01 st Aug Artificial Intelligence in Healthcare
Artificial Intelligence in Healthcare
The term ‘Artificial Intelligence’ or ‘AI’ may bring to mind science fiction films such as ‘Terminator’ and ‘The Matrix’ in which machines became sentient, enlightened and then turned on their creators. Ignoring the apocalyptic scenarios of Science fiction, AI undoubtedly has many ways that it can be utilised to improve everyday life. With the ability to learn and solve problems the potential uses of AI range from improving artificial limbs to predicting climate change and preventing disasters. AI is already a part of everyday life from ‘Siri’ on the iPhone to the predictive algorithms used on Netflix and Amazon. So, what is the science ‘fact’ behind the current and potential uses of AI in the pharmaceutical industry? The concept of AI being utilised as a tool in medicine is not a new one. An internet search will produce scientific papers dating back to the 1980’s, discussing the use of AI in medical diagnosis. ‘Artificial Intelligence in Medicine’ conferences have been held biennially for more than 30 years. So, what has changed to thrust the topic back into the spotlight? One factor has been the explosion in data. It has been estimated that in 2016 we generated as much data as in the previous history of mankind. So-called ‘Big Data’ goes hand-in-hand with AI. This is because a massive amount of data is required for the computer to learn and computers are now learning for themselves as opposed to being programmed line by line to ‘think’.
AI has a symbiotic relationship with data, as well as learning from it, AI’s strength is also in being able to sort through it rapidly and spot and predict pattern and trends. This is why AI is useful for medicine. One of the hopes for AI is that it will greatly improve the efficiency of the drug discovery process, decreasing both the 15-year timeline and $1 billion plus costs in getting a drug to market. To this end, Sanofi, GSK and Evotec have recently entered into collaboration with Exscentia who specialise in AI driven drug discovery. Exscentia claim to be able to select promising drug candidates within a quarter of the time of traditional approaches using AI design and assessment of molecules for potency, selectivity and other key criteria. From this information, a select number of molecules can be synthesised and the data from testing in vitro entered back into the design cycle allowing it to progress rapidly.
Both Novartis and Pfizer have entered into a collaboration with IBM Watson in the field of Oncology. Novartis hope to use cognitive computing (i.e. AI or ‘machine learning’) to “optimize cancer care and improve patient outcomes”. It is hoped that the analysis of real-world data will lead to better understanding of the expected outcomes of oncology treatments and will allow the doctor to select the best treatment option for the patient. Pfizer hope that using IBM Watson’s cloud-based AI platform to analyse a huge amount of data they will be able to accelerate the identification of new immune-oncology therapies. Pfizer tell us that “The average researcher reads between 200 and 300 articles in a given year, while Watson for Drug Discovery has ingested 25 million Medline abstracts, more than 1 million full-text medical journal articles, 4 million patents and is regularly updated. Watson for Drug Discovery can be augmented with an organisation’s private data such as lab reports and can help researchers look across disparate data sets to surface relationships and reveal hidden patterns through dynamic visualizations.” This is the power of AI- to rapidly assimilate and wade through this vast amount of information- identifying the trends and devising solutions. Utilising this strength, a UK government project has partnered with the US biotech company, Berg, to use their AI platform to study 100,000 genomes of 70,000 people. By linking this data to NHS records the hope is to identify the genetic factors behind rare diseases and the 6 most common types of cancer.
In a rather different use of AI, the Alder Hey children’s hospital in Liverpool has teamed up with IBM Watson to produce an app to reassure patients and to improve the overall experience of the hospital. Children can ask any question to the app and Watson will study the tone as well as the literal content of the question to identify any anxieties of the patient. The app will reassure the patient while providing feedback to doctors so that they can improve the child’s stay.
With Elon Musk, Stephen Hawking and Bill Gates all recently expressing concerns over AI, are further controls or regulation required? We obviously aren’t expecting the embracing of AI by a pharmaceutical company to lead to a nuclear apocalypse as imagined in science fiction films, however there will be issues to resolve. One of these issues is data privacy. The Royal Free NHS Foundation Trust was found to have breached the Data Protection Act by the Information Commissioner’s Office, in supplying Google’s DeepMind personal and identifiable information on approximately 1.6 million patients. As discussed in the academic paper ‘Google DeepMind and healthcare in an age of algorithms’, neither the Information Commissioner’s Office, the Health Research Authority’s Confidential Advisory Group or the Medicines and Healthcare products Regulatory Agency were consulted before the patients’ data was transferred to third party servers. The patients involved were not asked for their consent, consulted or even informed before their data was transferred. Consent was required because for many patients (approximately 5/6th of the 1.6 million affected) it could not be reasonably argued that they were in a ‘direct care’ relationship with DeepMind. The paper also expresses its concerns over the visibility and lack of independent oversight of what the data is being used for once handed over to DeepMind.
In fact, one of the biggest risks of AI is of making ourselves redundant and the consequential loss of jobs. A paper from Oxford University researchers suggested that as much as 47% of US jobs could be automated in the next one to two decades. Recent predictions by Pricewaterhouse Coopers suggested that the figure of job losses was closer to 38% in the US by 2030 and 30% in the UK. Even if these figures are out by a factor of ten we can see that globally many millions of people will be affected by the advance in technology. It is predicted that most of these jobs will be lost in sectors such as transportation and storage, manufacturing, wholesale and retail with only 17% of the job losses predicted to be in the area of healthcare and social work. While the potential to utilise AI is there, as pointed out in the report by Pricewaterhouse Coopers, not all the jobs may be automated for a variety of legal, economic and regulatory reasons.
In conclusion, it is predicted that Artificial Intelligence will have a significant impact on all our lives in the coming years. Much will undoubtedly be positive but there will also be challenges regarding the loss of jobs, concerns over data privacy and whether current regulation is sufficient for a device that can think for itself.