Transforming oesophageal cancer diagnosis using optical coherence elastography

1) Brief description of the context of the research including potential impact

Oesophageal cancer is recognised as one of the "unmet need" cancers by Cancer Research UK due to poor survival and late diagnoses. These problems are due to technological deficits along all stages of the medical process. From surveillance to accurate staging, and treatment. The research aims to overcome this problem by combining deep learning, which has revolutionised medical imaging, with Optical Coherence Elastography (OCE), which has already used its almost cell-level resolution imaging of tissue elasticity to push breast cancer research and systemic sclerosis in skin. However, despite combined success of these technologies, their potential in oesophageal cancer is still largely unexplored.

Built on the impressive spatial resolution at which optical coherence tomography (OCT) can image tissue, OCE images tissue to measure the tissue deformation and infer slight differences in the mechanical properties of the tissue and does so with a greater sensitivity than any other imaging modality. However, despite its technical impressiveness and success in some areas, the modality has not yet received a largescale, prospective clinical trail which would be needed before it could become clinically viable.

Despite the short comings, OCE's capacities give reason to expect that the potential impact of this research is substantial, addressing a critical need in healthcare. Early Oesophageal cancer diagnosis is pivotal for effective intervention, and deep learning enhances diagnostic impacts, potentially advancing OCE technologically.

2) Aims and Objectives

- The project aims to customise an existing OCE imaging system for imaging oesophageal biopsy samples, encompassing both hardware and software modifications.

- Evaluate and adapt an existing model-based technique for oesophageal tissue stiffness retrieval, comparing it with a novel deep learning approach.

- Investigate unpaired and unsupervised training schemes, along with generative adversarial methods and transfer learning, during the initial phase of the research.

- Explore the integration of learned and model-based approaches through a developed iterative scheme for stiffness retrieval. Evaluate the system's ability to differentiate between normal and malignant tissue in ex-vivo.

- Conduct a preliminary examination across various tissue types to understand optimal performance scenarios for OCE, including Barrett's oesophagus at several levels of severity

3) Novelty of Research Methodology

The project's novelty lies in enhancing OCE hardware components, tissue stiffness retrieval methods and in applying a OCE-domain deep learning model to oesophageal cancer, where OCE has not yet seen research.

4) Alignment to EPSRC's strategies and research areas

- The research aligns with artificial intelligence, digitisation, and data strategies, aiming to drive value and security through scientific and technical breakthroughs in AI. The potential impact also strengthens the UK's position as a "thought leader in the digital world".

- By providing a new method for oesophageal cancer diagnosis, the research aligns with the goal of transforming health and healthcare. It addresses a major health challenge and aims for a more affordable solution by automating diagnostic workload.

- Positioned at the frontiers of engineering and technology, the project utilises deep learning and OCE to address key societal challenges, contributes to a healthier future.

5) Any companies or collaborators involved

N/A