AI in Healthcare: How can AI help?
WHY IS AI IMPORTANT?
Artificial intelligence (AI) seems to be everywhere, everyone is talking about it. It’s part of the Fourth Revolution. AI allows systems to learn from previous and new experiences and then use this knowledge to perform human tasks. AI is already being used by IT, marketing, finance and customer services. The biggest user to date has been IT (see HBR’s graph; right). But, healthcare is not far behind, an analysis by Accenture found that the AI healthcare market is expected to be worth $6.6 billion by 2021. AI applications could save the US healthcare system up to $150 billion in costs by 2026. What about the world? The number of companies providing AI for healthcare has increased from less than 20 in 2012 to 100 by the end of 2017.
With all the promise of AI, there are a limited number of AI applications currently being used in the clinic, by clinicians and patients. The reasons for this will be discussed in the next post. This article focuses on where AI is likely to adopted and to do what. AI systems can focus on early detection & diagnosis, treatment and prognosis.
WHAT IS AI?
Rather than a detailed explanation of AI a few quotes and two videos help to explain AI in a more easier manner.
McKinsey’s guide to AI and ZD Net’s everything you need to know about AI are great overviews of AI.
John McCarthy, who first coined the term in 1956, defines it as “the science and engineering of making intelligent machines.”
When a machine (a computer) is able to mimic human cognitive functions and involve in learning, problem-solving and decision-making that is called “Artificial Intelligence (AI)”
Merriam Webster’s definition for AI
1: a branch of computer science dealing with the simulation of intelligent behavior in computers
2: the capability of a machine to imitate intelligent human behaviour
HOW CAN AI BE USED IN HEALTHCARE?
DIAGNOSING DISEASES WITH AI
One of the first narrow applications of AI (ANI) has been to assist diagnosis. AI is being developed to help diagnose; skin cancer, diabetic retinopathy, Alzheimer’s disease and many more.
The first AI technology for diagnosis was approved by the FDA in April 2018. IDx-DR is used to detect greater than mild diabetic retinopathy. IDx-DR is a software that uses AI algorithms to analyse images of the eye taken with a retinal camera. The software is held on a cloud-based server; which analyses the retinal scans and will provide one of two results. Either 1) “more than mild diabetic retinopathy detected: refer to an eye care professional” or 2) “negative for more than mild diabetic retinopathy; rescreen in 12 months.” This will prove very beneficial as data suggests that less than 50% of diabetic patients get yearly eye tests.
More technology is being developed and has is in advanced phases of testing. Ultromics have developed a system to diagnose heart diseases more accurately than doctors. Doctors diagnose one in five heart conditions incorrectly, Ultromics is hoping to improve that statistic, it uses echocardiogram for a deeper understanding of the data. Ultromics system was trained using heart scans from 1,000 patients and whether they went on to suffer from heart conditions. The system has been part of multiple trials with the results to be published later this year, but early results suggest it is outperforming human cardiologists. If the system is found to work, it will be rolled out in the NHS later in the year.
Another company, Optellum, has developed a system which can determine if large clumps of cells in the lungs are potentially cancerous or harmless. Current practice requires regular scans over several months to see how the clumps of cells develop. This would allow for patients to be informed early and reduce the number of scans and the anxious visits and waiting time about the big ‘C’ word. In clinical trials, Optellum has been able to detect which clumps are harmless. This would save the NHS money and resources as a lower number of scans performed, and that time could be used to provide care for other patients. Dr Kabir, CSO and CTO of Optellum, recently spoke to the BBC and said that the lung cancer diagnosis system could save £10bn if it was adopted in the US and the European Union.
TREATING DISEASES WITH AI
After diagnosing any disease the next question is what can we do? How quickly can we act? How effective will treatment be?
IBM’s Watson is being tested by the University of Pittsburgh for cancer treatment. It uses patient’s genomic tumour information to create individual treatment plans. The patient’s tumour is compared to more than 10,000 tumour samples in the NIH’s Cancer Genome Atlas. Rather than sticking to the standardised treatment regimes, the treatment is now these can be tailored to each patient. An early adoption of precision medicine. Another startup SOPHiA Genetics is running a similar service with 5 UK hospitals. An early idea of precision medicines is developing steadily, slowly the practice of one hat fits all is being replaced.
Without effective treatment options treatment would be ineffective. It takes more than 10 years, billions of dollars to go from molecule to medicine. AI provides hope to speed this up and make it more cost-effective. Bench Sci has found more than 76 companies using AI for drug discovery. Big pharma has also jumped at the opportunity with various collaborations. Bench Sci has found 18 Big Pharma companies using AI to aid drug discovery.
An extension to drug disovery is find more indications for existing drug candidates. A partnership between Johnson & Johnson with BenevolentAI has the same aims. BenevolentAI develops, manufactures and commercialises a number of novel clinical stage drug candidates. A phase IIB trial has started for a medication to treat sleepiness in Parkinson’s disease patients.
PREDICTION AND PROGNOSIS OF DISEASES USING AI
AI has can also help with prediction and prognosis of diseases:
Predictive analytics — The ability to monitor patients and prevent patient emergencies before they occur, by analyzing data for key indicators
Predictive health trackers — The ability to monitor patients health status using real-time data collection
Google’s AI algorithm featured in The Verge can predict heart disease using a retinal scan. Published in Nature; from a retinal scan the system could tell the person’s age, blood pressure, and if they smoked. Just take a minute to think and imagine; you go for an eye scan (maybe takes 30 seconds) and it can predict your risk of heart disease.
This was then used to calculate the risk of them suffering a major cardiac event such as a heart attack. The system was trained using data from 284,335 patients and validated on two independent datasets of 12,026 and 999 patients. This not only signifies a big step for AI but also a paradigm shift. This algorithm has found a new use of AI, rather just speeding up existing methods, making current ones more effective, AI can be used to analyze existing medical data from a different sources for a different applications.
Another way AI can help to predict and prevent is has been developed by Lumiata. Lumiata has developed a tool called Risk Matrix to analyse patient’s medical records, then by using AI and risk algorithms it can predict an individual’s likely future health. The tool is designed to help insurers and healthcare providers have more confidence in the decisions they make. So far it has the ability to provide predictions from more than 20 major diseases. Predicting the risks and likelihood will give a basis for more personalised care by mitigating the key risks in an individual’s health.
FINAL THOUGHTS
The world of AI and healthcare is only going to keep on growing, it is not feasible to cover each idea and use. CB Insights reported on 106 startups applying AI algorithms to improve healthcare, with the millions of dollars invested in each.
Although these are never going to replace the medical team, they will assist them vastly from diagnosing to treatment to the prediction to prevention. I think we need to be aware of what the future of healthcare is shaping to be so we can prepare our care for the future and be able to offer the best possible solutions. This will also help us to be less overwhelmed by the incorporation of technology.
Originally published on Novate Healthcare’s blog.