran)
Iwsm2014   cosmic approximate sizing using a fuzzy logic approach (alain abran)Iwsm2014   cosmic approximate sizing using a fuzzy logic approach (alain abran)
Iwsm2014 cosmic approximate sizing using a fuzzy logic approach (alain abran)Nesma
 
Prototype-based classifiers and their applications in the life sciences
Prototype-based classifiers and their applications in the life sciencesPrototype-based classifiers and their applications in the life sciences
Prototype-based classifiers and their applications in the life sciencesUniversity of Groningen
 
Learning Sparse Networks using Targeted Dropout
Learning Sparse Networks using Targeted DropoutLearning Sparse Networks using Targeted Dropout
Learning Sparse Networks using Targeted DropoutSeunghyun Hwang
 
Bridging the Gap: Machine Learning for Ubiquitous Computing -- Evaluation
Bridging the Gap: Machine Learning for Ubiquitous Computing -- EvaluationBridging the Gap: Machine Learning for Ubiquitous Computing -- Evaluation
Bridging the Gap: Machine Learning for Ubiquitous Computing -- EvaluationThomas Ploetz
 
NEURAL Network Design Training
NEURAL Network Design  TrainingNEURAL Network Design  Training
NEURAL Network Design TrainingESCOM
 
Moving object detection in complex scene
Moving object detection in complex sceneMoving object detection in complex scene
Moving object detection in complex sceneKumar Mayank
 
Mx net image segmentation to predict and diagnose the cardiac diseases karp...
Mx net image segmentation to predict and diagnose the cardiac diseases   karp...Mx net image segmentation to predict and diagnose the cardiac diseases   karp...
Mx net image segmentation to predict and diagnose the cardiac diseases karp...KannanRamasamy25
 

Similar to A Probabilistic U-Net for Segmentation of Ambiguous Images (20)

End-to-End Object Detection with Transformers
End-to-End Object Detection with TransformersEnd-to-End Object Detection with Transformers
End-to-End Object Detection with Transformers
 
How useful is self-supervised pretraining for Visual tasks?
How useful is self-supervised pretraining for Visual tasks?How useful is self-supervised pretraining for Visual tasks?
How useful is self-supervised pretraining for Visual tasks?
 
DeepStrip: High Resolution Boundary Refinement
DeepStrip: High Resolution Boundary RefinementDeepStrip: High Resolution Boundary Refinement
DeepStrip: High Resolution Boundary Refinement
 
FickleNet: Weakly and Semi-supervised Semantic Image Segmentation using Stoch...
FickleNet: Weakly and Semi-supervised Semantic Image Segmentation using Stoch...FickleNet: Weakly and Semi-supervised Semantic Image Segmentation using Stoch...
FickleNet: Weakly and Semi-supervised Semantic Image Segmentation using Stoch...
 
Declarative data analysis
Declarative data analysisDeclarative data analysis
Declarative data analysis
 
Your Classifier is Secretly an Energy based model and you should treat it lik...
Your Classifier is Secretly an Energy based model and you should treat it lik...Your Classifier is Secretly an Energy based model and you should treat it lik...
Your Classifier is Secretly an Energy based model and you should treat it lik...
 
Deep Generative model-based quality control for cardiac MRI segmentation
Deep Generative model-based quality control for cardiac MRI segmentation Deep Generative model-based quality control for cardiac MRI segmentation
Deep Generative model-based quality control for cardiac MRI segmentation
 
Lec16: Medical Image Registration (Advanced): Deformable Registration
Lec16: Medical Image Registration (Advanced): Deformable RegistrationLec16: Medical Image Registration (Advanced): Deformable Registration
Lec16: Medical Image Registration (Advanced): Deformable Registration
 
Fa19_P1.pptx
Fa19_P1.pptxFa19_P1.pptx
Fa19_P1.pptx
 
ResNeSt: Split-Attention Networks
ResNeSt: Split-Attention NetworksResNeSt: Split-Attention Networks
ResNeSt: Split-Attention Networks
 
