Lung Biomechanical Modelling Driven by Machine Learning Algorithm Towards Effective Lung Cancer Radiation Therapy

点击翻译按钮获取中文摘要

Lung cancer is the most common cause of cancer death in both men and women as its 5-year survival rate is as low as 14%. External Beam Radiation Therapy is widely used for lung cancer treatment. However, it is extremely challenging due to tumor motion and deformation during respiration. To apply sufficiently high radiation dose to tumor to destroy cancer cells while keeping dose of healthy tissue at minimum, radiation therapy systems are designed such that radiation beam targets moving tumor while patients breathes during therapy session. This can be achieved only if the moving tumor position and its varying shape are estimated accurately throughout the session. Since no medical imaging method is capable of harmless visualisation of the tumor during therapy, we pursue another approach which involves estimating the varying tumor position and shape throughout the session. Having this information the radiation beam can be made to follow the tumor position while it is confined to its shape continuously. In this approach, motion data of the visible chest's surface is measured and input in a computer model to estimate the tumor position and shape over time. An effective model to be used for this estimation is the most important elements that we aim to develop in this research. The model will be a computer program developed based on the biomechanics of respiration. It is patient specific (i.e. considers the patient's specific anatomy etc.) to estimate the tumor motion and shape over time using the chest motion data. Another issue with lung radiation therapy is that radiation dose planning maybe associated with significant harm to healthy tissue. Such planning can be improved by accurate identification of lung gas trapping normally coexisting with cancer. This can be achieved using image processing methods that we will develop to identify such regions using patient medical image before planning to have highly concentrated beams through these regions. As such, the primary objective of the proposed research is to develop and rigorously validate computer models of the respiration system using biomechanics to accurately predict lung tumor motion and shape. The model will input motion data of the patient's chest surface that can be measured using optical tracking systems to output the tumor's varying location and shape throughout the radiation therapy procedure. Another objective is accurate identification of gas trapping regions in the lung where only little tissue maybe exposed to radiation within their volume. Such regions can be utilized for effective therapy planning where radiation beam concentration within these regions is maximized. A long term objective of the research is to incorporate these developments into clinical applications where the tumor motion/deformation data is fed to radiation machines with motion compensation capability for optimal therapy outcome. The research is expected to have major impact on health care of lung cancer patients.