E-Thesis 49 views 37 downloads
Modernising the skin cancer MDT using novel technologies / Stephen Ali
Swansea University Author: Stephen Ali
DOI (Published version): 10.23889/SUthesis.69479
Abstract
Aims Multidisciplinary team meetings (MDTs) are an integral component of contemporary cancer care. However, variations in treatment uptake still exist and as such, MDTs have not been entirely successful in their aim of reducing variation in access to care. There are significant direct and indirect c...
| Published: |
Swansea, Wales, UK
2025
|
|---|---|
| Institution: | Swansea University |
| Degree level: | Doctoral |
| Degree name: | Ph.D |
| Supervisor: | Whitaker, Iain ; Hutchings, Hayley ; Dobbs, Thomas |
| URI: | https://cronfa.swan.ac.uk/Record/cronfa69479 |
| Abstract: |
Aims Multidisciplinary team meetings (MDTs) are an integral component of contemporary cancer care. However, variations in treatment uptake still exist and as such, MDTs have not been entirely successful in their aim of reducing variation in access to care. There are significant direct and indirect costs of MDT working and evidence demonstrates that some MDTs function more effectively than others. Regular meetings to discuss patients also present an opportunity cost to the National Health Service (NHS). The evidence base for cancer treatment is accumulating constantly, with the National Institute for Health and Care Excellence (NICE) regularly updating their advice. Innovative solutions to these problems would have health economic benefit, free up specialist time and improve the reproducibility of evidence-based decision making. It is the aim of this thesis to develop and validate methods of automating primary and secondary functions of the MDT using natural language processing (NLP) techniques. Methods Initial steps included conducting scoping surveys to understand the current state and challenges of conventional and remote skin cancer MDTs. A systematic review and meta-analysis of existing literature on ruled based NLP clinical decision support systems (CDSS) in cancer care were also undertaken to compare the effectiveness of these systems against human clinicians in medical decision-making. Further, development and validation of an NLP-based CDSS for basal cell carcinoma (BCC) were carried out, focusing on providing treatment recommendations post-primary surgical interventions via a virtual MDT (vSMDT) platform. Additionally, an NLP-based information extraction system for BCC was created and validated to improve quality assurance and outcome benchmarking post-surgery. Results Survey responses uncovered prevailing practices and variances within specialist skin cancer multidisciplinary team meetings (SSMDTs) in the UK. It was discovered that only 26.0% of the SSMDTs were quorate by membership, with the major obstacle being the absence of clinical oncology presence. There was also a notable 69.0% achieving quoracy by meeting frequency, revealing considerable discrepancies and emphasising the significant need for standardisation and adherence to NICE quoracy standards. Further, the thesis provides insights into the effectiveness of MDT meetings, revealing a uniform belief in the importance of risk stratification and prioritisation of complex cases. There is a consensus on the need for protocolised treatment pathways and enhancements in meeting preparation and attendance. Evaluation of remote skin cancer MDTs in the post-COVID-19 era illustrated a comparative efficacy in communication and decision-making between virtual and in-person meetings, with a preference for a hybrid format for future interactions, emphasising the necessity for improvements in connectivity and integration. Innovations in automated information extraction in BCC histopathology were explored using an NLP system, which displayed high precision and recall, offering promising applications in improving the quality of cancer registry data. Moreover, the automation of a web-based model of care demonstrated high accuracy comparable with the literature, supporting the feasibility of a fully automated, virtual, web-based service model for hosting the skin MDT. Additionally, advancements like ChatGPT have proven the ability to generate clinically accurate and human-like clinical letters, paving the way for a significant reduction in clinical and administrative workloads and a standardization in the dissemination of patient information. Conclusion The synthesis of the findings in this thesis underscores the transformative potential of advancements in computational methods like NLP in surgical practice. It illustrates the critical role of restructuring MDT discussions, emphasising the importance of tumour-specific guidance and protocolised treatment pathways. The insights derived from this research support the transformative potential of virtual MDTs and AI-driven solutions in refining service provision, clinical communications, and decision-making processes. It also highlights the continuous need for reassessing and revising clinical standards through innovative technologies to achieve optimal excision rates and enhance patient care. Overall, this thesis advocates for the accelerated integration of novel technologies like NLP, with a human in the loop to enable clinicians to give more time to care for patients in the digital era of healthcare. |
|---|---|
| Item Description: |
ORCiD identifier: https://orcid.org/0000-0002-9917-3432 |
| College: |
Faculty of Medicine, Health and Life Sciences |
| Funders: |
Welsh Clinical Academic Training Fellowship, the Paton Masser grant from the British Association of Plastic, Reconstructive, and Aesthetic Surgeons, and the Topol fellowship. The Reconstructive Surgery & Regenerative Medicine Research Centre's pivotal work is funded by The Scar Free Foundation and Health and Care Research Wales |