MEME – An Integrated Tool For Advanced Computational Experiments
MEME – An Integrated Tool For Advanced Computational ExperimentsMEME – An Integrated Tool For Advanced Computational Experiments
MEME – An Integrated Tool For Advanced Computational Experiments
 
Iwsm2014 cosmic approximate sizing using a fuzzy logic approach (alain abran)
Iwsm2014   cosmic approximate sizing using a fuzzy logic approach (alain abran)Iwsm2014   cosmic approximate sizing using a fuzzy logic approach (alain abran)
Iwsm2014 cosmic approximate sizing using a fuzzy logic approach (alain abran)
 
Prototype-based classifiers and their applications in the life sciences
Prototype-based classifiers and their applications in the life sciencesPrototype-based classifiers and their applications in the life sciences
Prototype-based classifiers and their applications in the life sciences
 
Learning Sparse Networks using Targeted Dropout
Learning Sparse Networks using Targeted DropoutLearning Sparse Networks using Targeted Dropout
Learning Sparse Networks using Targeted Dropout
 
01.pdf
01.pdf01.pdf
01.pdf
 
crossvalidation.pptx
crossvalidation.pptxcrossvalidation.pptx
crossvalidation.pptx
 
Bridging the Gap: Machine Learning for Ubiquitous Computing -- Evaluation
Bridging the Gap: Machine Learning for Ubiquitous Computing -- EvaluationBridging the Gap: Machine Learning for Ubiquitous Computing -- Evaluation
Bridging the Gap: Machine Learning for Ubiquitous Computing -- Evaluation
 
NEURAL Network Design Training
NEURAL Network Design  TrainingNEURAL Network Design  Training
NEURAL Network Design Training
 
Moving object detection in complex scene
Moving object detection in complex sceneMoving object detection in complex scene
Moving object detection in complex scene
 
Mx net image segmentation to predict and diagnose the cardiac diseases karp...
Mx net image segmentation to predict and diagnose the cardiac diseases   karp...Mx net image segmentation to predict and diagnose the cardiac diseases   karp...
Mx net image segmentation to predict and diagnose the cardiac diseases karp...
 

More from Seunghyun Hwang

An annotation sparsification strategy for 3D medical image segmentation via r...
An annotation sparsification strategy for 3D medical image segmentation via r...An annotation sparsification strategy for 3D medical image segmentation via r...
An annotation sparsification strategy for 3D medical image segmentation via r...Seunghyun Hwang
 
Do wide and deep networks learn the same things? Uncovering how neural networ...
Do wide and deep networks learn the same things? Uncovering how neural networ...Do wide and deep networks learn the same things? Uncovering how neural networ...
Do wide and deep networks learn the same things? Uncovering how neural networ...Seunghyun Hwang
 
Deep Learning-based Fully Automated Detection and Quantification of Acute Inf...
Deep Learning-based Fully Automated Detection and Quantification of Acute Inf...Deep Learning-based Fully Automated Detection and Quantification of Acute Inf...
Deep Learning-based Fully Automated Detection and Quantification of Acute Inf...Seunghyun Hwang
 
Diagnosis of Maxillary Sinusitis in Water’s view based on Deep learning model
Diagnosis of Maxillary Sinusitis in Water’s view based on Deep learning model Diagnosis of Maxillary Sinusitis in Water’s view based on Deep learning model
Diagnosis of Maxillary Sinusitis in Water’s view based on Deep learning model Seunghyun Hwang
 
Energy-based Model for Out-of-Distribution Detection in Deep Medical Image Se...
Energy-based Model for Out-of-Distribution Detection in Deep Medical Image Se...Energy-based Model for Out-of-Distribution Detection in Deep Medical Image Se...
Energy-based Model for Out-of-Distribution Detection in Deep Medical Image Se...Seunghyun Hwang
 
