Seminar: “Biomedisa AI: Automated Segmentation of Micro-CT Image Data using the Biomedisa Online Platform” by Dr. Philipp David Lösel

Dr. Philipp David Lösel, The Australian National University (ANU), Canberra, Australia

Biomedisa AI: Automated Segmentation of Micro-CT Image Data using the Biomedisa Online Platform

Analysing volumetric biological or geological image data, such as from X-ray micro-tomography, often requires isolating specific structures from the 3D volume through segmentation. Manual or semi-automated segmentation by experts is a common approach, but it is time-consuming and limits large-scale analysis of morphological data.
Biomedisa (https://biomedisa.info) is an online segmentation platform designed for effortless accessibility via web browsers, eliminating the need for complex configurations. Tailored for scientists with diverse expertise levels, it accommodates those without in-depth computer or software knowledge. Additionally, it can also be installed locally for enhanced versatility and offers seamless integration into custom Python projects.
While Biomedisa's smart interpolation can aid in generating training data for subsequent machine learning or segmenting individual samples [1], Biomedisa's deep learning capabilities enables automated segmentation, facilitating extensive analysis of distinct objects within the volume, such as insect brains, and can provide insights into animal behaviour, ecology, and evolution [2]. Notably, Biomedisa's functionality extends to segmenting repetitive structures in large volumes, like cells in biological samples, rock particles, or roots in soil, without requiring image resizing. Here, we demonstrate Biomedisa's patch-based deep learning approach to segment distinct objects, including 187 bee brains from 3D image data and repetitive structures in volumes of up to 2,500 x 2,500 x 10,000 voxels. Our approach successfully identifies plant roots in fertilised soil and separates hundreds of thousands of particles from a crushed rock scan by predicting particle boundaries and separating connected regions.

References
1.    Lösel, P.D. et al. Introducing Biomedisa as an open-source online platform for biomedical image segmentation. Nat Commun 11, 5577 (2020).
2.    Lösel, P.D. et al. Natural variability in bee brain size and symmetry revealed by micro-CT imaging and deep learning. PLoS Comput. Biol. 19, e1011529 (2023).