Neurohistology Database

GIF references: All images taken by H Muzart, using equipment and biosamples owned by H Muzart, bought from third-party labs.

Collecting Neurohistology Images

I have strong experience in using microscopy tools ever since school in 2005 (for microbes and body tissues), and in combination with histology, mainly at Pfizer in 2008 and my first 2 years at UCL in 2013-2015, and some visits to the BrainBank at KCL. Histology/neurohistology sprouts from the science of cellular biology, it was an inspiration for the artificial computational models in neural networks I worked on, and also silicon hardware computer chips as a basis for potential future neuromorphic computing.

A tool for digitised histology (SlideSurfer), the use of which was pioneered by one of my advisors (G Campbell, neuronal biology and digital education) at UCL, had been used extensively at universities. Similar to the Google Earth satellite view, one may hope to get that for the human brain using histology samples an 3D models for photorealistic mapping of the brain. Since 2015, neurohistology has inspired my initial visualisation of real biological neural networks, such as the hippocampus, cortical layers, etc, especially with modern-day techniques and staining, and new state-of-the-art imaging techniques. Many of these (well-preserved) samples can now be ordered at a relatively cheap cost online for analysis by anyone.

Since 2016, I have decided to start using available ones and new (non-neural) samples, to try to work on cataloguing a novel set of images, using my own microscope (then later upgraded to a new digital microscope), and to start to document a large database of neurohistology for educational and research purposes. Prior manufacturers/storers I browsed from include AmScope, JiuSion, Swift, eMarth, Amcap. This is still ongoing as more time is needed and money for data storage to get tens of thousands of high-quality images, and I need to sort out a better infrastructure to organise the images. The next step would be to classify using human micro-tasking (e.g. using MTurk/Zooniverse) or AI-based techniques (Matlab tools, Google DCNNs API, etc).