Segmenting Medical MRI via Recurrent Decoding Cell
Segmenting Medical MRI via Recurrent Decoding CellSegmenting Medical MRI via Recurrent Decoding Cell
Segmenting Medical MRI via Recurrent Decoding CellSeunghyun Hwang
 
Progressive learning and Disentanglement of hierarchical representations
Progressive learning and Disentanglement of hierarchical representationsProgressive learning and Disentanglement of hierarchical representations
Progressive learning and Disentanglement of hierarchical representationsSeunghyun Hwang
 
A Simple Framework for Contrastive Learning of Visual Representations
A Simple Framework for Contrastive Learning of Visual RepresentationsA Simple Framework for Contrastive Learning of Visual Representations
A Simple Framework for Contrastive Learning of Visual RepresentationsSeunghyun Hwang
 
Mix Conv: Mixed Depthwise Convolutional Kernels
Mix Conv: Mixed Depthwise Convolutional KernelsMix Conv: Mixed Depthwise Convolutional Kernels
Mix Conv: Mixed Depthwise Convolutional KernelsSeunghyun Hwang
 
Large Scale GAN Training for High Fidelity Natural Image Synthesis
Large Scale GAN Training for High Fidelity Natural Image SynthesisLarge Scale GAN Training for High Fidelity Natural Image Synthesis
Large Scale GAN Training for High Fidelity Natural Image SynthesisSeunghyun Hwang
 

More from Seunghyun Hwang (10)

An annotation sparsification strategy for 3D medical image segmentation via r...
An annotation sparsification strategy for 3D medical image segmentation via r...An annotation sparsification strategy for 3D medical image segmentation via r...
An annotation sparsification strategy for 3D medical image segmentation via r...
 
Do wide and deep networks learn the same things? Uncovering how neural networ...
Do wide and deep networks learn the same things? Uncovering how neural networ...Do wide and deep networks learn the same things? Uncovering how neural networ...
Do wide and deep networks learn the same things? Uncovering how neural networ...
 
Deep Learning-based Fully Automated Detection and Quantification of Acute Inf...
Deep Learning-based Fully Automated Detection and Quantification of Acute Inf...Deep Learning-based Fully Automated Detection and Quantification of Acute Inf...
Deep Learning-based Fully Automated Detection and Quantification of Acute Inf...
 
Diagnosis of Maxillary Sinusitis in Water’s view based on Deep learning model
Diagnosis of Maxillary Sinusitis in Water’s view based on Deep learning model Diagnosis of Maxillary Sinusitis in Water’s view based on Deep learning model
Diagnosis of Maxillary Sinusitis in Water’s view based on Deep learning model
 
Energy-based Model for Out-of-Distribution Detection in Deep Medical Image Se...
Energy-based Model for Out-of-Distribution Detection in Deep Medical Image Se...Energy-based Model for Out-of-Distribution Detection in Deep Medical Image Se...
Energy-based Model for Out-of-Distribution Detection in Deep Medical Image Se...
 
Segmenting Medical MRI via Recurrent Decoding Cell
Segmenting Medical MRI via Recurrent Decoding CellSegmenting Medical MRI via Recurrent Decoding Cell
Segmenting Medical MRI via Recurrent Decoding Cell
 
Progressive learning and Disentanglement of hierarchical representations
Progressive learning and Disentanglement of hierarchical representationsProgressive learning and Disentanglement of hierarchical representations
Progressive learning and Disentanglement of hierarchical representations
 
A Simple Framework for Contrastive Learning of Visual Representations
A Simple Framework for Contrastive Learning of Visual RepresentationsA Simple Framework for Contrastive Learning of Visual Representations
A Simple Framework for Contrastive Learning of Visual Representations
 
Mix Conv: Mixed Depthwise Convolutional Kernels
Mix Conv: Mixed Depthwise Convolutional KernelsMix Conv: Mixed Depthwise Convolutional Kernels
Mix Conv: Mixed Depthwise Convolutional Kernels
 
