Topological hierarchies in complex image data

Developing new techniques for understanding complex image data to enable and advance biological discovery


Imaging is a cornerstone of modern biology, and generates huge quantities of rich data. Understanding what this data tells us about the structure and function of the underlying biological system is a challenging and complex task. This project is developing new methods to interpret these complex images, drawing on recent advances in machine learning and computational topology to attempt to understand the sample in terms of its component 'parts' and how these combine together to form more complex structures. Developing an understanding of the biology in these terms could reveal important information about the structure-function relationship with potential long-term applications in drug discovery.

Explaining the science

Biological systems are fundamentally compositional in nature - simplistically we have molecules composed from atoms; organelles from molecules; cells from organelles; tissues from cells. This hierarchical organising principle is reflected in how a sample appears when it is imaged, and complex images are built up from increasingly involved combinations of quite simple building blocks, or parts. This idea has been used in computer vision to learn 'compositional hierarchies' of complex visual scenes.

This project will extend these ideas and combine them with ideas drawn from topology, which provides a formal framework that will allows us to understand how the parts are connected with each other.

Project aims

Modern imaging techniques that are now widely used to understand complex biological samples can generate huge quantities of extraordinarily rich and informative data. The ability to understand what the images tell us about the sample is currently limited by the computational methods that are essential to the analysis of the images.

This project will develop a new approach to analysing these images that will allow us to understand how the complex biological sample - for example a living cell, or a piece of tissue from a biopsy - is composed from simpler parts that combine together in certain ways that control how the sample behaves. The new approach will improve understanding of what the fundamental parts are and how they combine together in different ways under different conditions to give rise to different functions. This understanding is crucial to working out how the parts, their relationships, and the function are connected to each so that researchers can discover and understand the processes that control our bodies.

This could, in turn, allow us to determine what is different between healthy and unhealthy samples and, in the long-term, to develop therapies that specifically target the key processes that cause disease and don't interfere with normal function. 


Advanced imaging techniques are being used in academic biology labs around the world, and in the biotechnology and pharmaceutical industries. This project will develop a new methodological approach that will enable these labs to probe the sample's constituent parts and their interactions in a new way. This could provide valuable new information about the underlying biological mechanisms that are controlling the sample's behaviour. In turn, this could allow them to understand how to control specific mechanisms or interactions that lead to the abnormal behaviour associated with disease.


Contact info

[email protected]