Agenda

2
2
 MONAI Bootcamp – Day 1 - Agenda
MONAI Introduction - 20 minutes total - 8:05 - 8:25
MONAI Core - 2 hours and 30 minutes - 8:30 am - 11:00 am
• Intro to MONAI Core - 60 minutes
• End-to-End workflow with Hands-on - 60 minutes
• MONAI Bundles and Model Zoo - 30 minutes
Coffee Break - 7 minute
MONAI Label - 60 minutes total - 11:10 am - 12:10 pm
• Intro to MONAI Label - 40 minutes
• 3D Slicer Demo - 20 minutes
Lunch Break - 40 minutes
MONAI Deploy - 2 hours total - 1 pm - 3:10 pm
• TotalSegmentator Teaser - 20 minutes
• MONAI Deploy Overview - 70 minutes
• MONAI Deploy in the Cloud on AWS - 40 minutes
3
 MONAI Bootcamp – Day 2 - Agenda

MONAI and Federated Learning Intro

• Brief Intro to MONAI and Recap of Bootcamp - 5 minutes - 8:05 am - 8:10 am
• Intro to Federated Learning - 20 minutes - 8:10 am - 8:30 am

MONAI + FLARE Technical Overview

• MONAI Federated Learning APIs - 30 minutes - 8:30 am - 9:00 am
• MONAI + NVIDIA FLARE Demo - 30 minutes - 9:00 am - 9:30 am

Real-World Use Cases of Federated Learning

• MONAI and Federated Learning on AWS - 30 minutes - 9:30 am to 10:00 am
• Canon - 30 minutes - 10:00 am - 10:30 am

4
 MONAI Discord
Server Rules

General Chat

Introduce Yourself

Q&A Channel

Leave a voice channel

https://discord.gg/8kxFN4tP
5
 MONAI Bootcamp – Hands-on

https://github.com/Project-MONAI/monai-bootcamp

Three Options:

Clone Repo to your workstation and setup

dependencies

Google Colab

AWS Studio Lab

- Message Michael Zephyr (mzephyr) for code

6
 1st FDA Prostate Cancer
Detection
Paige Prostate Clinical Grade AI
Paige.AI

Data in Healthcare

Medical Object Detection Vision Transformers GNNs for Deformable

nnDetection Localization and UNETR Multi-Organ Registration

PROMISE
Characterization Segmentation SMNet
DKFZ NVIDIA, Vanderbilt Peking University

OF AI IN
IMAGING
Multi-modal Brain Tumor
Vision
Federated Learning Segmentation
Transformers
EXAM COVID-19 Oxygen Prediction UNET | SegResnet | Swin UNETR
20+ Institutions NVIDIA
30% of the World’s Data is from Healthcare

90% of Healthcare Data is from Medical Imaging

7
 Domain Expertise

COLLABORATION

AI IS
NOT
Team of Researchers
Team of Clinicians

A HAMMER. Learning
Rates
Expert
Annotations

We need to bring together the domain experts to help

guide the creation of AI. This means that Researchers
and Clinicians need to have a consistent, easy, and
reproducible way to bring their expertise to the
workflow.

8
 What is
MONAI?

Project MONAI is a
collaborative open-source
initiative built by academic and
industry leaders
for deep learning in healthcare
imaging.

9
 Goal of MONAI

Establish an inclusive community

of AI researchers to develop and
exchange best practices for AI in
healthcare imaging across
academia and enterprise
researchers with the goal of
accelerating the pace of
innovation.
10
 Stephen Aylward Sebastien Ourselin Klaus Maier-Hein Jayashree Kalpathy- Jorge Cardoso Daniel Rubin
Chair of the Advisory Board
Cramer

Kevin Zhou Nassir Navab Andrew Feng Nasir Rajpoot Justin Kirby Keyvan Farahani

MONAI Advisory MONAI brings together the effort to build a common and open foundation. It is mission-critical for
MONAI’s success to be guided by thought leaders in the domain.

Board.

11
 MONAI
Working Groups.
Imaging I/O Federated Learning Community Development

Focus: define how data is read into Focus: Unify the disparate methods of Focus: Establish MONAI as a common
and written out from memory in Federated Learning in a common software foundation that the medical
MONAI. MONAI framework. imaging research and development
community can build upon.

Data Evaluation, Reproducibility, Deploy

and Benchmarking
Focus: Defining support for Focus: Provide the infrastructure and Focus: Close the existing gap from
bioinformatics, biomarkers, and tools for quality-controlled validation research and development to clinical
metadata that are in scope for MONAI. and benchmarking of medical image production environments by bringing
analytics methods. AI models into the medical workflow.

Transformations Research Digital Pathology

Focus: Topics related to data Focus: Establish MONAI as a catalyst Focus: Creating a standard pipeline
preprocessing and augmentation for scientific progress and real-life for preprocessing, analysis, and
modules in MONAI. impact. visualization of pathology images.

12
 MONAI Helping guide MONAI's vision and mission, an Advisory Board and nine working groups, have
been created and are led by thought leaders throughout the medical research community.

Decision These focused working groups allow leaders in those fields to concentrate their efforts and
bring the most impactful contributions to the community. These working groups are open for
anyone to attend and contribute.

Process.
Have an idea? Draft Document Working Group Leads

GitHub discussion MONAI Strategy

Create
proposal
Goes to Communicate with other
WGs

MONAI
Community support
Advisory Board
Big disruptive
idea
Approved by general
consensus

GitHub repo
Present to

Goes to GitHub issues

MONAI
Smaller scope idea Working Group Pull Requests

Approved by 2+
13 institutions
What MONAI Stream

is
Real-Time Inference

MONAI?
MONAI Core
Deep Learning
Training

MONAI Deploy
End-to-End Clinical Deployment
MONAI Label
AI-Assisted Annotation
 00
Data
MONAI Data is the basis for all medical imaging workflows. Whether that’s your data or public data, you need a way to get the
data into the MONAI as quickly as possible.

