SlideShare a Scribd company logo
1 of 36
馭繁為簡
- 在 AI 雲上實現高效能的弱監
督式數位病理影像分析
Sean & C.C Chen
What is Digital Pathology
What is Digital Pathology
What is Digital Pathology
What is Digital Pathology
What is Digital Pathology
Digital Pathology Applications - Cancer Screening
• Goals
– Whether this Whole Slide Image (WSI) has cancer cells
– Location of cancer cells?
100,000 pixels
80,000 pixels
Cancer /
Non-Cancer
---
Segments1
Segments2
…
512 pixels
512 pixels
Annotations
Cancer Screening
Traditional Method:Patch based model
Annotations  Patching  Training
Two-Level AI Model for
Cancer Detection on
Whole Slide Image
Cancer Screening
Traditional Method:Patch based model
Annotations  Patching  Training
Two-Level AI Model for
Cancer Detection on
Whole Slide Image
Patch-level model (>10M
Patches)
Background, Benign, Cancer
Performance
- Accuracy: 98%
- AUC: 0.99
Ground Truth : Cancer, Normal Tissue
Shadowed area : Cancer predicted by AI
Cancer Screening
Traditional Method:Patch based model
Annotations  Patching  Training
Two-Level AI Model for
Cancer Detection on
Whole Slide Image
Patch-level model (>10M
Patches)
Background, Benign, Cancer
Performance
- Accuracy: 98%
- AUC: 0.99
Slide-level model
260 Training, 100 Testing
Performance
- Accuracy: 97%
- AUC: 0.98
Benign or NPC ?
Ground Truth : Cancer, Normal Tissue
Shadowed area : Cancer predicted by AI
Cancer Screening
Traditional Method:Patch based model
Annotation Example
Cancer Screening - Patch based model
Pros Cons
Good Performance
Good Localization
Two stages model
Need “LOTS OF”
segment annotations
 1hr/slide
 Confused
annotations
Low inference speed
 15-30 mins/slide
標註難
專業的標註更難
如何既降低專業標註需求又維持 AI 的品質?
• Given a image-level ground-truth
• Where is the target?
Weakly Supervised Learning
[Gupta et.al, 2019]
Digital Pathology Applications - Cancer Screening
Positive
Positive
Negative
Positive
Image Label
Cancer region? (if any)
WSOD through Class Activation Map
[Rajpurkar et.al, 2018]
[Zhou et.al, 2016]
[Gondal et.al, 2017]
Digital Pathology Applications - Cancer Screening
Positive
Positive
Negative
Positive
Image Label Residual Networks
Positive?
Class Activation Map
[He et.al, 2016]
[Zhou et.al, 2016]
Issue of Out-Of-Memory
Assume a WSI (60k x 60k)
• It will take ~ 320 Gb Memory
GPU Memory/GPU Cores
Memory
Bandwidth
Quadro RTX
8000
48 GB 4608 672 GB/sec
Tesla V100 v2 32 GB 5120 900 GB/sec
Tesla V100 16 GB 5120 900 GB/sec
P100 16 GB 3584 732 GB/sec
P40 24GB 3840 346 GB/sec
K80 24GB 4992 480 GB/sec
GPU cannot hold the super-resolution images
Issue of Out-Of-Memory
Assume a WSI (60k x 60k)
• It will take ~ 320 Gb Memory
GPU Memory/GPU Cores
Memory
Bandwidth
Quadro RTX
8000
48 GB 4608 672 GB/sec
Tesla V100 v2 32 GB 5120 900 GB/sec
Tesla V100 16 GB 5120 900 GB/sec
P100 16 GB 3584 732 GB/sec
P40 24GB 3840 346 GB/sec
K80 24GB 4992 480 GB/sec
GPU cannot hold the super-resolution images
• Unified Memory:Use system RAM to keep parts of
model/weights and load them onto GPU RAM when
needed.
Issue of Out-Of-Memory
GPU cannot hold the super-resolution images
• Unified Memory:Use system RAM to keep parts of
model/weights and load them onto GPU RAM when
needed.
We need a server that has
• Large system RAM & Storage
• Fast CPU-GPU transfer
Assume a WSI (60k x 60k)
• It will take ~ 320 Gb Memory
GPU Memory/GPU Cores
Memory
Bandwidth
Quadro RTX
8000
48 GB 4608 672 GB/sec
Tesla V100 v2 32 GB 5120 900 GB/sec
Tesla V100 16 GB 5120 900 GB/sec
P100 16 GB 3584 732 GB/sec
P40 24GB 3840 346 GB/sec
K80 24GB 4992 480 GB/sec
How Taiwania2 Help us
8x GPU (V100-v2, 32Gb)
768 Gb RAM
2 NUMA nodes
8x GPU (V100-v2, 32Gb)
768 Gb RAM
2 NUMA nodes
Storage (6T)
Login Node
…
4x Infiniband
Scheduling:Slurm
Comm:OpenMPI & Horovod
Framework:Tensorflow
Ultra-Patch Workflow
1112 images (positive:557; negative: 555)
QuantaGrid D52G nodes on Taiwania2
8 Tesla V100 (32gb) and 768 Gb system memory per node
With batch size = 1
~320 Gb system memory is used for training through Unified Memory
Training:0.0067 Image/sec. (each update takes 2.5 mins)
Inferencing:~20 sec/WSI
Hardware
Data
Parameter
Time
Ultra-Patch Method - Result
• Compared to the two-stages model,
our Ultra-Patch Method with 10k
inputs got a competitive result
• Compared to different input-size, larger
inputs generally got better results
Ultra-Patch Method - Result
• Compared to the two-stages model,
our Ultra-Patch Method with 10k
inputs got a competitive result
• Compared to different input-size, larger
inputs generally got better results
Ultra-Patch Method - Result
Ultra-Patch Method - Result
• Unified Memory
• Transfer cost between CPU-GPU
• Group-Execution and Group-Prefetching
Optimization for training speed
Load Tensors Do Compute Load Tensors …
…
Load Tensors Do Compute Load Tensors
Load Tensors Do Compute
…
Typical workflow
With Prefetching
Optimization for training speed
Typical workflow
Optimized workflow GPU Page Faults Reduced
• Unified Memory
• Transfer cost between CPU-GPU
• Group-Execution and Group-Prefetching
• Parallel Training
• Distributed training through Horovod
Optimization for training speed
Load Tensors Do Compute Load Tensors …
…
Load Tensors Do Compute Load Tensors
Load Tensors Do Compute
…
Typical workflow
With Prefetching
Ultra-Patch workflow
1112 images (positive:557; negative: 555)
QuantaGrid D52G nodes on Taiwania2
8 Tesla V100 (32gb) and 768 Gb system memory per node
With batch size = 4 (mixed-precision)
~700 Gb system memory is used for training through Unified Memory
Training:1.06 Image/sec
Inferencing:~20 sec/WSI
Hardware
Data
Parameter
Time
• ~5.11x acceleration on a single GPU optimization.
• ~147.28x acceleration on 8 nodes
• ~498x acceleration on 32 nodes
Experiments & Optimization Results
8 nodes
1 nodes
Data-size comparison
• More data, better
performance.
Impact of Ultra-Patch On Medical Image Analysis
• Speed up the developing time for EACH cancer
classifier/detector
• Laborious annotation tasks will be reduced
• Results will be available faster
• 國家高速網路中心提供台灣杉二號運算環境與 HPC 技術支援
• 林口長庚醫院提供訓練與測試資料集
Acknowledgement
Prevision the Future of AI

