Clinical evaluation of Atlas and Deep Learning-based autocontouring for lung cancer

Radiation Therapy is one of the most effective cancer treatments in oncology, but there are many lengthy yet time-consuming steps within this multi-task workflow. Organ-at-risk (OAR) manual contouring is just one of these critical tasks which can take a clinical expert anywhere from thirty minutes to two hours for a complex case.  Contouring OAR means that health organs close to the ‘target’ (i.e. tumor) are spared from radiation, critical to a patient’s longer-term health and outcomes. This paper examines and contrasts quality and time savings for manual contouring with 2 forms of auto-segmentation or autocontouring solutions: Atlas-based methods introduced in 2014 and more advanced Deep Learning methods (a form of artificial intelligence), introduced clinically in 2018. Mirada’s DLCExpertTM was the first clinically approved model. This paper concludes that user-adjusted software-generated contours reduce contouring time for lung OAR, and deep learning contouring showed promising results compared to existing solutions.  Read more here: https://www.sciencedirect.com/science/article/pii/S0167814017327299