Large Scale GAN Training for High Fidelity Natural Image Synthesis
Large Scale GAN Training for High Fidelity Natural Image SynthesisLarge Scale GAN Training for High Fidelity Natural Image Synthesis
Large Scale GAN Training for High Fidelity Natural Image Synthesis
 

Recently uploaded

08448380779 Call Girls In Civil Lines Women Seeking Men
08448380779 Call Girls In Civil Lines Women Seeking Men08448380779 Call Girls In Civil Lines Women Seeking Men
08448380779 Call Girls In Civil Lines Women Seeking MenDelhi Call girls
 
Top 5 Benefits OF Using Muvi Live Paywall For Live Streams
Top 5 Benefits OF Using Muvi Live Paywall For Live StreamsTop 5 Benefits OF Using Muvi Live Paywall For Live Streams
Top 5 Benefits OF Using Muvi Live Paywall For Live StreamsRoshan Dwivedi
 
Strategies for Unlocking Knowledge Management in Microsoft 365 in the Copilot...
Strategies for Unlocking Knowledge Management in Microsoft 365 in the Copilot...Strategies for Unlocking Knowledge Management in Microsoft 365 in the Copilot...
Strategies for Unlocking Knowledge Management in Microsoft 365 in the Copilot...Drew Madelung
 
From Event to Action: Accelerate Your Decision Making with Real-Time Automation
From Event to Action: Accelerate Your Decision Making with Real-Time AutomationFrom Event to Action: Accelerate Your Decision Making with Real-Time Automation
From Event to Action: Accelerate Your Decision Making with Real-Time AutomationSafe Software
 
Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...
Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...
Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...apidays
 
🐬 The future of MySQL is Postgres 🐘
🐬  The future of MySQL is Postgres   🐘🐬  The future of MySQL is Postgres   🐘
🐬 The future of MySQL is Postgres 🐘RTylerCroy
 
Finology Group – Insurtech Innovation Award 2024
Finology Group – Insurtech Innovation Award 2024Finology Group – Insurtech Innovation Award 2024
Finology Group – Insurtech Innovation Award 2024The Digital Insurer
 
How to convert PDF to text with Nanonets
How to convert PDF to text with NanonetsHow to convert PDF to text with Nanonets
How to convert PDF to text with Nanonetsnaman860154
 
Driving Behavioral Change for Information Management through Data-Driven Gree...
Driving Behavioral Change for Information Management through Data-Driven Gree...Driving Behavioral Change for Information Management through Data-Driven Gree...
Driving Behavioral Change for Information Management through Data-Driven Gree...Enterprise Knowledge
 
04-2024-HHUG-Sales-and-Marketing-Alignment.pptx
04-2024-HHUG-Sales-and-Marketing-Alignment.pptx04-2024-HHUG-Sales-and-Marketing-Alignment.pptx
04-2024-HHUG-Sales-and-Marketing-Alignment.pptxHampshireHUG
 
CNv6 Instructor Chapter 6 Quality of Service
CNv6 Instructor Chapter 6 Quality of ServiceCNv6 Instructor Chapter 6 Quality of Service
CNv6 Instructor Chapter 6 Quality of Servicegiselly40
 
IAC 2024 - IA Fast Track to Search Focused AI Solutions
IAC 2024 - IA Fast Track to Search Focused AI SolutionsIAC 2024 - IA Fast Track to Search Focused AI Solutions
IAC 2024 - IA Fast Track to Search Focused AI SolutionsEnterprise Knowledge
 
Factors to Consider When Choosing Accounts Payable Services Providers.pptx
Factors to Consider When Choosing Accounts Payable Services Providers.pptxFactors to Consider When Choosing Accounts Payable Services Providers.pptx
Factors to Consider When Choosing Accounts Payable Services Providers.pptxKatpro Technologies
 
