< Back to previous page

Project

Benefits of deep learning algorithms for delineation of target volumes in radiotherapy treatment for head and neck cancer.

Head and neck squamous cell carcinoma (HNSCC) is the seventh most common cause of cancer-related death worldwide and is often treated with radio(chemo)therapy. Precise delineation of target volumes (TVs) is necessary for accurate radiotherapy treatment, however delineation is highly experience and perception dependent and therefore prone to human errors. Inadequate TV delineation is proven to be a main cause of local recurrence. A first step in limiting the variability in delineation was recently taken with the publication of international delineation guidelines. However, these guidelines have not yet been validated and implemented in Belgian practice. The main objective of our project is to tackle the issue of delineation variability and errors and to improve prognosis in patients with HNSCC. First, we plan the validation of the delineation guidelines by a retrospective study analysing their effect on the dose-volume histograms and expected toxicity. Next, a 3D convolutional neural network for automated delineation of TVs will be developed with the help of artificial intelligence. We aim to develop and validate state-of-the-art machine learning approaches, more specifically based on deep learning, for auto-delineation of TVs in HNSCC in both the planning CT and daily CT images (cone beam CT). Information acquired by clinical examination and anatomical and functional imaging will be included in the algorithms as well as expert feedback to improve auto-delineation quality.

Date:1 Aug 2020 →  Today
Keywords:Radiotherapy, autodelineation, artificial intelligence, AI, deep learning, head and neck cancer
Disciplines:Cancer therapy
Project type:PhD project