Google DeepMind teams up with London hospitals to put machine learning to work against head and neck cancers
Google’s machine learning subsidiary DeepMind has kicked off a new research partnership with the radiotherapy department at the University College London Hospitals NHS Foundation Trust, a provider organization that specializes in cancer treatment.
DeepMind and clinicians in UCLH’s radiotherapy team are exploring whether machine learning methods can reduce the amount of time it takes to plan radiotherapy treatment for cancers of the head and neck.
To that end, 1 in 75 men and 1 in 150 women will be diagnosed with oral cancer during their lifetime, and oral cavity cancer has risen by 92 percent since the 1970s, DeepMind said. Head and neck cancer in general affects more than 11,000 patients in the U.K. alone each year, the firm added.
[Also: AI, cognitive computing, machine learning are coming to healthcare: Time to invest?]
“Advances in treatment such as radiotherapy have improved survival rates, but because of the high number of delicate structures concentrated in this area of the body, clinicians have to plan treatment extremely carefully to ensure none of the vital nerves or organs are damaged,” DeepMind said. “That makes a cancer at the back of the mouth or in the sinuses, for example, particularly hard to treat with radiotherapy.”
Before radiotherapy can be administered, clinicians must create a detailed map of the areas of the body to be treated, as well as the areas to avoid. This process is called segmentation, and involves drawing around different parts of the anatomy and feeding this information to a radiotherapy machine, which then can target cancers while avoiding healthy tissue. When a tumor and vital anatomical structures are found in extremely close proximity, as in the head and neck, these outlines that clinicians must create have to be made with tremendous detail. For these cancers, segmentation can take around four hours.
“Even though UCLH’s specialist team at its dedicated head and neck cancer center is a national leader in this process, there is still potential for innovation,” DeepMind said. “We think machine learning could make a difference. Our collaboration will see us carefully analyze anonymized scans from up to seven hundred former patients at UCLH to determine the potential for machine learning to make radiotherapy planning more efficient. Clinicians will remain responsible for deciding radiotherapy treatment plans, but it is hoped that the segmentation process could be reduced from up to four hours to around an hour.”
Ultimately, DeepMind and UCLH hope the research could lead to two benefits: freeing up clinicians’ time to focus more on patient care, education and research; and developing a radiotherapy segmentation algorithm that can potentially be applied to other areas of the body.
“We will treat the patient data we are using in this project with the utmost care and respect,” DeepMind said. “All scans will be anonymized in line with the UCLH Information Governance policy before they are shared with DeepMind. This kind of research is still exploratory, but we think it has great potential to help both clinicians and patients.”
Twitter: @SiwickiHealthIT
Email the writer: bill.siwicki@himssmedia.com
