16. Image processing & segmentation for machine learning evaluation


Martin Held

Our group provides electron microscopy and atomic force microscopy to the institute of membrane research. We collect a lot of image data, which is relevant to get a first idea of the membrane morphology and to evaluate the membrane fabrication process. We also perform 3D imaging to compare the real membrane structure to the simulated one, as well as in-situ-swelling measurements to judge the performance pH-responsive membranes.

What is the data science project’s research question? What steps are necessary to prepare the images & the corresponding fabrication data in order to be evaluated with a machine learning algorithm?

What data will be worked on? Electron microscopy images and the corresponding fabrication data

What tasks will this project involve? 

1. Clean up and filter the table with processing parameters
2. Adjust the histograms of the images to be comparable, without destorting the information
3. Segmenting the electronmicroscopic images of the membranes into pores and no-pores
4. If time allows: Evaluation of the fabrication data and images with machine learning

What makes this project interesting to work on?  The first 3 tasks are absolutly straight-forward and set the ground for an evaluation with a machine learning network.
If time remains to the machine learning evaluation: The fabrication parameters and resulting membrane morphology are really difficult to correlate and generations of PhD students have more or less used trial & error to find an optimim membrane morphology. We have collected 1850 comparable images of 20 PhD students and would like to shed some systematic light on the fabrication process.

What is the expected outcome? Contribution to research paper

What infrastructure, programs and tools will be used? Can they be used remotely?  Fabrication data can be provided as an Excel or ASCII table. The images are in tiff format. Ideally, the tasks are solved with open source software such as ImageJ, Python etc. No servers are required.
Access to our image database can be granted via remote access to a computer in our institute, but will rarely be necessary.

What skills are necessary for this project? Data analytics / statistics, Data mining / Machine learning, Visualization

Is the data open source? No

Interested candidates should be at Bachelor level (3+) . Martin Held  is looking for 1 visiting scientist , working on the project.