More Related Content

Similar to Weakly Supervised Whole Slide Image Analysis Using Cloud Computing

2018AOI論壇_深度學習於表面瑕疪檢測_元智大學蔡篤銘
2018AOI論壇_深度學習於表面瑕疪檢測_元智大學蔡篤銘2018AOI論壇_深度學習於表面瑕疪檢測_元智大學蔡篤銘
2018AOI論壇_深度學習於表面瑕疪檢測_元智大學蔡篤銘CHENHuiMei
 
"Energy-efficient Hardware for Embedded Vision and Deep Convolutional Neural ...
"Energy-efficient Hardware for Embedded Vision and Deep Convolutional Neural ..."Energy-efficient Hardware for Embedded Vision and Deep Convolutional Neural ...
"Energy-efficient Hardware for Embedded Vision and Deep Convolutional Neural ...Edge AI and Vision Alliance
 
Introducing Amazon EC2 P3 Instance - Featuring the Most Powerful GPU for Mach...
Introducing Amazon EC2 P3 Instance - Featuring the Most Powerful GPU for Mach...Introducing Amazon EC2 P3 Instance - Featuring the Most Powerful GPU for Mach...
Introducing Amazon EC2 P3 Instance - Featuring the Most Powerful GPU for Mach...Amazon Web Services
 
Deep Dive on Delivering Amazon EC2 Instance Performance
Deep Dive on Delivering Amazon EC2 Instance PerformanceDeep Dive on Delivering Amazon EC2 Instance Performance
Deep Dive on Delivering Amazon EC2 Instance PerformanceAmazon Web Services
 
MIT's experience on OpenPOWER/POWER 9 platform
MIT's experience on OpenPOWER/POWER 9 platformMIT's experience on OpenPOWER/POWER 9 platform
MIT's experience on OpenPOWER/POWER 9 platformGanesan Narayanasamy
 
Meetup web scale architecture quantum computing (Part 1 16-10-2018)
Meetup web scale architecture quantum computing (Part 1 16-10-2018)Meetup web scale architecture quantum computing (Part 1 16-10-2018)
Meetup web scale architecture quantum computing (Part 1 16-10-2018)Rolf Huisman
 
STAR CCM GLOBAL CONFERENCE UBERCLOUD
STAR CCM GLOBAL CONFERENCE UBERCLOUDSTAR CCM GLOBAL CONFERENCE UBERCLOUD
STAR CCM GLOBAL CONFERENCE UBERCLOUDThomas Francis
 
20211119 ntuh azure hpc workshop final
20211119 ntuh azure hpc workshop final20211119 ntuh azure hpc workshop final
20211119 ntuh azure hpc workshop finalMeng-Ru (Raymond) Tsai
 
Scalable Multi-Node Deep Learning Training in the Cloud (CMP368-R1) - AWS re:...
Scalable Multi-Node Deep Learning Training in the Cloud (CMP368-R1) - AWS re:...Scalable Multi-Node Deep Learning Training in the Cloud (CMP368-R1) - AWS re:...
Scalable Multi-Node Deep Learning Training in the Cloud (CMP368-R1) - AWS re:...Amazon Web Services
 