Mastering MySQL Database Architecture: Deep Dive into MySQL Shell and MySQL R...
Mastering MySQL Database Architecture: Deep Dive into MySQL Shell and MySQL R...Mastering MySQL Database Architecture: Deep Dive into MySQL Shell and MySQL R...
Mastering MySQL Database Architecture: Deep Dive into MySQL Shell and MySQL R...Miguel Araújo
 
Slack Application Development 101 Slides
Slack Application Development 101 SlidesSlack Application Development 101 Slides
Slack Application Development 101 Slidespraypatel2
 
The Codex of Business Writing Software for Real-World Solutions 2.pptx
The Codex of Business Writing Software for Real-World Solutions 2.pptxThe Codex of Business Writing Software for Real-World Solutions 2.pptx
The Codex of Business Writing Software for Real-World Solutions 2.pptxMalak Abu Hammad
 
Raspberry Pi 5: Challenges and Solutions in Bringing up an OpenGL/Vulkan Driv...
Raspberry Pi 5: Challenges and Solutions in Bringing up an OpenGL/Vulkan Driv...Raspberry Pi 5: Challenges and Solutions in Bringing up an OpenGL/Vulkan Driv...
Raspberry Pi 5: Challenges and Solutions in Bringing up an OpenGL/Vulkan Driv...Igalia
 
How to Troubleshoot Apps for the Modern Connected Worker
How to Troubleshoot Apps for the Modern Connected WorkerHow to Troubleshoot Apps for the Modern Connected Worker
How to Troubleshoot Apps for the Modern Connected WorkerThousandEyes
 
Data Cloud, More than a CDP by Matt Robison
Data Cloud, More than a CDP by Matt RobisonData Cloud, More than a CDP by Matt Robison
Data Cloud, More than a CDP by Matt RobisonAnna Loughnan Colquhoun
 
Unblocking The Main Thread Solving ANRs and Frozen Frames
Unblocking The Main Thread Solving ANRs and Frozen FramesUnblocking The Main Thread Solving ANRs and Frozen Frames
Unblocking The Main Thread Solving ANRs and Frozen FramesSinan KOZAK
 

Recently uploaded (20)

08448380779 Call Girls In Civil Lines Women Seeking Men
08448380779 Call Girls In Civil Lines Women Seeking Men08448380779 Call Girls In Civil Lines Women Seeking Men
08448380779 Call Girls In Civil Lines Women Seeking Men
 
Top 5 Benefits OF Using Muvi Live Paywall For Live Streams
Top 5 Benefits OF Using Muvi Live Paywall For Live StreamsTop 5 Benefits OF Using Muvi Live Paywall For Live Streams
Top 5 Benefits OF Using Muvi Live Paywall For Live Streams
 
Strategies for Unlocking Knowledge Management in Microsoft 365 in the Copilot...
Strategies for Unlocking Knowledge Management in Microsoft 365 in the Copilot...Strategies for Unlocking Knowledge Management in Microsoft 365 in the Copilot...
Strategies for Unlocking Knowledge Management in Microsoft 365 in the Copilot...
 
From Event to Action: Accelerate Your Decision Making with Real-Time Automation
From Event to Action: Accelerate Your Decision Making with Real-Time AutomationFrom Event to Action: Accelerate Your Decision Making with Real-Time Automation
From Event to Action: Accelerate Your Decision Making with Real-Time Automation
 
Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...
Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...
Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...
 
🐬 The future of MySQL is Postgres 🐘
🐬  The future of MySQL is Postgres   🐘🐬  The future of MySQL is Postgres   🐘
🐬 The future of MySQL is Postgres 🐘
 
Finology Group – Insurtech Innovation Award 2024
Finology Group – Insurtech Innovation Award 2024Finology Group – Insurtech Innovation Award 2024
Finology Group – Insurtech Innovation Award 2024
 
How to convert PDF to text with Nanonets
How to convert PDF to text with NanonetsHow to convert PDF to text with Nanonets
How to convert PDF to text with Nanonets
 
Driving Behavioral Change for Information Management through Data-Driven Gree...
Driving Behavioral Change for Information Management through Data-Driven Gree...Driving Behavioral Change for Information Management through Data-Driven Gree...
Driving Behavioral Change for Information Management through Data-Driven Gree...
 
