Segmentation of head and neck radiotherapy scans with machine learning

Full title Applying machine learning to automated segmentation of head and neck tumour volumes and organs at risk on medical images including computed tomography (CT) and magnetic resonance imaging (MRI) scans captured for radiotherapy planning.
Research type Research study
IRAS ID 202595
Contact Name Stanislav Nikolov
Contact Email stanislavn@google.com
Sponsor organisation DeepMind Technologies Ltd
Eudract number
ISRCTN number
Clinicaltrials.gov identifier
Additional reference number fields
Research summary Radiotherapy is one of the main ways cancers are treated. Radiation is used to kill cancerous cells and prevent their recurrence. In over 40% of cancers, radiotherapy is used as first line treatment, usually alongside surgery or chemotherapy, with curative intent. Radiotherapy is particularly important in head and neck cancers, where most will be treated with radiotherapy during the course of treatment. In many other cancers radiotherapy has a role in reducing symptoms and controlling pain even if the cancer cannot be treated.

Complex treatment planning is required to ensure enough radiation is given to the tumour while easily damaged areas (known as organs at risk, e.g. the eyes and nerves) are not damaged. It can take radiotherapy experts at least three to four hours to pick out the important areas on planning computed tomography (CT) scans (known as segmentation). This process and the backlog created can result in longer times between diagnosis and treatment.

Cutting edge machine learning algorithms, along with increasing availability of large digital datasets, allows the creation of sophisticated image recognition tools. DeepMind Technologies Ltd is at the forefront of these developments internationally and wishes to apply these algorithms to segmentation of radiotherapy planning via it’s DeepMindHealth division, in collaboration with NHS researchers.

The work will focus on applying novel machine learning algorithms to automatic detection and segmentation of both cancerous head and neck areas and organs at risk in University College London Hospitals NHS Foundation Trust patients. Should the research be successful, implementation of the outcomes would help patients get treatment more quickly and ease pressures on time and resources in radiotherapy units.
REC Name South Central - Oxford C Research Ethics Committee
REC Reference 16/SC/0189
REC Opinion

Favourable Opinion

Favourable Opinion

Date of REC Opinion 6 April 2016