Performance Analysis and Optimizations of CAE Applications (Case Study: STAR_...
Performance Analysis and Optimizations of CAE Applications (Case Study: STAR_...Performance Analysis and Optimizations of CAE Applications (Case Study: STAR_...
Performance Analysis and Optimizations of CAE Applications (Case Study: STAR_...Fisnik Kraja
 
Clipper: A Low-Latency Online Prediction Serving System
Clipper: A Low-Latency Online Prediction Serving SystemClipper: A Low-Latency Online Prediction Serving System
Clipper: A Low-Latency Online Prediction Serving SystemDatabricks
 
AWS Summit Bogotá Track Avanzado: EC2 avanzado
AWS Summit Bogotá Track Avanzado: EC2 avanzadoAWS Summit Bogotá Track Avanzado: EC2 avanzado
AWS Summit Bogotá Track Avanzado: EC2 avanzadoAmazon Web Services
 
Octnews featured article
Octnews featured articleOctnews featured article
Octnews featured articleKangZhang
 
Foundations of Amazon EC2 - SRV319
Foundations of Amazon EC2 - SRV319 Foundations of Amazon EC2 - SRV319
Foundations of Amazon EC2 - SRV319 Amazon Web Services
 
Cache Optimization Techniques for General Purpose Graphic Processing Units
Cache Optimization Techniques for General Purpose Graphic Processing UnitsCache Optimization Techniques for General Purpose Graphic Processing Units
Cache Optimization Techniques for General Purpose Graphic Processing UnitsVajira Thambawita
 
Nervana and the Future of Computing
Nervana and the Future of ComputingNervana and the Future of Computing
Nervana and the Future of ComputingIntel Nervana
 
High Performance Erlang - Pitfalls and Solutions
High Performance Erlang - Pitfalls and SolutionsHigh Performance Erlang - Pitfalls and Solutions
High Performance Erlang - Pitfalls and SolutionsYinghai Lu
 
Imaging automotive 2015 addfor v002
Imaging automotive 2015   addfor v002Imaging automotive 2015   addfor v002
Imaging automotive 2015 addfor v002Enrico Busto
 

Similar to Weakly Supervised Whole Slide Image Analysis Using Cloud Computing (20)

2018AOI論壇_深度學習於表面瑕疪檢測_元智大學蔡篤銘
2018AOI論壇_深度學習於表面瑕疪檢測_元智大學蔡篤銘2018AOI論壇_深度學習於表面瑕疪檢測_元智大學蔡篤銘
2018AOI論壇_深度學習於表面瑕疪檢測_元智大學蔡篤銘
 
"Energy-efficient Hardware for Embedded Vision and Deep Convolutional Neural ...
"Energy-efficient Hardware for Embedded Vision and Deep Convolutional Neural ..."Energy-efficient Hardware for Embedded Vision and Deep Convolutional Neural ...
"Energy-efficient Hardware for Embedded Vision and Deep Convolutional Neural ...
 
Introducing Amazon EC2 P3 Instance - Featuring the Most Powerful GPU for Mach...
Introducing Amazon EC2 P3 Instance - Featuring the Most Powerful GPU for Mach...Introducing Amazon EC2 P3 Instance - Featuring the Most Powerful GPU for Mach...
Introducing Amazon EC2 P3 Instance - Featuring the Most Powerful GPU for Mach...
 
Deep Dive on Delivering Amazon EC2 Instance Performance
Deep Dive on Delivering Amazon EC2 Instance PerformanceDeep Dive on Delivering Amazon EC2 Instance Performance
Deep Dive on Delivering Amazon EC2 Instance Performance
 
MIT's experience on OpenPOWER/POWER 9 platform
MIT's experience on OpenPOWER/POWER 9 platformMIT's experience on OpenPOWER/POWER 9 platform
MIT's experience on OpenPOWER/POWER 9 platform
 
Meetup web scale architecture quantum computing (Part 1 16-10-2018)
Meetup web scale architecture quantum computing (Part 1 16-10-2018)Meetup web scale architecture quantum computing (Part 1 16-10-2018)
Meetup web scale architecture quantum computing (Part 1 16-10-2018)
 
STAR CCM GLOBAL CONFERENCE UBERCLOUD
STAR CCM GLOBAL CONFERENCE UBERCLOUDSTAR CCM GLOBAL CONFERENCE UBERCLOUD
STAR CCM GLOBAL CONFERENCE UBERCLOUD
 
20211119 ntuh azure hpc workshop final
20211119 ntuh azure hpc workshop final20211119 ntuh azure hpc workshop final
20211119 ntuh azure hpc workshop final
 
Scalable Multi-Node Deep Learning Training in the Cloud (CMP368-R1) - AWS re:...
Scalable Multi-Node Deep Learning Training in the Cloud (CMP368-R1) - AWS re:...Scalable Multi-Node Deep Learning Training in the Cloud (CMP368-R1) - AWS re:...
Scalable Multi-Node Deep Learning Training in the Cloud (CMP368-R1) - AWS re:...
 
Performance Analysis and Optimizations of CAE Applications (Case Study: STAR_...
Performance Analysis and Optimizations of CAE Applications (Case Study: STAR_...Performance Analysis and Optimizations of CAE Applications (Case Study: STAR_...
Performance Analysis and Optimizations of CAE Applications (Case Study: STAR_...
 