04-2024-HHUG-Sales-and-Marketing-Alignment.pptx
04-2024-HHUG-Sales-and-Marketing-Alignment.pptx04-2024-HHUG-Sales-and-Marketing-Alignment.pptx
04-2024-HHUG-Sales-and-Marketing-Alignment.pptx
 
CNv6 Instructor Chapter 6 Quality of Service
CNv6 Instructor Chapter 6 Quality of ServiceCNv6 Instructor Chapter 6 Quality of Service
CNv6 Instructor Chapter 6 Quality of Service
 
IAC 2024 - IA Fast Track to Search Focused AI Solutions
IAC 2024 - IA Fast Track to Search Focused AI SolutionsIAC 2024 - IA Fast Track to Search Focused AI Solutions
IAC 2024 - IA Fast Track to Search Focused AI Solutions
 
Factors to Consider When Choosing Accounts Payable Services Providers.pptx
Factors to Consider When Choosing Accounts Payable Services Providers.pptxFactors to Consider When Choosing Accounts Payable Services Providers.pptx
Factors to Consider When Choosing Accounts Payable Services Providers.pptx
 
Mastering MySQL Database Architecture: Deep Dive into MySQL Shell and MySQL R...
Mastering MySQL Database Architecture: Deep Dive into MySQL Shell and MySQL R...Mastering MySQL Database Architecture: Deep Dive into MySQL Shell and MySQL R...
Mastering MySQL Database Architecture: Deep Dive into MySQL Shell and MySQL R...
 
Slack Application Development 101 Slides
Slack Application Development 101 SlidesSlack Application Development 101 Slides
Slack Application Development 101 Slides
 
The Codex of Business Writing Software for Real-World Solutions 2.pptx
The Codex of Business Writing Software for Real-World Solutions 2.pptxThe Codex of Business Writing Software for Real-World Solutions 2.pptx
The Codex of Business Writing Software for Real-World Solutions 2.pptx
 
Raspberry Pi 5: Challenges and Solutions in Bringing up an OpenGL/Vulkan Driv...
Raspberry Pi 5: Challenges and Solutions in Bringing up an OpenGL/Vulkan Driv...Raspberry Pi 5: Challenges and Solutions in Bringing up an OpenGL/Vulkan Driv...
Raspberry Pi 5: Challenges and Solutions in Bringing up an OpenGL/Vulkan Driv...
 
How to Troubleshoot Apps for the Modern Connected Worker
How to Troubleshoot Apps for the Modern Connected WorkerHow to Troubleshoot Apps for the Modern Connected Worker
How to Troubleshoot Apps for the Modern Connected Worker
 
Data Cloud, More than a CDP by Matt Robison
Data Cloud, More than a CDP by Matt RobisonData Cloud, More than a CDP by Matt Robison
Data Cloud, More than a CDP by Matt Robison
 
Unblocking The Main Thread Solving ANRs and Frozen Frames
Unblocking The Main Thread Solving ANRs and Frozen FramesUnblocking The Main Thread Solving ANRs and Frozen Frames
Unblocking The Main Thread Solving ANRs and Frozen Frames
 

