AI powered automated imaging and analysis workflows for EM systems - - PowerPoint PPT Presentation

ai powered automated imaging and analysis workflows for
SMART_READER_LITE
LIVE PREVIEW

AI powered automated imaging and analysis workflows for EM systems - - PowerPoint PPT Presentation

AI powered automated imaging and analysis workflows for EM systems Jrme Hummel, Guilhem Preti, David Tourais, Rmi Blanc, Quentin Lavigne, Eric Goergen* The world leader in serving science Introducing Avizo2D with AI Blurring the line


slide-1
SLIDE 1

The world leader in serving science

Jérôme Hummel, Guilhem Péretié, David Tourais, Rémi Blanc, Quentin Lavigne, Eric Goergen*

AI powered automated imaging and analysis workflows for EM systems

slide-2
SLIDE 2

2

Imaging & Analysis Introducing Avizo2D with AI

Blurring the line between

slide-3
SLIDE 3

3

Avizo2D Software – Overview

  • Utilize a variety of image processing approaches in a single, easy to modify workflow
  • The intuitive UI and modular approach enables ability to evaluate the accuracy of processing steps in real time
  • Attach Avizo 2D directly to your system for on-the-fly analysis
slide-4
SLIDE 4

4

Integrated Solutions To ensure you leave our systems with answers Or, are armed with the best data to ask the right questions

Create a rich, integrated user experience that blends acquisition with advanced analysis

Imaging Analysis

Our Goal

slide-5
SLIDE 5

5

Avizo2D Software – The Platform

Input Image Characterized Output Interactive Workflow Player

  • Access the full power of Avizo’s unmatched processing libraries in a completely redesigned, intuitive UI
  • Build recipes on the fly and visualize the impact of each module choice or setting change in real time
  • Quickly build processing routines for repeated tasks
  • Batch process images for maximum efficiency
  • Optimize Imaging conditions/segmentation while on the tool
slide-6
SLIDE 6

6

Introducing Avizo2D with AI

  • Dedicated tools for image Labeling and Training of deep learning models
  • Easy to use application suite to train custom-designed Python derived networks
  • Any Trained Models can be inserted into analysis workflows (e.g. Keras-Tensor Flow)

Why Artificial Intelligence?

ü AI is well-suited to solving segmentation on traditionally difficult sample types ü Provides a strong framework for objective, repeatable analysis ü Adds access to cutting edge techniques via easy integration with established models

AI in Avizo2D

slide-7
SLIDE 7

7

Avizo2D Software – Deep Learning Models – Step 1: Labeler Cutting-edge technologies wrapped in a user friendly environment

  • Simple tools for feature labeling
  • Quickly populate data with target structures

300 nm

slide-8
SLIDE 8

8

Avizo2D Software – Deep Learning Models – Step 2: Trainer Easy to implement advanced AI workflows

  • Choose base topology
  • Label training set
  • Input data and Ground Truth
  • Set up training parameters
slide-9
SLIDE 9

9

Avizo2D Software – Deep Learning Models – Step 3: Analyzer using DL model

Model trained on a TEM stack of neural tissue Model applied on a Volume Scope stack

  • f a mouse heart muscle

300 nm 2 µm

slide-10
SLIDE 10

10

AI Electron Imaging Denoiser

  • Enhance image quality using AI-powered denoising tools
  • Lower dwell time
  • Quality data on sensitive samples

1 µm 1 µm

slide-11
SLIDE 11

11

  • Go away from the tool with augmented insight
  • Link direct to Avizo2D from the microscope
  • Run image by image to tune results
  • Batch process as images are obtained
  • Connect directly to Maps for full automation

360 µm

slide-12
SLIDE 12

12

Avizo2D Software – The Online Recipe Library

  • Access pre-built workflows
  • Upload and share workflows
  • Access pre-trained models
  • Quickly get up to speed with image processing

approaches

  • Benefit from workflows created by your peers

and application experts

slide-13
SLIDE 13

13

End-to-End Automation System Automation Visualize, Annotate, Share Correlative Microscopy Maps Software

ü Compatible with our full line of Electron Microscopes ü Make the most efficient use of your platform ü Integrated EDS Mapping on SEM and TEM

Automated Large Area imaging Templates for quick set up Automate single to thousands of frames Register and co-visualize imagery Import images from any source Import 3D stacks All imagery in one place Use multi-modal data as navigational aids Annotate and share results with colleagues

slide-14
SLIDE 14

14

Building Efficient Workflows

  • Integrated workflows to decrease time to result
  • Access advanced image processing pipelines while on the tool
  • Tune and optimize imaging and processing workflows
  • Automated, end to end experience
slide-15
SLIDE 15

15

Avizo2D Software – Connectivity with MAPS | Real-time Data Analysis

slide-16
SLIDE 16

16

Classify Au-Ni-Cl Nanoparticles from EDS map

Sample courtesy J. Bursik, Institute of Physics of Materials, Brno

  • Analyzed using pre-built processing recipes
  • Separated by composition
  • Full size and shape statistical extraction
slide-17
SLIDE 17

17

Characterize porosity in separator polymer - Battery

  • Repeatable analysis
  • Simple traditional workflow
  • Live tracking of morphological changes

300 nm

slide-18
SLIDE 18

18

Additive Manufacturing Powder – Statistical Characterization

  • Easily automated workflow
  • Full morphological characterization
  • Compensation for focal depth

5 µm

slide-19
SLIDE 19

19

Fluorescence Microscopy: cell count in well plate

  • Avizo2D can be used in optical imaging workflows
  • Support for import and processing of common optical data outputs
  • Import and batch process from a single folder or image stack

40 µm

slide-20
SLIDE 20

20

Summary

  • Avizo2D is the digital eco-system for AI-powered, workflow-based imaging data analysis
  • Low learning curve environment. No high-end image processing expertise required
  • Unique inter-operability with Thermo Scientific EM acquisition software
  • Solve analysis cases that are challenging for conventional techniques with AI
  • Enables repeatable, standardized analysis approaches

AI Integration Tight System Integration Walk away with ANSWERS