Predicting CAD Severity Using Patient Symptom Descriptions and BERT

Heeres, Maria (2024) Predicting CAD Severity Using Patient Symptom Descriptions and BERT. Master's Thesis / Essay, Artificial Intelligence.

Preview	Text mAI2024HeeresAM.pdf Download (2MB) | Preview
	Text Toestemming Heeres.pdf Restricted to Registered users only Download (126kB)

Abstract

Coronary Artery Disease is one of the leading causes of death worldwide. With such a common condition it is incredibly important to be able to quickly diagnose and treat any patient suffering from it. One way to identify the severity of a patient’s CAD risk is with a CT scan that looks at the calcium scores of the coronary arteries. These CT scans often take time and resources, so the CONCRETE project seeks to streamline this process by analyzing patient chest pain complaints using different natural language processing models to try and reduce the number of patient that need to be referred for additional testing. In this paper we use the BERTje architecture to try and predict whether a patient has a high or low calcium score based on a patient’s complaints. These complaints would be the same as those a general practitioner would hear when interviewing a patient. Our experiments did not produce a model that could reliably predict CAD severity, nor did it indicate whether any specific questions were more indicative of a high calcium score. We can conclude that either the questions analyzed can not be used with the BERTje model to identify high CAD risk cases, or more likely that the current amount of data that the CONCRETE project has collected is not yet sufficient to fine tune a BERTje model.

Item Type:	Thesis (Master's Thesis / Essay)
Supervisor name:	Cnossen, F.
Degree programme:	Artificial Intelligence
Thesis type:	Master's Thesis / Essay
Language:	English
Date Deposited:	05 Sep 2024 11:16
Last Modified:	05 Sep 2024 11:16
URI:	https://fse.studenttheses.ub.rug.nl/id/eprint/34202

Actions (login required)

View Item