AI-based Analysis of Cardiovascular Disease Risk in CT of Breast Cancer Patients

Abstract

Cardiovascular disease (CVD) is common in breast cancer survivors. Early detection of elevated CVD risk in patients may help with prevention of complications. In this thesis we investigated whether we could automatically identify patients with an elevated CVD risk by using the planning CT scan: a scan that is made before radiotherapy treatment. We further developed a software algorithm that analyses CT scans and measures the amount of coronary artery calcification. A large amount of coronary artery calcification is a marker for elevated CVD risk. After an extensive evaluation of the software, we used it to investigate the risk of CVD in a large group of breast cancer patients. We found a strong relation between the amount of coronary artery calcification and CVD risk. Moreover, certain breast cancer treatments elevate the risk of CVD. This shows that the software algorithm is an effective tool for identifying breast cancer patients with an elevated CVD risk. In the second part of the thesis, we have developed a software algorithm that can automatically segment the heart chambers and coronary arteries. This way we could automatically calculate radiation dose during radiotherapy treatment per cardiac structure. We found that with higher radiation dose to the structures the risk of CVD increased. The developed software and results presented in this thesis can help develop strategies to reduce the burden of CVD in breast cancer patients.