Clipper: A Low-Latency Online Prediction Serving System
Clipper: A Low-Latency Online Prediction Serving SystemClipper: A Low-Latency Online Prediction Serving System
Clipper: A Low-Latency Online Prediction Serving System
 
AWS Summit Bogotá Track Avanzado: EC2 avanzado
AWS Summit Bogotá Track Avanzado: EC2 avanzadoAWS Summit Bogotá Track Avanzado: EC2 avanzado
AWS Summit Bogotá Track Avanzado: EC2 avanzado
 
Octnews featured article
Octnews featured articleOctnews featured article
Octnews featured article
 
Foundations of Amazon EC2 - SRV319
Foundations of Amazon EC2 - SRV319 Foundations of Amazon EC2 - SRV319
Foundations of Amazon EC2 - SRV319
 
Cache Optimization Techniques for General Purpose Graphic Processing Units
Cache Optimization Techniques for General Purpose Graphic Processing UnitsCache Optimization Techniques for General Purpose Graphic Processing Units
Cache Optimization Techniques for General Purpose Graphic Processing Units
 
6monitor_NYMIIS
6monitor_NYMIIS6monitor_NYMIIS
6monitor_NYMIIS
 
QCon London.pdf
QCon London.pdfQCon London.pdf
QCon London.pdf
 
Nervana and the Future of Computing
Nervana and the Future of ComputingNervana and the Future of Computing
Nervana and the Future of Computing
 
High Performance Erlang - Pitfalls and Solutions
High Performance Erlang - Pitfalls and SolutionsHigh Performance Erlang - Pitfalls and Solutions
High Performance Erlang - Pitfalls and Solutions
 
Imaging automotive 2015 addfor v002
Imaging automotive 2015   addfor v002Imaging automotive 2015   addfor v002
Imaging automotive 2015 addfor v002
 

More from Sean Yu

AI-powered Medical Imaging Analysis for Precision Medicine
AI-powered Medical Imaging Analysis for Precision MedicineAI-powered Medical Imaging Analysis for Precision Medicine
AI-powered Medical Imaging Analysis for Precision MedicineSean Yu
 
NTU DBME5028 Week8 Transfer Learning
NTU DBME5028 Week8 Transfer LearningNTU DBME5028 Week8 Transfer Learning
NTU DBME5028 Week8 Transfer LearningSean Yu
 
NTU DBME5028 Week5 Introduction to Machine Learning
NTU DBME5028 Week5 Introduction to Machine Learning NTU DBME5028 Week5 Introduction to Machine Learning
NTU DBME5028 Week5 Introduction to Machine Learning Sean Yu
 
Practical aspects of medical image ai for hospital (IRB course)
Practical aspects of medical image ai for hospital (IRB course)Practical aspects of medical image ai for hospital (IRB course)
Practical aspects of medical image ai for hospital (IRB course)Sean Yu
 
Baisc Deep Learning HandsOn
Baisc Deep Learning HandsOnBaisc Deep Learning HandsOn
Baisc Deep Learning HandsOnSean Yu
 
[Taiwan AI Academy] Machine learning and deep learning application examples
[Taiwan AI Academy] Machine learning and deep learning application examples[Taiwan AI Academy] Machine learning and deep learning application examples
[Taiwan AI Academy] Machine learning and deep learning application examplesSean Yu
 
[Python - Deep Learning] Data generator
[Python - Deep Learning] Data generator[Python - Deep Learning] Data generator
[Python - Deep Learning] Data generatorSean Yu
 
日常生活中的機器學習與 AI 應用 - 院區公開演講
日常生活中的機器學習與 AI 應用 - 院區公開演講日常生活中的機器學習與 AI 應用 - 院區公開演講
日常生活中的機器學習與 AI 應用 - 院區公開演講Sean Yu
 
台灣人工智慧年會 - 兩位跨領域者的深度學習之旅
台灣人工智慧年會 - 兩位跨領域者的深度學習之旅台灣人工智慧年會 - 兩位跨領域者的深度學習之旅
台灣人工智慧年會 - 兩位跨領域者的深度學習之旅Sean Yu
 
R 語言教學: 探索性資料分析與文字探勘初探
R 語言教學: 探索性資料分析與文字探勘初探R 語言教學: 探索性資料分析與文字探勘初探
R 語言教學: 探索性資料分析與文字探勘初探Sean Yu
 

More from Sean Yu (10)

AI-powered Medical Imaging Analysis for Precision Medicine
AI-powered Medical Imaging Analysis for Precision MedicineAI-powered Medical Imaging Analysis for Precision Medicine
AI-powered Medical Imaging Analysis for Precision Medicine
 
NTU DBME5028 Week8 Transfer Learning
NTU DBME5028 Week8 Transfer LearningNTU DBME5028 Week8 Transfer Learning
NTU DBME5028 Week8 Transfer Learning
 
NTU DBME5028 Week5 Introduction to Machine Learning
NTU DBME5028 Week5 Introduction to Machine Learning NTU DBME5028 Week5 Introduction to Machine Learning
NTU DBME5028 Week5 Introduction to Machine Learning
 
Practical aspects of medical image ai for hospital (IRB course)
Practical aspects of medical image ai for hospital (IRB course)Practical aspects of medical image ai for hospital (IRB course)
Practical aspects of medical image ai for hospital (IRB course)
 
Baisc Deep Learning HandsOn
Baisc Deep Learning HandsOnBaisc Deep Learning HandsOn
Baisc Deep Learning HandsOn
 
