Outlines to outcomes : Second-generation AI auto-contouring for radiotherapy

点击翻译按钮获取中文摘要

External Beam Radiotherapy (RT) is a cornerstone of modern cancer treatment and is offered to 50% of cancer patients in the UK. RT planning involves manually outlining the target to be irradiated and healthy tissues and organs to be avoided on medical images of the patient (such as CT scans). This planning is time-consuming and skill-intensive, and there is inevitable inter- and intra-observer variation. Technology to automatically contour (auto-contouring) organs can reduce variability and improve efficiency of the contouring process. Following on from a previous project funded by InnovateUK, Mirada Medical, as an established developer of innovative software in RT planning, was first-to-market with their CE marked AI-based auto-contouring solution, DLCExpert. With the use of DLCExpert, clinicians can contour organs-at-risk faster and more consistently than with manual contouring reducing the time required for treatment planning and allowing patients to be treated sooner.While substantially better than the previous class of technology, the AI solutions available on the market are first generation and still have room for improvement, with editing of the contours produced still being required in many instances. AI technology could be considered to still be in its infancy. In this project, we seek to further advance the state-of-the-art, pushing towards a second-generation solution which would represent another step change in performance and lead to auto-contouring that requires no manual adjustment, thus providing further time-saving in the cancer treatment pathway.