Will AI In Healthcare Work?

There are sceptics, but the rollout is slowly beginning


As in almost every industry imaginable, artificial intelligence (AI) is set to make a considerable mark. Its introduction is almost certain, but the degree to which machine learning will be accepted remains unclear. Automation is seeping its way into the public consciousness but, despite an almost guaranteed level of public acceptance, healthcare is a sensitive area. And Google, champions of AI, is attempting to roll out its automated technology into the UK's top hospitals.

Working alongside the NHS, Google plan to utilize an app designed to alert staff to patients at serious risk of complications due to kidney failure - Stream. Co-designed by Google's DeepMind, Royal Free Hospital London and engineers at digital product studio ustwo, the app is being used to 'present timely information that helps nurses and doctors detect cases of acute kidney injury.' The app almost instantly alerts medical staff to the results of routine blood tests which can reveal kidney problems, a contributor of up to 20% of emergency hospital admissions. Of these cases, 25% are preventable, and Streams aims to give staff the tools to do so.

The app is not a standalone tool, though, and this is key to the integration of automation into hospital work. Google DeepMind recently bought Hark - a task management app intended to move paper-based or pager-reliant hospital work onto mobile, developed by a team at Imperial College London over four years. Pushed by top surgeons Ara Darzi and Dominic King, the app is in early pilot stage at St Mary's Hospital London, where they found that clinicians were able to respond 37% faster than with traditional methods.

Generally, the priority is the treatment of kidney injuries, and Streams can be integrated with Hark to achieve that ends. Hark does not yet utilize artificial intelligence or machine learning to predict which patients are the most 'likely' to become seriously ill and require immediate treatment, but DeepMind claimed 'that may change in the future.' Some are sceptical. After all, the NHS has a history of failed software investments, but the coming together of communication systems and automated diagnoses has the potential to effectively streamline hospital practices. And this is precisely how the technology will have to be introduced; to supplement already existing and functional systems, rather than to overhaul procedures. For AI in healthcare to be utilized safely, it must begin life as a useful add-on rather than a central pillar.

Demis Hassabis, head of DeepMind, has been lauded for his project to build software capable of rivalling the human brain for intelligence. The technology is of course still some way away, but exciting developments have been made. For example, DeepMind has recently been displaying its capabilities by mastering video games. Unlike most game-playing bots, DeepMind had no access to the internal code of the games, rather it observed and acted 'as a human would'. Most recently, it learnt to traverse 3D, randomly generated mazes in a game called Labyrinth, learning a 'reasonable strategy for exploring random 3D mazes using only a visual input', according to its creators.

No such machine learning has been deployed within the NHS, but as the technology is developed new opportunities for the integration of automation will present themselves. Mustafa Suleyman, co-founder of DeepMind, said: 'it's super early. But if we get it right, we can help doctors and nurses spend more time providing care, and less time juggling to-do lists'. Changes are likely to be just as that, more administrative than clinical, at least for now - automated treatment is still some way off. Changes in the near future are unlikely to be revolutions, but additions, such as Hark's ability to pick up relevant information from the electronic patient record, give the staff a more complete picture of the condition and medical history of the patient as they respond.

The involvement of tech giants in healthcare - particularly one that regularly harvests data for commercial gain - will bring with it a host of ethical concerns. Suleyman is adamant that the company are taking adequate precautions in the protection of patient data, though, insisting that 'the hospital will always own and control that data...the data will never leave the UK and it will never be linked or associated to Google accounts, products or services.' Such assurances will be vital if public opinion is to warm to Google and DeepMind's involvement in the healthcare system.

There is no AI or machine learning currently deployed in the NHS. But, with incremental steps both inside and outside of the hospital, there is no reason AI cannot be introduced to supplement, rather than replace, existing practices and help streamline an at times disorganized - but always brilliant - public system. 

Pharma small

Read next:

R&D Innovation Approach In Pharmaceutical Industry