[Taiwan AI Academy] Machine learning and deep learning application examples
[Taiwan AI Academy] Machine learning and deep learning application examples[Taiwan AI Academy] Machine learning and deep learning application examples
[Taiwan AI Academy] Machine learning and deep learning application examples
 
[Python - Deep Learning] Data generator
[Python - Deep Learning] Data generator[Python - Deep Learning] Data generator
[Python - Deep Learning] Data generator
 
日常生活中的機器學習與 AI 應用 - 院區公開演講
日常生活中的機器學習與 AI 應用 - 院區公開演講日常生活中的機器學習與 AI 應用 - 院區公開演講
日常生活中的機器學習與 AI 應用 - 院區公開演講
 
台灣人工智慧年會 - 兩位跨領域者的深度學習之旅
台灣人工智慧年會 - 兩位跨領域者的深度學習之旅台灣人工智慧年會 - 兩位跨領域者的深度學習之旅
台灣人工智慧年會 - 兩位跨領域者的深度學習之旅
 
R 語言教學: 探索性資料分析與文字探勘初探
R 語言教學: 探索性資料分析與文字探勘初探R 語言教學: 探索性資料分析與文字探勘初探
R 語言教學: 探索性資料分析與文字探勘初探
 

Recently uploaded

Kantar AI Summit- Under Embargo till Wednesday, 24th April 2024, 4 PM, IST.pdf
Kantar AI Summit- Under Embargo till Wednesday, 24th April 2024, 4 PM, IST.pdfKantar AI Summit- Under Embargo till Wednesday, 24th April 2024, 4 PM, IST.pdf
Kantar AI Summit- Under Embargo till Wednesday, 24th April 2024, 4 PM, IST.pdfSocial Samosa
 
办理学位证中佛罗里达大学毕业证,UCF成绩单原版一比一
办理学位证中佛罗里达大学毕业证,UCF成绩单原版一比一办理学位证中佛罗里达大学毕业证,UCF成绩单原版一比一
办理学位证中佛罗里达大学毕业证,UCF成绩单原版一比一F sss
 
DBA Basics: Getting Started with Performance Tuning.pdf
DBA Basics: Getting Started with Performance Tuning.pdfDBA Basics: Getting Started with Performance Tuning.pdf
DBA Basics: Getting Started with Performance Tuning.pdfJohn Sterrett
 
04242024_CCC TUG_Joins and Relationships
04242024_CCC TUG_Joins and Relationships04242024_CCC TUG_Joins and Relationships
04242024_CCC TUG_Joins and Relationshipsccctableauusergroup
 
VIP Call Girls Service Miyapur Hyderabad Call +91-8250192130
VIP Call Girls Service Miyapur Hyderabad Call +91-8250192130VIP Call Girls Service Miyapur Hyderabad Call +91-8250192130
VIP Call Girls Service Miyapur Hyderabad Call +91-8250192130Suhani Kapoor
 
Effects of Smartphone Addiction on the Academic Performances of Grades 9 to 1...
Effects of Smartphone Addiction on the Academic Performances of Grades 9 to 1...Effects of Smartphone Addiction on the Academic Performances of Grades 9 to 1...
Effects of Smartphone Addiction on the Academic Performances of Grades 9 to 1...limedy534
 
VIP High Class Call Girls Jamshedpur Anushka 8250192130 Independent Escort Se...
VIP High Class Call Girls Jamshedpur Anushka 8250192130 Independent Escort Se...VIP High Class Call Girls Jamshedpur Anushka 8250192130 Independent Escort Se...
VIP High Class Call Girls Jamshedpur Anushka 8250192130 Independent Escort Se...Suhani Kapoor
 
ASML's Taxonomy Adventure by Daniel Canter
ASML's Taxonomy Adventure by Daniel CanterASML's Taxonomy Adventure by Daniel Canter
ASML's Taxonomy Adventure by Daniel Cantervoginip
 
vip Sarai Rohilla Call Girls 9999965857 Call or WhatsApp Now Book
vip Sarai Rohilla Call Girls 9999965857 Call or WhatsApp Now Bookvip Sarai Rohilla Call Girls 9999965857 Call or WhatsApp Now Book
vip Sarai Rohilla Call Girls 9999965857 Call or WhatsApp Now Bookmanojkuma9823
 
Amazon TQM (2) Amazon TQM (2)Amazon TQM (2).pptx
Amazon TQM (2) Amazon TQM (2)Amazon TQM (2).pptxAmazon TQM (2) Amazon TQM (2)Amazon TQM (2).pptx
Amazon TQM (2) Amazon TQM (2)Amazon TQM (2).pptxAbdelrhman abooda
 
GA4 Without Cookies [Measure Camp AMS]
GA4 Without Cookies [Measure Camp AMS]GA4 Without Cookies [Measure Camp AMS]
GA4 Without Cookies [Measure Camp AMS]📊 Markus Baersch
 
Beautiful Sapna Vip Call Girls Hauz Khas 9711199012 Call /Whatsapps
Beautiful Sapna Vip  Call Girls Hauz Khas 9711199012 Call /WhatsappsBeautiful Sapna Vip  Call Girls Hauz Khas 9711199012 Call /Whatsapps
Beautiful Sapna Vip Call Girls Hauz Khas 9711199012 Call /Whatsappssapnasaifi408
 