Workflow. MONAI provides easy access to datasets like the Medical Segmentation Decathlon and MedNIST datasets through
wrapper APIs. MONAI also provides easy methods to load your data with performant libraries for most common medical
image formats.

01 02 03 04
Labeling Training App Development Deployment

MONAI Label is an intelligent open- MONAI is the flagship PyTorch-based MONAI Deploy App SDK enables MONAI Deploy is also building open
source image labeling, and learning library for deep learning in healthcare developers to take an AI model and reference implementations of an
tool that helps researchers and imaging. It provides domain-optimized turn them into AI applications. inference orchestration engine,
clinicians collaborate, create foundational capabilities for informatics gateway, and a workflow
annotated datasets and build AI developing healthcare imaging training MONAI Deploy App SDK v0.4 manager to help drive clinical
models in a standardized MONAI workflows integration.
paradigm.
MONAI Core v1.0 MONAI Inference Service v0.1
MONAI Label v0.5 MONAI Workflow Manager v0.2
MONAI Informatics Gateway v0.2

15
 MONAI gives developers an AI Factory.
MONAI Model Zoo
Cloud
Datasets and
Services
Pre-trained Models

MONAI Deploy

MAP

Accelerate Development Foundation for an Ecosystem

MONAI Label
Create Reproducible Models Access to Research Community
Hospitals
Active Learning
Labeling
Path to Commercialization Accelerating time to Clinical
Impact

MONAI Core Medical

Devices
Model Training

16
 MONAI Stack.

MONAI Model Zoo MONAI Label Samples MONAI Research Clinical workflow spec Customer applications
MONAI Apps

MIC Training pipelines CAI Training Pipelines MONAI Tutorials MONAI Applications (MAPs)

MONAI Label MONAI Core

MONAI Deploy MONAI
Auto Seg | Interactive Segmentation | Continuous Self supervised learning | Auto3D | Federated
MONAI Tools Batch| Stream Model Zoo
Learning Learning

MONAI Core Components

Transforms | Data Writers | Losses & Metrics | Network Architectures | Inferers | Visualization | Interoperability

System PyTorch | NumPy | CUDA | Docker K8s | Argo

Software
Base OS (Ubuntu | Windows | Mac)

Systems
NGC CERTFIED OEMs NVIDIA DGX FAMILY CLOUD CSPs HOLOSCAN DEVKITS

SOTA Research to Clinical Deployments

17
MONAI
Model Zoo.
CT Pathology
Use pre-trained models to
jumpstart AI development
for all organs and disease
types

Modalities

MRI X-Ray Endoscopy

Model Zoo
Contributors.
Segmentation Detection

Applications

Classification Registration Reconstruction

 MONAI
Design Goals.

Customizable Composable Domain Specialized

Abstracted for customizable design for Portable with ease of Integration into 3D Transformations, Network
varying user expertise existing workflows Architectures, and workflows for
Medical Imaging

GPU Optimized Reproducible High Quality

Multi-GPU CUDA accelerated data Built for reproducibility and Tutorials for getting started, robust
and model parallel processing comparison with state of the art validation, and documentation

20
 MONAI Downloads
and
Contributors.

780k+ 20k+ 10k+

MONAI Core MONAI Label MONAI Deploy

150+ 30+
Individual Contributors Institutions
 Partners
and
Growth.
22
Federated Learning
with MONAI.
Federation Participants
Privacy
Local Model Private Data
Preserving
Build collaborative AI with privacy-
preserving NVIDIA FLARE integration

AI Research and Development

Hospitals

MONAI FL
MONAI FL
and
Provisioning Client APIs Academic Medical Centers

Trained Model Federated Server with Federated

customized workflows Workflows

Cancer Treatment Centers

Interoperability Distributed Multi-Party Collaboration Generalizable Models

MONAI Federated learning (FL) client Mitigate risk of compromising data Adapt existing AI workflows to an FL
algorithm APIs can run AI applications security and privacy with built-in paradigm and build generalizable
on any FL platform algorithms and workflows models for any data set
Getting involved in the
MONAI Community.

Start using MONAI! Submit Issues or Bugs Contribute code, features,

or research

How? How? How?

Start with tutorials in whichever You’ve been using MONAI and You’ve submitted a paper or
piece of MONAI fits your current found an issue or think we could trained a model? Start a discuss to
workflow. write a tutorial? Tell us! integrate it into MONAI

Get started with MONAI Contribute to MONAI Commit to MONAI

 MONAI Resources.
● MONAI Website: https://monai.io/

● MONAI Slack: https://forms.gle/QTxJq3hFictp31UM9

● MONAI Docs:

○ MONAI Core: https://docs.monai.io/en/stable/

○ MONAI Label: https://docs.monai.io/projects/label/en/latest/index.html

○ MONAI Deploy App SDK: https://docs.monai.io/projects/monai-deploy-app-sdk/en/latest/

● MONAI Github: https://github.com/Project-MONAI

○ MONAI Core: https://github.com/Project-MONAI/MONAI

○ MONAI Label: https://github.com/Project-MONAI/MONAILabel

○ MONAI Deploy: https://github.com/Project-MONAI/monai-deploy

● MONAI YouTube: https://www.youtube.com/c/Project-MONAI

○ Overview Videos, Deep Dive Series, Bootcamp and other event recordings

● MONAI Twitter: https://twitter.com/ProjectMONAI

○ Follow for the latest announcements

● MONAI Medium: https://monai.medium.com/

○ Read about our latest releases and our upcoming research interview series

25
 Let’s get
started!

26
26