Artificial intelligence could help to plan radiotherapy treatment for head and neck cancer 

13/09/2018 00:00 
An artificial intelligence (AI) system that could help to plan radiotherapy treatment for head and neck cancer patients has shown promising early results.

Clinicians and researchers from DeepMind Health and University College London Hospitals (UCLH) are developing an algorithm which initial testing suggests can perform to a similar standard as radiographers and oncologists but in a fraction of the time.

At present, it can take clinicians up to four hours to manually mark-up cancerous and healthy tissues on CT scans of head and neck cancer patients ahead of radiotherapy.

This process, known as segmentation, is particularly difficult for cancer of the head and neck, such as mouth cancer, because these tumours are situated extremely closely to healthy structures such as the eyes and salivary glands.

But it is essential that scans are marked-up accurately so that the highest radiation dose possible is given to the tumour, while avoiding healthy tissue.

In a non-clinical environment, the algorithm can rapidly segment a wide range of anatomy on a scan, creating the possibility that in future, patients could begin radiotherapy more quickly.

The model also opens up the possibility of adaptive radiotherapy becoming more widely used in clinical practice. Adaptive radiotherapy is the process by which radiotherapy treatment is repeatedly amended as anatomy changes, for example if the tumour shrinks.

At present, the time it takes to manually mark-up scans is a barrier to this type of treatment because to be effective, re-planning needs to happen quickly and regularly – sometimes daily.

Therapeutic Radiographer Kevin Sullivan, who is UCLH’s head of cancer services, said: “It is very early days for this research but the findings so far are really exciting.

“Having an automated system which could reduce part of the radiotherapy planning process from hours to minutes is a potentially game-changing development.

“In future, it could enable us to treat patients more quickly and effectively, with the ultimate goal of improving the outlook for people diagnosed with head and neck cancer.”

Kevin added: “This system will not replace clinicians. We will still need radiographers and doctors to review the results of the algorithm, and physicists will continue to develop radiotherapy treatment plans. But the algorithm will save time by doing some important and time-consuming groundwork.”

The algorithm has been developed using more than 800 scans of 500 former head and neck cancer patients of UCLH who consented to their anonymised data being used for research purposes.

It has also been tested on publicly available, anonymised scans from the US, demonstrating that it can work effectively on new patient groups.

The initial results of this research have been published on the arXiv website.

The next phase of this work will be to test the system in real clinical practice.

Mustafa Suleyman, co-founder and head of applied AI at DeepMind, said: “Worldwide around 550,000 people a year in the UK receive the devastating news that they have cancer in the head and neck. Making sure they receive the right treatment as quickly as possible is not an easy task.

“We began our partnership with the radiotherapy department at UCLH in 2016 in the hope that advanced technologies might help solve this challenge, improve care and ultimately save lives.

“While the initial findings of our research to date are exciting, for our system to have an impact of people diagnosed with cancer, we need to improve and expand it.”

UCLH’s chief executive Professor Marcel Levi said: “UCLH’s partnership with DeepMind Health is one of many pieces of ground-breaking work we are undertaking as part of our ambitious research hospital programme.

“This means embedding clinical research across all of our disciplines, and harnessing the power of data science and artificial intelligence to support clinical decision-making to make services safer, quicker and more effective.”

In September 2019, UCLH’s partnership with DeepMind transferred, or ‘novated’, to Google Health UK. Read more

 Latest news

 Contact details

Communications unit
2nd floor central
250 Euston Road
London NW1 2PG

Media enquiries

Switchboard: 020 3456 7890
Media enquiries: 020 3447 7542 / 020 3447 9506

Out of hours
The normal working hours for the Communications Unit are Monday to Friday 9am – 5pm. The only media enquiries that will be answered outside of these working hours are urgent enquiries and those relating to major incidents. To access the out-of-hours service call switchboard on 0845 155 5000.

Share this story