Supervised by Professor Gill Dobbie, Dr Vithya Yogarajan, and Professor John Windsor (Faculty of Medical and Health Sciences, the University of Auckland), this project seeks to leverage machine learning for healthcare.
Title: Machine Learning to Predict Severe Acute Pancreatitis
Summary:
Acute Pancreatitis (AP), an inflammatory disorder of the pancreas associated with substantial morbidity and mortality, is one of the most common gastrointestinal causes of hospital admission in the USA. In New Zealand, AP continues to have a high incidence rate, with Māori patients reporting the highest incidence of AP. Early identification of patients who are at a greater risk of developing a severe course of the disease is crucial to reducing the risk of adverse disease outcomes and death. Currently, clinical scoring systems are used by health professionals in hospitals to quantify the severity of AP. However, such scoring systems have limitations: these require several variables which are not always easily accessible or, in some cases, need 2-3 days for evaluation. Research has been conducted utilising machine learning in predicting the severity of AP, positing the potential usefulness of models compared to clinical scoring systems.
This research aims to explore the validity of machine learning in predicting the severity of AP using routinely collected patient data. Models such as Support Vector Machine, Random Forest, and XGBoost are trained, tuned, and utilised. This research shows that the accuracy of predicted results from multiple machine learning algorithms is consistently better than those of the clinical scoring systems.
About Natalie: a 4th year student pursuing the Bachelor of Engineering (Honours) and Arts, Engineering Science and Asian Studies. More info: Linkedin.
Photo of Natalie