An in-depth look at the myriad ways artificial intelligence is changing the industry
Artificial intelligence is poised to change the pharmaceutical and healthcare industries as it looks to streamline, speed-up and improve overall efficiency. It might still be early days but it’s a bandwagon that organisations are quickly jumping on board.
According to a CB Insights report, about 86% of healthcare organisations, life science companies and med tech firms were using artificial intelligence technology in 2016. Big pharma names announcing deals and applications include Bayer, J&J, Merck, Sanofi, Genentech and Pfizer.
Meanwhile, more than 50% of healthcare industry executives anticipate broad-scale adoption of the technology by 2025, a TechEmergence study recently revealed, with nearly half of the respondents noting that chronic conditions will be the initial target. CB Insights estimates that companies will spend on average $54 million on AI projects by 2020, while Frost & Sullivan expects AI to generate savings of more than $150 billion for the healthcare industry by 2025.
“The predictive aspects of AI have been around for a long time but there are broad-scale developments – such as big data, computational power, genomic insights, wearables, integration with electronic health records, data quality and collection – that are driving the application of AI in healthcare,” says Chris Coburn, chief innovation office at Partners HealthCare, a not-for-profit hospital system and physician network in the US, which is exploring a range of AI applications. “These tools offer important opportunities for new discoveries, applications and financial returns so it’s only natural that industry embraces it to provide new patient benefit and economic promise. It’s early days so we have to keep an eye on the hype but the signals are encouraging.”
Although there is no standard or universally agreed definition, artificial intelligence is essentially a set of advanced technologies that allow machines to do very complex tasks that resemble processes associated with human intelligence such as reasoning, learning, adaptation, sensory understanding and interaction. At the moment, most applications are narrow and specific to a particular task or problem.
Types of AI include: machine learning (perhaps the most successful type which involves data analysis to identify patterns and rules); natural language processing (which understands, analyses and interprets human language); expert systems (which emulate the decision-making ability of a human expert); vision perception (such as image recognition); speech recognition (such as speech to text and text to speech), automated planning and scheduling; and robotics to replicate human behaviour.
The potential application of AI in pharma and healthcare is broad with a multitude of benefits – notably the ability to reduce time and increase accuracy and efficiency, which will cut costs. AI also has the potential to provide the industry with more insights and help drive the transformation of value-based healthcare.
One of the most visible areas where AI is being applied in the pharma industry is in drug discovery, where Coburn explains the technology will allow the discovery of new drugs that hadn’t been possible before. Using machine learning algorithms, the technology can analyse large and complex datasets – such as scientific literature, genetic and clinical data – to quickly identify patterns and relationships to discover new targets, drug entities or the repurposing of already marketed drugs to treat other diseases. Pfizer and Genentech are two companies that are using this approach to drug discovery.
This ability promises to cut the time involved in the discovery process. For instance, this was seen during the 2014 West African Ebola epidemic when an AI programme identified two medicines with the potential to reduce Ebola infectivity. The discovery took a day. Under normal non-computerised conditions, this sort of finding would take months or even years.
Drug discovery costs should also reduce as a result of artificial intelligence. Indeed, a study by researchers at Carnegie Mellon University and Albert Ludwig University in Germany, estimates AI could cut the cost of drug discovery by about 70%.
Artificial intelligence is also being used at a more basic research level. Scientists at London’s Institute of Cancer Research, for instance, are using AI to predict how cancers will progress and evolve. According to Dr Andrea Sottoriva, team leader in evolutionary genomics and modelling at the ICR, being able to predict cancer evolution “would give us a chance to stay one step ahead of cancer” and reduce the chance of the tumour developing a resistance to drugs.
As such, the team has developed a new machine learning technique to analyse genetic changes. “Our new AI method analyses cancer DNA data from patients’ tumours by combining AI approaches with Darwinian evolutionary theory, with the aim of identifying hidden sequences of evolutionary steps in patients’ tumours that determine the way in which their disease evolves,” says Sottoriva. “Finding these hidden sequences means that we have the potential of predicting the next evolutionary step when we see a new patient presenting with earlier disease where only part of the sequence is observable. The potential of anticipating cancer’s next step implies that we can design the best treatment for a given patient, intervene early with effective therapies, thus optimising patient care and sparing treatment to those patients who would not benefit from it.”
AI in trials
Besides drug discovery and optimised treatment, artificial intelligence also has potential benefits in clinical trials. “AI has the potential to reduce the cost and time of clinical trials by identifying patients most likely to benefit from the therapy and avoiding ones who will not,” says Coburn. Essentially AI could be used to mine medical records or genomic data to match suitable patients to relevant clinical trials. Figures from various AI programmes suggest more than 50 patients could be found and validated for a biomarker-led clinical trial within ten minutes, a process that traditionally would take six months.
