Tencent partners London tech outfit to improve Parkinson's treatment

London

ONE of China's tech mammoths is tapping into the London digital hub for help with using artificial intelligence (AI) to improve Parkinson's disease treatment.

Tencent Holdings Ltd is teaming with closely held Medopad Ltd to develop a system to remotely monitor patients with the severe movement disorder. Their goal, according to Wei Fan, executive director for medical AI research at Shenzhen-based Tencent, is to allow doctors to set drug doses and modify care without patients coming into the hospital.

Healthcare is increasingly "not only about what happens in the clinic, but what happens in the real world", Mr Fan said in an interview before the announcement. "The reason we want to work with Medopad is they're experts in monitoring patients."

Medopad, which was founded in 2011 and has received funding from German pharmaceuticals firm Bayer AG, said in February that it had signed deals worth £100 million (S$180 million) with a number of Chinese companies, including Tencent.

Tencent is shopping around the world for partners in revolutionising medicine. The maker of the WeChat social media and payment app has already forged agreements with drug giants Novartis AG and AstraZeneca plc to help address product quality and spot dangerous counterfeits. It's backing WeDoctor, the US$6 billion AI-enabled startup that helps detect disease by analysing streams of patient data, and Blue, the insurer offering health policies via apps.

The collaboration with Medopad, along with Ray Chaudhuri, director of the Parkinson's Center of Excellence at King's College Hospital in London, is another tech effort to give doctors meaningful information about patients who may be hundreds of miles away, cutting out cumbersome travel and clinic scheduling. Medopad has developed mobile apps for monitoring patients with diabetes, cancer, heart disease and rare diseases.

For Parkinson's patients, Medopad has developed apps to allow doctors to replicate strength and movement tests that they normally observe in person. The centrepiece is a video test of fine hand movements - opening and closing the patients' fists - that Medopad analyses digitally for range and speed.

The AI software that Medopad is developing will eventually be taught to automatically score these tests. For the moment, though, the technology is primarily used to translate the video images from a patient's smartphone into a graph that doctors can then use to evaluate the patient's performance.

Trials are underway to study whether the approach can show whether patients' symptoms are worsening or improving, allowing doctors to tweak treatment doses, said Dan Vahdat, Medopad's chief executive officer. Ideally, it would help prevent severe complications by ensuring that changes in disease status are addressed promptly, Mr Vahdat said. "We can raise an alert," he said in an interview. "It all happens through your phone, no matter where you are. You get flagged and your doctors say you should come in."

Mr Fan and Mr Vahdat envision the same remote technology being applied for childhood brain cancers, where young patients' movements are studied to make sure that medications aren't harming mobility and dexterity, and management of knee surgeries. Similar systems could be used to watch children for signs of developing scoliosis, Mr Fan said.

Computer vision, a type of AI that can analyse images, is making rapid inroads into healthcare, with dozens of companies working on AI software that can read X-rays, mammograms, CT scans and other sorts of medical imaging. In addition to its work with Medopad, Tencent's AI medical research group has several other projects under way using computer vision, Mr Fan said, including the creation of software that can diagnose skin cancer from a photograph taken with a mobile phone.

Most AI software requires large quantities of data to train. And China, with its massive population and fewer qualms about data privacy than in Europe or the US, has been pushing aggressively to use data on that population to train AI algorithms.

Some medical researchers have raised concerns that AI software trained exclusively on Chinese patients may not be accurate for treating those from other ethnic or demographic backgrounds. Tencent and Medopad want to conduct a "multicultural, multicentre" test of their software, Mr Fan said, working with hospitals in the US, New Zealand and China. A large-scale clinical trial of the Parkinson's product would begin later this year, he said. BLOOMBERG