In a recent article, we talked about the ways in which Natural Language Processing could be an administrative saviour in the healthcare system. However, this technology is just one of the aspects of Medicine and Healthcare that are seeing the rise of AI and robotics promising a major positive impact.
So far, we have seen deep learning algorithms accurately (96%) diagnose tuberculosis from chest x-rays. Google has trained an AI to detect the spread of breast cancer into lymph node tissue on microscopic specimen images. Neural networks have demonstrated that they are capable of competing with human physicians at detecting changes in diabetes from images of a patient’s retina.
These are merely a few examples of how AI is being used in diagnosis. However, one of the most major applications of AI in diagnosis comes via its ability to discover new relationships between data points to establish associations previously invisible to the human eye. The sheer amount of data analysis required to scrape the surface of these relationships make it prohibitive both on a cognitive and a financial level. But deep learning neural networks have access to this information in barely any time at all. Unstructured data in the form of imagery, text, test results, and more besides can be organised, processed, and analysed in unprecedented depth.
The result of this could mean the ability to establish causation, to predict future conditions on an individual basis, to recommend the best treatment and medication based on deeply analysed historical patient records, and to detect conditions and illness from the very earliest stage. The life-saving and cost-reducing benefits will be huge.
UK researchers, for example, recently fed four machine learning algorithms data on 295,000 patients. The aim was to allow the algorithms to correlate medical history with rates of heart attacks. Once the algorithms had processed this data, they were then fed records from a further 82,000 patients in order to predict which ones would have heart attacks. These were patients with a history of heart attacks, but this data was not included in that fed to the algorithms. The algorithm results were compared against factors identified by American College of Cardiology/American Heart Association (ACC/AHA) guidelines which take into account factors such as smoking history, age, cholesterol, and diabetes.
All four of the algorithms performed significantly better than the ACC/AHA guidelines. They correctly predicted 7.6% more events, raising 1.6% fewer alarms. This amounts to no fewer than 355 patients’ lives being saved… by AI.
These examples demonstrate the early stage of research at which we are now. So how much more accurate will deep learning algorithms be in years to come? Technology, as we all know, evolves extremely fast. It’s mind-boggling to wonder quite where we will be with AI in medicine in just five years’ time.
What’s more, as technology evolves, hardware costs fall and processor speeds rise. Soon, the technology will undoubtedly be as, if not more, sophisticated than a human physician. What’s more, it will be affordable. Medical physicians are expensive, understandably, given their level of training and expertise. This means that medical costs are correspondingly expensive, both for governments and medical insurance companies.
In the UK, as we know, the NHS is struggling beyond belief. Cost-saving measures are rife, and patient care suffers. AI has the ability to fix this indefinitely. For those with private medical insurance, and in countries where that is the only option, membership costs may fall considerably. The result? Better health, wellbeing, and longevity for all.
AI Health Assistants
In the UK, getting a doctor’s appointment can be akin to getting an audience with the Pope. Online self-diagnosis can leave one with the distinct sensation of being on the cusp of death. Call the NHS helpline, and the chances are you’ll be told to go see a doctor.
Clearly, we need a solution. And, you guessed it, AI might just be that solution.
One AI-powered mobile app, whose information has been approved by the NHS, may be one tool to combat these issues. The Your.MD app features a chatbot that uses natural language processing and generation, which it uses to ask patients about their symptoms and provide clear, easy-to-understand information regarding their medical condition. Its machine learning algorithms can generate a complex map of the user’s condition and provide a personalised experience. Effectively, it gets to know you. Where necessary, of course, it urges the patient to see a human doctor.
Your.MD is not the only AI health assistant promising to alleviate the strain on the GP surgery. Ada is an AI-powered health assistant that is integrated with the Amazon Alexa. It learns the user’s medical history, generates a detailed symptom assessment report, and provides the option to contact a human doctor.
Babylon Health is another, which also follows up on past symptoms, and can put you in touch with a doctor via video call if need be.
Data and Blockchain
Of course, medical data is incredibly sensitive, as Google DeepMind found out to their detriment when their data-sharing with the NHS became the target of privacy campaigners and UK authorities. So how can we ensure that the data being collected and analysed by artificial intelligence systems is safe?
The answer may lie in the blockchain, which is a distributed ledger which ensures both privacy and transparency of data. Data is stored at each ‘block’ in the chain, only accessible to approved individuals and immune from manipulation. AI systems like Morpheo, which has been used to diagnose sleep disorders, use blockchain to keep patient data safe, private, and secure.
Whilst blockchain is an emergent technology that promises positive disruption to industries as diverse as Banking and Agriculture, as well as Healthcare, of course, there are still some issues to iron out before we can be certain that it is as secure as it needs to be. Nonetheless, as, like AI, it evolves, we can expect vastly improved iterations that will have massive implications.
With ever-increasing computational power and vast reams of incoming data, combined with artificial intelligence technology, the tools are arriving to enable physicians across the full range of medical disciplines to improve patient care. Less expensive care, with faster, more accurate results will have an instrumental impact on the overall health and wellbeing of our society. The only issue will be what to do with the swollen population as everybody starts living longer…
link to “How The Use of Natural Language Processing in Healthcare Can Revive The NHS”
link to “What If We Live Too Long? Life Extension and the Problem of Population“
Link to Full Article: Read Here