Dubai Call Girls Wifey O52&786472 Call Girls Dubai
Dubai Call Girls Wifey O52&786472 Call Girls DubaiDubai Call Girls Wifey O52&786472 Call Girls Dubai
Dubai Call Girls Wifey O52&786472 Call Girls Dubaihf8803863
 
From idea to production in a day – Leveraging Azure ML and Streamlit to build...
From idea to production in a day – Leveraging Azure ML and Streamlit to build...From idea to production in a day – Leveraging Azure ML and Streamlit to build...
From idea to production in a day – Leveraging Azure ML and Streamlit to build...Florian Roscheck
 
꧁❤ Greater Noida Call Girls Delhi ❤꧂ 9711199171 ☎️ Hard And Sexy Vip Call
꧁❤ Greater Noida Call Girls Delhi ❤꧂ 9711199171 ☎️ Hard And Sexy Vip Call꧁❤ Greater Noida Call Girls Delhi ❤꧂ 9711199171 ☎️ Hard And Sexy Vip Call
꧁❤ Greater Noida Call Girls Delhi ❤꧂ 9711199171 ☎️ Hard And Sexy Vip Callshivangimorya083
 
RadioAdProWritingCinderellabyButleri.pdf
RadioAdProWritingCinderellabyButleri.pdfRadioAdProWritingCinderellabyButleri.pdf
RadioAdProWritingCinderellabyButleri.pdfgstagge
 
Call Girls In Mahipalpur O9654467111 Escorts Service
Call Girls In Mahipalpur O9654467111  Escorts ServiceCall Girls In Mahipalpur O9654467111  Escorts Service
Call Girls In Mahipalpur O9654467111 Escorts ServiceSapana Sha
 
Call Girls in Defence Colony Delhi 💯Call Us 🔝8264348440🔝
Call Girls in Defence Colony Delhi 💯Call Us 🔝8264348440🔝Call Girls in Defence Colony Delhi 💯Call Us 🔝8264348440🔝
Call Girls in Defence Colony Delhi 💯Call Us 🔝8264348440🔝soniya singh
 
High Class Call Girls Noida Sector 39 Aarushi 🔝8264348440🔝 Independent Escort...
High Class Call Girls Noida Sector 39 Aarushi 🔝8264348440🔝 Independent Escort...High Class Call Girls Noida Sector 39 Aarushi 🔝8264348440🔝 Independent Escort...
High Class Call Girls Noida Sector 39 Aarushi 🔝8264348440🔝 Independent Escort...soniya singh
 

Recently uploaded (20)

Kantar AI Summit- Under Embargo till Wednesday, 24th April 2024, 4 PM, IST.pdf
Kantar AI Summit- Under Embargo till Wednesday, 24th April 2024, 4 PM, IST.pdfKantar AI Summit- Under Embargo till Wednesday, 24th April 2024, 4 PM, IST.pdf
Kantar AI Summit- Under Embargo till Wednesday, 24th April 2024, 4 PM, IST.pdf
 
办理学位证中佛罗里达大学毕业证,UCF成绩单原版一比一
办理学位证中佛罗里达大学毕业证,UCF成绩单原版一比一办理学位证中佛罗里达大学毕业证,UCF成绩单原版一比一
办理学位证中佛罗里达大学毕业证,UCF成绩单原版一比一
 
DBA Basics: Getting Started with Performance Tuning.pdf
DBA Basics: Getting Started with Performance Tuning.pdfDBA Basics: Getting Started with Performance Tuning.pdf
DBA Basics: Getting Started with Performance Tuning.pdf
 
04242024_CCC TUG_Joins and Relationships
04242024_CCC TUG_Joins and Relationships04242024_CCC TUG_Joins and Relationships
04242024_CCC TUG_Joins and Relationships
 
VIP Call Girls Service Miyapur Hyderabad Call +91-8250192130
VIP Call Girls Service Miyapur Hyderabad Call +91-8250192130VIP Call Girls Service Miyapur Hyderabad Call +91-8250192130
VIP Call Girls Service Miyapur Hyderabad Call +91-8250192130
 
Effects of Smartphone Addiction on the Academic Performances of Grades 9 to 1...
Effects of Smartphone Addiction on the Academic Performances of Grades 9 to 1...Effects of Smartphone Addiction on the Academic Performances of Grades 9 to 1...
Effects of Smartphone Addiction on the Academic Performances of Grades 9 to 1...
 
VIP High Class Call Girls Jamshedpur Anushka 8250192130 Independent Escort Se...
VIP High Class Call Girls Jamshedpur Anushka 8250192130 Independent Escort Se...VIP High Class Call Girls Jamshedpur Anushka 8250192130 Independent Escort Se...
VIP High Class Call Girls Jamshedpur Anushka 8250192130 Independent Escort Se...
 
