McCabe, P (2022) Diagnostic and Predictive Modelling in Osteoarthritis using Statistical and Machine Learning Tools. Doctoral thesis, Liverpool John Moores University.

Preview

Text 2021mccabephd.pdf - Published Version Available under License Creative Commons Attribution Non-commercial. Download (6MB) | Preview

Abstract

Osteoarthritis (OA) is a degenerative bone disease that affects joints. OA is one of the most common diseases affecting people in old age. Between 12% and 30% of over 65s are affected by OA, with the knees being the most commonly affected joint. The process for making a diagnosis of knee osteoarthritis is time consuming and somewhat subjective. Clinicians assess a variety of clinical symptoms and information and establish if the patient meets the criteria for having the disease. The utilisation of a machine learning tool could potentially enhance the experience of patients in a clinical setting by reducing the amount of testing required to arrive at a firm diagnosis.
In clinical settings where patient education and behaviour modification are at the forefront, interpretable models are key, as it is vital to be able to explain a decision that leads to any course of action related to an individual patient. In Chapter 3, a model that could be used to aid clinicians in making a diagnosis is developed, and Chapter 4 a model to identify risk cohorts of people who do not yet have the disease is described. Chapter 5 takes those models and uses a different dataset to validate them and develop interactive web-based applications that have easy to explain results.
These models are expanded to consider the effect of gender in presentation of knee osteoarthritis and how this can influence the likelihood of presenting with the disease. Also, the use of multitask learning aims to describe the usefulness of combining datasets to enhance model performance.
Together, these models and approaches utilise both clinical and demographic features to help identify those with knee osteoarthritis and those who are at risk of developing the disease in a five-year timeframe. The models and apps have potential use in clinical settings both as a decision support tool and as a resource for patient education following UK validation of the model.

Item Type:	Thesis (Doctoral)
Uncontrolled Keywords:	Machine learning; Survival Analysis; Data Science; Applied Mathematics
Subjects:	Q Science > QA Mathematics > QA75 Electronic computers. Computer science R Medicine > R Medicine (General) T Technology > T Technology (General) T Technology > TA Engineering (General). Civil engineering (General)
Divisions:	Computer Science & Mathematics
SWORD Depositor:	A Symplectic
Date Deposited:	30 Sep 2022 08:51
Last Modified:	01 Oct 2023 00:50
DOI or ID number:	10.24377/LJMU.t.00017656
Supervisors:	Lisboa, P, Olier-Caparroso, I and Baltzopoulos, V
URI:	https://researchonline.ljmu.ac.uk/id/eprint/17656

View Item