A Fiji pipeline to segment 3D objects and retrieve shape parameters in biomedical images

Biologists often lack image processing tools that are easy to understand and to adjust to their specific problem. With that in mind, we established a detailed modus operandi, based on free open-source programs. It involves a trainable algorithm: this kind of algorithms are best suited to extract elements that are not easily distinguishable (including non-specific signal intensity, varying textures, etc.).

In our case, it was designed to assess the morphology of amyloid-β plaques in transgenic mice, which are rodent models for Alzheimer’s disease. The images were acquired as X-Ray Phase Contrast Tomography (XPCT) on synchrotron beamlines. Our Fiji pipeline was built with the following plugins: segmentation editor (to isolate hippocampus), trainable WEKA segmentation 3D (to identify plaques), MorpholibJ and 3D ImageJ suite (to label objects and extract relevant shape parameters). Since these plugins were not developed for this specific application, the present pipeline is likely to be well-suited for any morphometric analysis of small 3D objects.