ASML's Taxonomy Adventure by Daniel Canter
ASML's Taxonomy Adventure by Daniel CanterASML's Taxonomy Adventure by Daniel Canter
ASML's Taxonomy Adventure by Daniel Canter
 
vip Sarai Rohilla Call Girls 9999965857 Call or WhatsApp Now Book
vip Sarai Rohilla Call Girls 9999965857 Call or WhatsApp Now Bookvip Sarai Rohilla Call Girls 9999965857 Call or WhatsApp Now Book
vip Sarai Rohilla Call Girls 9999965857 Call or WhatsApp Now Book
 
Amazon TQM (2) Amazon TQM (2)Amazon TQM (2).pptx
Amazon TQM (2) Amazon TQM (2)Amazon TQM (2).pptxAmazon TQM (2) Amazon TQM (2)Amazon TQM (2).pptx
Amazon TQM (2) Amazon TQM (2)Amazon TQM (2).pptx
 
GA4 Without Cookies [Measure Camp AMS]
GA4 Without Cookies [Measure Camp AMS]GA4 Without Cookies [Measure Camp AMS]
GA4 Without Cookies [Measure Camp AMS]
 
Beautiful Sapna Vip Call Girls Hauz Khas 9711199012 Call /Whatsapps
Beautiful Sapna Vip  Call Girls Hauz Khas 9711199012 Call /WhatsappsBeautiful Sapna Vip  Call Girls Hauz Khas 9711199012 Call /Whatsapps
Beautiful Sapna Vip Call Girls Hauz Khas 9711199012 Call /Whatsapps
 
Dubai Call Girls Wifey O52&786472 Call Girls Dubai
Dubai Call Girls Wifey O52&786472 Call Girls DubaiDubai Call Girls Wifey O52&786472 Call Girls Dubai
Dubai Call Girls Wifey O52&786472 Call Girls Dubai
 
From idea to production in a day – Leveraging Azure ML and Streamlit to build...
From idea to production in a day – Leveraging Azure ML and Streamlit to build...From idea to production in a day – Leveraging Azure ML and Streamlit to build...
From idea to production in a day – Leveraging Azure ML and Streamlit to build...
 
꧁❤ Greater Noida Call Girls Delhi ❤꧂ 9711199171 ☎️ Hard And Sexy Vip Call
꧁❤ Greater Noida Call Girls Delhi ❤꧂ 9711199171 ☎️ Hard And Sexy Vip Call꧁❤ Greater Noida Call Girls Delhi ❤꧂ 9711199171 ☎️ Hard And Sexy Vip Call
꧁❤ Greater Noida Call Girls Delhi ❤꧂ 9711199171 ☎️ Hard And Sexy Vip Call
 
RadioAdProWritingCinderellabyButleri.pdf
RadioAdProWritingCinderellabyButleri.pdfRadioAdProWritingCinderellabyButleri.pdf
RadioAdProWritingCinderellabyButleri.pdf
 
Call Girls In Mahipalpur O9654467111 Escorts Service
Call Girls In Mahipalpur O9654467111  Escorts ServiceCall Girls In Mahipalpur O9654467111  Escorts Service
Call Girls In Mahipalpur O9654467111 Escorts Service
 
E-Commerce Order PredictionShraddha Kamble.pptx
E-Commerce Order PredictionShraddha Kamble.pptxE-Commerce Order PredictionShraddha Kamble.pptx
E-Commerce Order PredictionShraddha Kamble.pptx
 
Call Girls in Defence Colony Delhi 💯Call Us 🔝8264348440🔝
Call Girls in Defence Colony Delhi 💯Call Us 🔝8264348440🔝Call Girls in Defence Colony Delhi 💯Call Us 🔝8264348440🔝
Call Girls in Defence Colony Delhi 💯Call Us 🔝8264348440🔝
 
High Class Call Girls Noida Sector 39 Aarushi 🔝8264348440🔝 Independent Escort...
High Class Call Girls Noida Sector 39 Aarushi 🔝8264348440🔝 Independent Escort...High Class Call Girls Noida Sector 39 Aarushi 🔝8264348440🔝 Independent Escort...
High Class Call Girls Noida Sector 39 Aarushi 🔝8264348440🔝 Independent Escort...
 

