How to… image analysis

Towards quantifiable results - a teaser to image analysis
images
how-to
Author

Christopher Schmied

Published

September 20, 2023

A guest post by Christopher Schmied, find him on linkedin or social media.

Hidden image analysis

As a Research Data Scientist I engage in many different collaborative scientific projects, particularly focusing on the analysis of microscopy image data (Fig. 1). Turning to the scientific literature, I noticed that image analysis methods were often absent or hidden by only briefly mentioning the used image analysis software. This made reproduction often very difficult.

In publications image analysis is only one of many different methods scientists utilize to produce quantitative results. This scientific skill is increasingly taught in many courses at academic institutions. Also many online resources in different formats exist that help scientists to perform image analysis correctly:

Furthermore, there is also a growing community of scientists specialized in image analysis. You can get directly in touch with this community via image.sc. In this forum you can find many solutions and helpful materials, for instance in the following post you will find a vast collection of training materials for image analysis. However, clear instructions or guidelines how image analysis should be documented in research papers were missing. This is another reason that is making it hard for authors to effectively publish good image analysis methods.

Figure 1: Image Analysis: (A) With image analysis workflows quantitative data is extracted from images. (B) Images are data here intensity information is displayed in different ways. (C) Image segmentation allows to extract information such as objects numbers, object sizes, shapes etc.

Bringing image analysis into the light

Helena Jambor then started a fantastic effort within Quarep-LiMi to improve publications of microscopy images. She brought me in as a Co-chair and together with many different scientists and experts in image analysis and microscopy we were able to distill core aspects of good image analysis methods in publications. We focused on giving practical and easy to execute advice in three different categories of image analysis methods that we commonly see used in research papers (Fig. Publication of image analysis):
  • Established workflows
  • New workflows
  • Machine learning workflows

We hope that these checklist will enhance the work of scientists and make their life easier by giving clear and actionable instructions for the publication of images and image analysis.

Read the Nature Methods paper here
Find the checklist to download as printable files here.
A companion Jupyter Book is available here.

Figure 2: Publication of image analysis: Image data is transformed into results using image analysis workflows. In the literature we see that authors use established analysis workflows, new workflows and machine learning workflows. In our checklists we give essential advice on how to publish these methods.