Skip to content

Day4.md

Latest commit

 

History

History
103 lines (55 loc) · 5.77 KB

Day4.md

File metadata and controls

103 lines (55 loc) · 5.77 KB

This page collects instructions, links and materials for the Day 4

Pre-workshop tasks:

1. Install Fiji and plugins

  • Make sure you have Fiji installed
  • Activate LOCI update site

2. Make sure you have a Google account

  • Workshop materials may occupy several GBs. If you don't have enough space on your google drive consider setting up a new Google account for the workshop.

3. Download the workshop materials

  • Download the workshop materials form here -> Download workshop materials
  • Upload the materials to your google drive, keep the same folder structure

Topic 1: Deep learning using BiaPy

Contents:

  • What is Deep Learning?
  • What is BiaPy?
  • How can you use BiaPy?
  • Quality control in deep learning: From dataset validation to model performance monitoring.
  • Tools available in BiaPy for segmentation, denoising, and super-resolution.
  • Example of deep learning in microscopy: Analyzing cancer cell behavior in microfluidics.
  • By the end of the lecture, you’ll have a basic understanding of deep learning and how to train an instance segmentation model using BiaPy.

Instructor

Joanna Pylvänäinen, Åbo Akademi University. Turku, Finland, joanna.pylvanainen@abo.fi

Session 1: Lecture: Deep Learning for Microscopy Image Analysis using BiaPy

In this lecture, we will explore BiaPy, a powerful and accessible toolbox designed to help train deep learning models for microscopy image analysis. We will start with an introduction to deep learning and its applications in microscopy, followed by a detailed discussion of how BiaPy simplifies the process of building and applying neural networks, even for those with limited computational resources.

Lecture here

Session 2: Hands-on: Preparing, Training, and Evaluating a Deep Learning Model for Instance Segmentation in Microscopy

In this comprehensive hands-on session, you will gain practical experience with the full workflow of training a deep learning model for instance segmentation tasks in microscopy, using the BiaPy platform. This session integrates two crucial stages of deep learning workflows: 1) data preparation and training and 2) model evaluation and deployment.

Workshop walk-through here
BiaPy platform
Download workshop materials

Part1: Preparing Data and Training an Instance Segmentation Model

You will begin by learning how to prepare your data for deep learning, which includes annotating microscopy images and setting up a training dataset suitable for instance segmentation. Through guided practice, you will learn to:

  • Annotate images effectively for segmentation tasks.

  • Organize and format your dataset for model training.

  • Train a deep learning model to segment cells in microscopy images.

By the end of this segment, you will have hands-on experience in developing a deep learning model from scratch and applying it to microscopy data.

Part 2: Evaluating Model Quality and Applying It to New Data

Once your model is trained, you will move on to the next phase: Quality Control (QC). Ensuring that your model is robust and generalizes well to new data is essential in any deep learning project. This part of the session will cover:

  • The importance of QC in DL-based segmentation.

  • Techniques to evaluate the model’s performance using training history and validation results.

  • Visualization and inspection of model outputs for quality assessment.

  • Deployment of the trained model to segment new, unseen datasets.

By the end of this segment, you’ll be equipped with the knowledge and tools to conduct thorough QC and confidently apply your model to real-world microscopy data.

Link to Quality control notebook Open In Colab

Topic 2: Tracking with TrackMate I

This course is aimed at beginners in bioimage analysis, that wants to learn how to analyse time-lapse movies. Such movies can follow the movement of cells, in 2D or 3D over time, or the motility of organelles and large objects. They offer unique insight into the dynamics of biology objects, on the evolution of their morphology and of molecular processes, imaged via transmission light or fluorescence microscopy.

The course starts a brief theoretical introduction about tracking algorithms, and is followed by a hands-on tutorial. For the tutorial we will use TrackMate (https://imagej.net/plugins/trackmate/), which is a somewhat popular Fiji plugin for tracking, developed in the Image Analysis Hub.

Preparing for the practicals

If it is not done already, download and install a fresh Fiji: https://fiji.sc/

For the practicals of today, we will learn how to use TrackMate with demo images. Please download them from here in advance:

https://dl.pasteur.fr/fop/QIc4l6FH/PasteurNEUBIAScourse-TrackMateTutorialsMaterias.zip (165 MB)

Screenshot 2025-04-30 at 17 09 13

The zipped folder contains the demo files required for the practicals, and PDFs of the TrackMate manuals (you don't have to read them!!). You can also follow the documentation online: https://imagej.net/plugins/trackmate/ (not required for the course!!)