Weakly Supervised Whole Slide Image Analysis Using Cloud Computing

  • 1. 馭繁為簡 - 在 AI 雲上實現高效能的弱監 督式數位病理影像分析 Sean & C.C Chen
  • 2. What is Digital Pathology
  • 3. What is Digital Pathology
  • 4. What is Digital Pathology
  • 5. What is Digital Pathology
  • 6. What is Digital Pathology
  • 7. Digital Pathology Applications - Cancer Screening • Goals – Whether this Whole Slide Image (WSI) has cancer cells – Location of cancer cells? 100,000 pixels 80,000 pixels Cancer / Non-Cancer --- Segments1 Segments2 … 512 pixels 512 pixels Annotations
  • 8. Cancer Screening Traditional Method:Patch based model Annotations  Patching  Training Two-Level AI Model for Cancer Detection on Whole Slide Image
  • 9. Cancer Screening Traditional Method:Patch based model Annotations  Patching  Training Two-Level AI Model for Cancer Detection on Whole Slide Image Patch-level model (>10M Patches) Background, Benign, Cancer Performance - Accuracy: 98% - AUC: 0.99 Ground Truth : Cancer, Normal Tissue Shadowed area : Cancer predicted by AI
  • 10. Cancer Screening Traditional Method:Patch based model Annotations  Patching  Training Two-Level AI Model for Cancer Detection on Whole Slide Image Patch-level model (>10M Patches) Background, Benign, Cancer Performance - Accuracy: 98% - AUC: 0.99 Slide-level model 260 Training, 100 Testing Performance - Accuracy: 97% - AUC: 0.98 Benign or NPC ? Ground Truth : Cancer, Normal Tissue Shadowed area : Cancer predicted by AI
  • 13. Cancer Screening - Patch based model Pros Cons Good Performance Good Localization Two stages model Need “LOTS OF” segment annotations  1hr/slide  Confused annotations Low inference speed  15-30 mins/slide
  • 15. • Given a image-level ground-truth • Where is the target? Weakly Supervised Learning [Gupta et.al, 2019]
  • 16. Digital Pathology Applications - Cancer Screening Positive Positive Negative Positive Image Label Cancer region? (if any)
  • 17. WSOD through Class Activation Map [Rajpurkar et.al, 2018] [Zhou et.al, 2016] [Gondal et.al, 2017]
  • 18. Digital Pathology Applications - Cancer Screening Positive Positive Negative Positive Image Label Residual Networks Positive? Class Activation Map [He et.al, 2016] [Zhou et.al, 2016]
  • 19. Issue of Out-Of-Memory Assume a WSI (60k x 60k) • It will take ~ 320 Gb Memory GPU Memory/GPU Cores Memory Bandwidth Quadro RTX 8000 48 GB 4608 672 GB/sec Tesla V100 v2 32 GB 5120 900 GB/sec Tesla V100 16 GB 5120 900 GB/sec P100 16 GB 3584 732 GB/sec P40 24GB 3840 346 GB/sec K80 24GB 4992 480 GB/sec GPU cannot hold the super-resolution images
  • 20. Issue of Out-Of-Memory Assume a WSI (60k x 60k) • It will take ~ 320 Gb Memory GPU Memory/GPU Cores Memory Bandwidth Quadro RTX 8000 48 GB 4608 672 GB/sec Tesla V100 v2 32 GB 5120 900 GB/sec Tesla V100 16 GB 5120 900 GB/sec P100 16 GB 3584 732 GB/sec P40 24GB 3840 346 GB/sec K80 24GB 4992 480 GB/sec GPU cannot hold the super-resolution images • Unified Memory:Use system RAM to keep parts of model/weights and load them onto GPU RAM when needed.
  • 21. Issue of Out-Of-Memory GPU cannot hold the super-resolution images • Unified Memory:Use system RAM to keep parts of model/weights and load them onto GPU RAM when needed. We need a server that has • Large system RAM & Storage • Fast CPU-GPU transfer Assume a WSI (60k x 60k) • It will take ~ 320 Gb Memory GPU Memory/GPU Cores Memory Bandwidth Quadro RTX 8000 48 GB 4608 672 GB/sec Tesla V100 v2 32 GB 5120 900 GB/sec Tesla V100 16 GB 5120 900 GB/sec P100 16 GB 3584 732 GB/sec P40 24GB 3840 346 GB/sec K80 24GB 4992 480 GB/sec
  • 22. How Taiwania2 Help us 8x GPU (V100-v2, 32Gb) 768 Gb RAM 2 NUMA nodes 8x GPU (V100-v2, 32Gb) 768 Gb RAM 2 NUMA nodes Storage (6T) Login Node … 4x Infiniband Scheduling:Slurm Comm:OpenMPI & Horovod Framework:Tensorflow
  • 23. Ultra-Patch Workflow 1112 images (positive:557; negative: 555) QuantaGrid D52G nodes on Taiwania2 8 Tesla V100 (32gb) and 768 Gb system memory per node With batch size = 1 ~320 Gb system memory is used for training through Unified Memory Training:0.0067 Image/sec. (each update takes 2.5 mins) Inferencing:~20 sec/WSI Hardware Data Parameter Time
  • 24. Ultra-Patch Method - Result • Compared to the two-stages model, our Ultra-Patch Method with 10k inputs got a competitive result • Compared to different input-size, larger inputs generally got better results
  • 25. Ultra-Patch Method - Result • Compared to the two-stages model, our Ultra-Patch Method with 10k inputs got a competitive result • Compared to different input-size, larger inputs generally got better results
  • 28. • Unified Memory • Transfer cost between CPU-GPU • Group-Execution and Group-Prefetching Optimization for training speed Load Tensors Do Compute Load Tensors … … Load Tensors Do Compute Load Tensors Load Tensors Do Compute … Typical workflow With Prefetching
  • 29. Optimization for training speed Typical workflow Optimized workflow GPU Page Faults Reduced
  • 30. • Unified Memory • Transfer cost between CPU-GPU • Group-Execution and Group-Prefetching • Parallel Training • Distributed training through Horovod Optimization for training speed Load Tensors Do Compute Load Tensors … … Load Tensors Do Compute Load Tensors Load Tensors Do Compute … Typical workflow With Prefetching
  • 31. Ultra-Patch workflow 1112 images (positive:557; negative: 555) QuantaGrid D52G nodes on Taiwania2 8 Tesla V100 (32gb) and 768 Gb system memory per node With batch size = 4 (mixed-precision) ~700 Gb system memory is used for training through Unified Memory Training:1.06 Image/sec Inferencing:~20 sec/WSI Hardware Data Parameter Time
  • 32. • ~5.11x acceleration on a single GPU optimization. • ~147.28x acceleration on 8 nodes • ~498x acceleration on 32 nodes Experiments & Optimization Results 8 nodes 1 nodes Data-size comparison • More data, better performance.
  • 33. Impact of Ultra-Patch On Medical Image Analysis • Speed up the developing time for EACH cancer classifier/detector • Laborious annotation tasks will be reduced • Results will be available faster
  • 34.
  • 35. • 國家高速網路中心提供台灣杉二號運算環境與 HPC 技術支援 • 林口長庚醫院提供訓練與測試資料集 Acknowledgement