Furthermore, artificial intelligence algorithms could also be used to predict clinical trial success and avoid potential risk, which can efficiently streamline the process. It could also be used, through the analysis of various data sets, to identify potential adverse events in subgroup populations, as well as predict which patients might have a higher chance of dropping out of a trial.
Then there is the development of real-world data through wearable devices and mobile sensors and apps, alongside electronic medical records, which can feed into clinical trial protocols that are powered by AI. This can not only remove the need for patients to travel long distances but can also improve data quality and disease insights, and more efficiently identify compliance issues, treatment efficacy and side effects that are not possible using traditional analytic techniques.
Healthcare goes sci-fi
Healthcare systems like the NHS are also benefiting from the application of artificial intelligence. Coburn says the introduction of AI to healthcare systems will change every element of point-of-care, and for patients with chronic conditions, continuous care through an AI-powered system has the potential to make the system more efficient, reduce hospital admissions and reduce costs, he says.
Examples of how AI is being used include: surgical robots; enhanced imaging, pathology and diagnosis; identifying infection patterns; virtual nurses for patient support; clinical decision-making; and data management.
Various studies show the benefits are profound. An AI-powered robotic procedure on orthopaedic patients, for instance, resulted in five times fewer complications than if a human surgeon did the operation. Meanwhile, a Danish deep-learning AI algorithm that monitored emergency calls could detect cardiac arrests with a 93% success rate compared with human dispatchers who had a success rate of 73%. And a Stanford University AI algorithm was able to accurately diagnose skin cancer based on visual images, matching the performance of dermatologists.
One area in the healthcare system where AI is making a real difference is in imaging. Experts believe that using AI in radiology will improve the speed and accuracy of diagnosis – and may even replace the need for invasive biopsies in some cases. According to an MIT study, a machine learning algorithm could analyse 3D brain scans up to 1,000 times faster than current analytic techniques.
Earlier this year, prime minister Theresa May announced that deploying artificial intelligence in the NHS would prevent more than 20,000 cancer-related deaths each year by 2033 based on the power to cross-reference patient data with national data to spot early-stage cancer. “The development of smart technologies to analyse great quantities of data quickly and with a higher degree of accuracy than is possible by human beings opens up a whole new field of medical research and gives us a new weapon in our armoury to fight against disease,” she said.
The PM’s announcement followed comments by NHS England chief executive Simon Stevens who said AI innovations were in the pipeline. However, the think tank Reform recently described the NHS’ approach to AI application as “piecemeal” and “patchy”, although a response from 30 NHS Trusts to a Freedom of Information request revealed that 43% were investing in AI, namely virtual assistants, speech recognition technology and chatbots. This is a move in the right direction given that a report by former health minister Lord Darzi noted the NHS could save £12.5 billion a year by using AI for repetitive and administrative tasks.
One area that could particularly help the healthcare system is the use of wearables, apps and other mobile devices powered by artificial intelligence, which could extract insights from continuous real-world data. This could include mobile phone cameras to take clinical-quality images remotely, while the technology could also use AI to detect, monitor and manage diseases with ‘smart assistants’ providing feedback on metrics. Digital healthcare consultations could also be powered by artificial intelligence.
The promise of AI is certainly exciting but there are still challenges and uptake has been slow. Frost & Sullivan notes that just 15-20% of end users have been actively using AI to drive real change in the way healthcare is delivered. This is due to a number of strategic, technological and infrastructure challenges that make application difficult. But there are also other challenges – notable among them is the need for accuracy, transparency in how the AI system works and accountability of the results and decisions.
There can be a data bias if the programme is not fed clean and informative data. Meanwhile consent, data privacy and security are a continuing concern, as is ensuring the technology is properly regulated and systems are interoperable. “A key challenge will be ensuring that AI is developed and used in a way that is transparent and compatible with the public interest, whilst stimulating and driving innovation in the sector,” a recent Nuffield Bioethics report noted.
Artificial Intelligence is still in its infancy but there is much promise for the pharmaceutical and healthcare space where AI has the potential to redefine the sector. According to Coburn the insights these systems will reveal will significantly advance multiple areas in understanding disease, improving efficiency and cutting costs. “Insight will increase, which will drive innovation, which in turn will drive further insights,” he says. “Expect a catalytic effect across disciplines in the longer term.”
Katrina Megget is a freelance journalist specialising in the healthcare industry