Workshop on AI in microscopy

AI in microscopy

Background: Recent years have given us tremendous advances in methods using artificial intelligence to solve many different tasks: The deep learning model chatGPT [OpenAI 2021] demonstrates how we can interact with computers naturally, and the Segment Anything Model (SAM) [Kirillov March 2023] has shown that most segmentation tasks can be completed given enough examples.

Content: The participants will 1) get a quick overview of image processing methods and their names to provide a vocabulary for large language models; 2) get hands-on experience with free or cheap tools for image analysis. We will start with the well-known QuPath analysis program, then we will investigate Jupyter Notebooks and Python as general tools for image analysis, and finally, we will explore whether large-language models are a shortcut to image analysis in Python; and 3) we will discuss the advantages and caveats of the three approaches and discuss what can be done easily and when to contact a data scientist.

Target group: Microscope life scientists

Format: Alternating lectures and hands-on exercises. Bring your laptop. 

Participation fee includes: Lunch and refreshments.

Venue: The lectures will be held in Room 4A069 (building 4a, ground floor (0), room 69). The Islands Brygge Metro station is 2 minutes from the entrance and the room is right inside the building. The exercise rooms will be in the neighboring rooms.

Program: We have prepared an exciting programs for you with a mixture of lectures and try-it-yourself exercises. Lectures will be in a joint room, and for the exercises, you will be divided into 3 exercise classes.

Time	Activity	Responsible
09.30 - 10.00	Check-in, coffee and morning snacks
10.00 - 10.05	Welcome and introduction to workshop	Sporring, Jon; Hasselholt, Stine
10.05 - 10.50	QuPath: Image processing and object segmentation	Loo, Tricia Yi Jun
10.50 - 11.50	QuPath exercise: Measurements in 2D
11.50 - 12.50	Lunch
12:50 - 13:10	Lecture: The importance of the pipeline	Nørgaard, Martin
13.10 - 13.40	Python: Image processing and object segmentation	Sporring, Jon
13.40 - 14.55	Python exercise: Hands on python with starter script
14.55 - 15.15	Coffee break
15.15 - 15.45	Large language models: ChatGPT for image analysis	Haase, Robert
15.45 - 16.50	ChatGPT exercise: object segmentation and measurements
16.50 - 17.00	Summary and outlook	Hasselholt, Stine

Material: You will be working with 3 tools: QuPath, Python, and chatGPT. To avoid installation time, we ask that you pre-install QuPath and Anaconda (a Python interpreter and editing tools), and that you sign up for chatGPT. The free version will be sufficient. Below is some helpful installation videos and documents we will use during the day:

• QuPath:
  • Download page: QuPath
  • Demo: Installation of QuPath on 64-bit Windows (MP4)
  • Installation support: online installation support(For Windows, Mac, Linux)
  • Dictionary of key image processing terms: Image processing terms
  • Optional tutorial: First Steps in QuPath
• Anaconda and Jupyter Notebook:
  • Download page: Anaconda Python Package
  • Demo: Download demo of installing Anaconda and Jupyter Notebook
  • Installation support: Anaconda installation support (For Windows, Mac, Linux)
  • Dictionary of key Python terms: Python Terms
  • Optional Jupyter Notebook Cheat Sheet: Jupyter notebook cheat sheet
• Suggested reading material:
  • Introduction to bioimage analysis by Pete Bankhead: Introduction to Bioimage analysis, Chapters Introducing images and Processing and analysis
• Slides and exercises (work in progress)
  • QuPath: Google drive folder
    
  • Python: Python Introduction Jupyter notebook exercise material
  • ChatGPT: Slides and exercises

Sponsor: We are happy to be sponsored in part by Danish Data Science Academy.

See also: The workshop is a part of the EMC2024 Conference.