Enhancing Aortic Valve Diameter Prediction: Accounting for Demographic Variability and Measurement Techniques

Dijkshoorn, Leah (2024) Enhancing Aortic Valve Diameter Prediction: Accounting for Demographic Variability and Measurement Techniques. Master's Thesis / Essay, Mathematics.

Preview	Text mMATH2024DijkshoornL.pdf Download (2MB) | Preview
	Text toestemming.pdf Restricted to Registered users only Download (130kB)

Abstract

Leveraging an extensive and diverse donor dataset from Cryolife Inc., previously unobservable patterns are analysed. This donor dataset is made up of physical AV diameter measurements, whereas diagnosis for aortic stenosis is done via echocardiographic AV diameter estimates. Thus, to address potential discrepancies between physical and echocardiographic measurements, a supplementary dataset from the University Medical Center Groningen (UMCG) was used for adults and models from existing literature were used for those 18 and under. The recent acquisition of the Lopez et al. dataset - which encompasses only those 18 and under - allowed for retroactive validation and a robust analysis of measurement biases. Following the data exploration of the donor dataset, an unexplainable trend was observed in the AV diameter measurements over time. Consequently, a bespoke segment neighbourhood algorithm was developed to objectively identify changepoints in the residuals. The trends between these changepoints were then corrected in the AV diameter measurements. It was found that the most suitable model to predict AV diameter was a generalised additive model (GAM) including tensor product smooth interaction terms. To account for heteroscedasticity due to the large demographic variability, a GAM was created to model the conditional standard deviation with respect to the demographic attributes. Combined, these models can be used for Z-score computation, as is standard in the cardiology field.

Item Type:	Thesis (Master's Thesis / Essay)
Supervisor name:	Grzegorczyk, M.A. and Lunter, G.A.
Degree programme:	Mathematics
Thesis type:	Master's Thesis / Essay
Language:	English
Date Deposited:	06 Aug 2024 06:59
Last Modified:	06 Aug 2024 06:59
URI:	https://fse.studenttheses.ub.rug.nl/id/eprint/33848

Actions (login required)

View Item