Improving the Prediction of Radiation-Induced Taste Loss in Head and Neck Cancer

Neh, Hendrike (2022) Improving the Prediction of Radiation-Induced Taste Loss in Head and Neck Cancer. Master's Thesis / Essay, Biomedical Engineering.

Preview	Text mBME_2022_NehH.pdf Download (1MB) | Preview
	Text toestemming.pdf Restricted to Registered users only Download (128kB)

Abstract

Purpose: Taste loss is a common side effect of head and neck cancer (HNC) radiotherapy treatment, and its prediction is important to increase the health and quality of life of survivors. This project aimed to improve the prediction of late taste loss at six months after treatment compared to previously developed logistic regression normal tissue complication probability (NTCP) models for patients undergoing radiation therapy for HNC by including a tongue mucosa structure in existing conventional NTCP models as well as developing a deep learning based NTCP model. Results: Of 949 included patients with HNC and available endpoint data, 26.5% (n = 252) patients reported moderate-severe taste loss at 6 months post treatment. A univariable analysis showed that the oral cavity mean dose was a more important predictor than the tongue mucosa mean dose. The logistic regression model with the tongue mucosa mean dose (AUC: 0.717; calibration slope: 1.02) did not perform better than the reference model with oral cavity mean dose (AUC: 0.724; calibration slope: 1.01). The DCNN (AUC: 0.682) and rCNN (AUC: 0.684) both performed worse than the reference model (AUC: 0.692) on the test set. Conclusion: This study presented two different NTCP models based on a new tongue mucosa structure and deep learning-based approach to predict taste loss in HNC patients at 6 months post treatment. No improvement to existing models was achieved and more work must be done to optimize the techniques used.

Item Type:	Thesis (Master's Thesis / Essay)
Supervisor name:	Ooijen, P.M.A. van
Degree programme:	Biomedical Engineering
Thesis type:	Master's Thesis / Essay
Language:	English
Date Deposited:	11 Aug 2022 12:43
Last Modified:	11 Aug 2022 12:43
URI:	https://fse.studenttheses.ub.rug.nl/id/eprint/28271

Actions (login required)

View Item