PhD thesis defense to be held on November 22, 2022, at 12:00 (Virtually in MS Teams)

Picture Credit: Vasiliki Iliadou

Thesis title: Radiation therapy image processing and analysis using image registration and machine learning teachnics for patients with head and neck cancer

Abstract: In this Doctoral Thesis the purpose was to analyze head and neck radiotherapy data, using image registration techniques and Machine Learning methods. The aim was to detect serious anatomical changes during patients' radiotherapy treatment sessions, as these could lead to low dosage to tumor and overexposure of surrounding healthy tissues, causing adverse side effects. In this direction, the weekly CBCT images taken during image-guided patient radiotherapy (IGRT) were aligned with the initial planning CT. The alignment was completed, firstly, with affine type transformations followed by elastic transformations. These transformations were used to transform the anatomical structures designed by clinicians to coincide with the patient's anatomical structures on the day of radiotherapy. In this way, the cancer volume and the parotid volume were calculated, as well as the percentage change after the end of each week of treatment, and the patients were divided into those who had severe anatomical changes after the 3rd week of treatment and those who did not present changes. The two patient classes were used to compute predictive models for the possible occurrence of anatomical changes in patients after the first week of treatment. For this purpose, image texture features were calculated for the anatomical structures of the tumor and the end of the first week of treatments. With those image features, a predictive model was constructed, able to accurately predict whether a patient can show severe anatomical changes after the 3rd week of treatments.

Supervisor: Professor George Matsopoulos

PhD Student: Vasiliki Iliadou