A Probabilistic U-Net for Segmentation of Ambiguous Images

  • 1. A Probabilistic U-Net for Segmentation of Ambiguous Images Hwang seung hyun Yonsei University Severance Hospital CCIDS DeepMind, Division of Medical Image Computing, German Cancer Research Center, Germany | NIPS 2018 2020.04.19
  • 2. Introduction Related Work Methods and Experiments 01 02 03 Conclusion 04 Yonsei Unversity Severance Hospital CCIDS Contents
  • 3. Probabilistic Unet Introduction – Limitations of prior methods • There exist ambiguities in segmentation task, especially in medical imaging applications • A lesion might be clearly visible, but ground truth labels can vary depending on radiologists. • Most existing segmentation algorithms either provide only consistent hypothesis of a pixel-wise probability(e.g. “each pixel is 50% cat, 50% dog) • Pixel wise probabilities ignores all co-variances between the pixels. • Existing methods are Ensemble Unet, dropout Unet, M heads model, etc. Introduction / Related Work / Methods and Experiments / Conclusion
  • 4. Probabilistic Unet Introduction – Probabilistic Unet Architecture • Probabilistic Unet provides multiple segmentation hypotheses for ambiguous images. • Combines conditional variational auto encoder(CVAE), and U-Net • First extract latent space and encodes the possible segmentation variants • Random sample from the space is injected into the Unet to produce segmentation map. Introduction / Related Work / Methods and Experiments / Conclusion
  • 5. Probabilistic Unet Introduction – Contributions • Provides consistent segmentation maps instead of pixel-wise probabilities, providing joint likelihood of modes. • Able to learn calibrated probabilities of segmentation modes. • Can produce diverse outputs for single image Introduction / Related Work / Methods and Experiments / Conclusion
  • 6. Related Work CVAE (Conditional Variational Auto Encoder) Introduction / Related Work / Methods and Experiments / Conclusion • Encoder를 통해 도출된 latent coding Z를 가우시 안 분포로 나타내기 위해 분산과 평균을 이용함 • Label 정보를 추가로 넣어준다
  • 7. Related Work U-Net Introduction / Related Work / Methods and Experiments / Conclusion • Encoding Phase
  • 8. Methods and Experiments Network Architecture Introduction / Related Work / Methods and Experiments / Conclusion • Sampling Process • Training Process
  • 9. Methods and Experiments Sampling Process Introduction / Related Work / Methods and Experiments / Conclusion • Prior Net (Unet’s encoding phase + global average pooling) produces Latent Space • Each position in this space encodes a segmentation variant • Broadcast the sample to feature map with the same shape as the segmentation map, and concatenate this map to the las activation map of U-Net * P : prior probability distribution * fcomb = three subsequent 1x1 convolutions * S: segmentation map corresponding to point z in latent space
  • 10. Methods and Experiments Training Process Introduction / Related Work / Methods and Experiments / Conclusion • Introduce Posterior Net that learns to recognize a useful segmentation variant • Posterior Net and Prior Net are updated through the standard training procedure for CVAE, by minimizing variational lower bound (Kullback-Leibler divergence) • Cross-entropy loss penalizes differences between S and Y • KL loss pulls the posterior distribution and prior distribution towards each other • Eventually covers the space of all useful segmentation variants for input image 21
  • 11. Methods and Experiments Sampling Process Introduction / Related Work / Methods and Experiments / Conclusion Output Samples Visualization of the Latent Space
  • 12. Methods and Experiments Introduction / Related Work / Methods and Experiments / Conclusion Performance Measures • Generalized Energy Distance Matrix • Not only compare deterministic prediction, but also compares distributions of segmentations * d: distance measure * Y, Y’ : Independent samples from the ground truth distribution * S, S’: independent samples from the predicted distribution * d(x,y) = 1 - IOU(x,y)
  • 13. Methods and Experiments Introduction / Related Work / Methods and Experiments / Conclusion Results
  • 14. Methods and Experiments Introduction / Related Work / Methods and Experiments / Conclusion Results • Energy Distance decreases as more samples are drawn indicating an improved matching of the GT distribution, as well as enhanced sample diversity.
  • 15. Conclusion Introduction / Related Work / Methods and Experiments / Conclusion • Each sample produced by probabilistic Unet is consistent segmentation result that closely match the multi-modal GT distributions • Employed energy distance matrix measures whether the model’s individual samples are both coherent, and whether they are produced with expected frequencies. • Can be used to assess annotations with model • Probabilistic U Net can replace the currently applied deterministic U Nets in large field of studies, especially in the medical domain • Guide steps to resolve ambiguities