2025-09-12 –, Ballroom 2
How do you reproducibly identify and count individual neurons in a brain region that’s tiny, diffuse, and surrounded by lookalike regions, especially when each 3D brain image is multiple terabytes in size?
This talk explores this very question by diving into the development of a Python-based, end-to-end pipeline for analysing whole mouse brains imaged using light-sheet fluorescence microscopy. The goal is to quantify the number of individual dopaminergic neurons in the substantia nigra pars compacta (SNpc), a small but clinically significant midbrain region implicated in Parkinson’s disease.
Built entirely with open-source Python tools, the workflow combines brainreg (from the BrainGlobe ecosystem) for atlas-based registration, dask for scalable image processing, and a custom-trained Cellpose model for 3D cell segmentation. To address the complexity of region extraction and alignment uncertainty, the pipeline includes parameter sweeps, pre-processing optimisation, and quantitative evaluation using expert-labelled ground truth masks.
This talk will also highlight how the integration of multiple Python open-source packages supports scalable, reproducible neuroimaging analysis, from parallel execution on HPC clusters to image registration and deep learning-based segmentation pipelines, as well as quantitative methods for assessing alignment fidelity.
Recent advancements in microscopy have revolutionised medical research by enabling imaging of whole organs at incredible resolution. In the world of neuroscience, this means that capturing entire mouse brains in 3D at cellular detail is now possible, and with this level of resolution, each dataset can easily reach multiple terabytes in size. While these rich images hold immense potential for understanding brain function and disease, extracting meaningful biological insights (such as counting specific neurons in tiny, diffuse regions like the substantia nigra pars compacta (SNpc)) poses significant computational and analytical challenges.
One of the toughest problems is accurately defining and extracting these small brain regions. The SNpc is surrounded by anatomically similar areas with overlapping cell types, making registration and segmentation highly sensitive to small errors. Misalignment or imprecise region extraction can lead to inaccurate cell counts, which risks compromising the biological conclusions drawn from the data.
To tackle these issues, an end-to-end Python pipeline combining atlas registration with the BrainGlobe ecosystem, scalable image processing using dask, and a custom-trained Cellpose deep learning model for 3D cell segmentation was developed. This workflow incorporates extensive parameter sweeps, preprocessing optimisations, and validation against expert-annotated ground truth masks using image similarity metrics to better understand and quantify the uncertainties in region extraction.
Despite these advances, assessing and improving region accuracy remains an ongoing challenge. This talk will share insights into the complexities of large-scale neuroimaging analysis and highlight that while Python provides a powerful toolkit, the problem of reliable, reproducible extraction of tiny, complex brain regions is far from solved, representing an active area of research and development in the world of bioimage analysis.
Ishrat Zaman is a scientist working in the field of medical research as a bioimage analyst. She is passionate about the intersection of neuroscience, image analysis, and open-source software as well as developing analysis pipelines that can address complex scientific and biological questions. With a background in both wet lab-based biology and computational science, Ishrat is passionate about bridging domains, making complex analysis pipelines more accessible, and looking at as many pretty microscopy images as possible.