Modelling Wireless Communication Channels - Oxford 3rd year group design project

From abstract:

It is currently impossible to extract and store the information that a human brain accumulates over one’s lifetime. Our research proposes an extensive pipeline to retrieve the locations of synapses, therefore providing us with the necessary information to map out an individual’s connectome. Ethical issues inevitably arise as this process requires the termination of a human life.

This is necessitated by the need for preservation, to prevent degradation of tissue and prepare for both storage and sectioning. The size of the human brain has required us to develop a new tool for the sectioning pipeline to work alongside those already proven in use throughout the neuroscience community. Further chemical processes will be needed to prepare the sample for imaging, each comes with benefits however the BROPA method is shown to be the best option. A clear record of techniques for high resolution imaging of brain tissue provides a reference for future endeavours and examines both the benefits and failings of each individually. Our current chosen technique is fully assessed for its suitability along with possible future developments and a small discussion with the manufacturer on viability.

Machine learning section - Wojciech Dziwulski

Various data organisation frameworks and querying tools proceed to be presented and contrasted. These are then used to estimate the storage requirement for one human brain. A range of machine learning techniques are described and compared. This justifies the choice of bagged tree ensembles in conjunction with hand-designed features for the image segmentation. The corresponding hardware requirements and running time efficiency are also presented. Finally, a detailed business case will consider the full project cost and sources of investment.

Credits to Alexander Dwornik, John Boyer and Judah Rand - team members.