The widespread use of Deep Learning (DL) models showcase unprecedent performance on a variety of complex tasks, achieving previously unattainable predictive accuracies at human and even superhuman levels. In particular, Computer Vision research has been dominated by the extensive use and exploration of DL. Moreover, the recent adoption of whole slide scanners by the pathologist's community has enabled the digitization of glass tissue section slides into whole slide images (WSIs). These are multi-gigabyte images with typical resolutions of 100000 x 100000 that, when coupled with visualization techniques, can provide invaluable information on the nature of the underlying tumour microenvironment. However, the use of DL on WSIs is challenging as the typical input images in DL models rarely exceed resolutions of 1000 x1000 due to the exponential increase in computational time. Therefore, conventional DL methods are insufficient in exploiting WSIs. In addition, multiple WSIs are captured from different regions of each patient's tissue. Considering the heterogeneity of the tumour microenvironment, all WSIs of a patient might need to be assessed for an accurate prognosis. This PhD project aims to address these challenges to improve the applicability of DL on large medical images and thereby facilitate accurate end-to-end personalized prognosis. The primary focus will be on Colorectal Cancer WSIs stained with Haematoxylin & Eosin. Nevertheless, the work should be generalizable to other types of cancer and other visualization techniques such as immunohistochemistry and immunofluorescence.

Doctoral Training Partnerships: a range of postgraduate training is funded by the Research Councils. For information on current funding routes, see the common terminology at https://www.ukri.org/apply-for-funding/how-we-fund-studentships/. Training grants may be to one organisation or to a consortia of research organisations. This portal will show the lead organisation only.

Abstracts are not currently available in GtR for all funded research. This is normally because the abstract was not required at the time of proposal submission, but may be because it included sensitive information such as personal details.

I'll be applying novel techniques involving tensor networks to improve the computational efficiency in simulation of open quantum systems involving coupled qubits connected to a Non-Markovian bath. The effect of such baths on the system is dependent on what the system has dissipated into them and therefore the former states of these baths. Thus when simulating these systems one has to store a certain number of past states of the system to accurately predict the next step in the evolution. This typically requires large amounts of memory and I'll be working on techniques that look to drastically reduce this demand. This will see me applying code already written to systems already studied and then adapting it to be more versatile in it's application to potentially more useful systems. These systems are relevant to the fast evolving area of quantum computing and any hopes of application of such systems will be reliant on good knowledge of how they evolve.