Please login to the form below

Not currently logged in
Email:
Password:

I, Robot: How AI is redefining the use of data in healthcare

As data moves beyond being simply ‘big’, artificial intelligence is becoming an increasingly influential set of technologies for healthcare
AI robot

From customer service bots, self-driving cars and playing ancient Chinese board games, AI seems to be making breakthroughs in all industries. However despite the huge amount of media interest in this technology, the phrase AI is all too often used as a catch-all term for many different technologies. In order to provide a useful critique of the developments in the healthcare sector, it is important that we take a look at what AI means when referring to different aspects of the industry.

Across healthcare and pharmaceutical development, AI brings with it the possibilities of significant improvements through marginal gains. The key to generating the maximum value of the technology is to identify where real problems lie and where real business opportunity exists.

A prescription for big data
Labelling data as 'big' doesn't really stand up to scrutiny anymore. When the first commentators were discussing the phenomenon, the data issue still only resonated with a few international conglomerates. However all this has changed and now the velocity with which data is being created means that the very data issue itself is well beyond 'big'. This flood of digital information is certainly a challenge for the healthcare industry.

The scale of the problem was well illustrated by a recent study completed at Icahn School of Medicine at Mount Sinai, New York City. Scientists attempted to isolate genetic sequences in order to find people resistant to particular strains of disease, and eventually they were able to identify 13 people from the 600,000 adults who were tested. However as illnesses continue to evolve and become resistant to the cures of today, the likelihood is that in the future 600,000 samples will not be adequate for successful research. Trying to place a ceiling on just how many data points may be required for such things in the future really is a futile exercise, but we can say with confidence that it will be comfortably more than one million.

Facing up to such a huge challenge, researchers are turning to the technological advancements that will allow them to bolster their analytical abilities, both in terms of handling volume and increasing accuracy. Acknowledging this opportunity, technology firms are more than happy to respond to the call for support. McKinsey's analysis of this space has recently suggested that the use of data handling strategies for pharmaceutical research could create up to $100bn in cost savings per year, and that is just in the US. The global potential for efficiency savings is huge.

Today the data problem for pharmaceutical firms is not just the volume but also its organisation within their databases. Following years of merger and acquisition activity, different research departments often work in silos, cut off from sharing information effectively between them. The result of this structural problem is that data from disparate studies relevant to each other is not easily collated. The knock-on effect of this is a huge inefficiency in research, which slows down the speed of breakthroughs and subsequently costs a lot of money.

In the frontline work of medical practitioners, having many different patient information systems means that doctors are unable to extract all existing data on that patient. This is a particularly acute problem when multiple healthcare institutions are brought together with a view to using one IT system. Simply put, a detailed and well-considered plan is required to overcome such a cumbersome problem.

The use of data handling strategies for pharmaceutical research could create up to $100bn in cost savings per year

We need bigger data
The issue can be addressed by implementing a more advanced use of metadata. Metadata is essentially a method of data tagging which will allow databases to better identify and index the information they hold. Rather than reinventing the wheel with a complete database re-factor, organisations can apply a layer of metadata to their software stack and then use that layer to manage its indexing, performed by powerful database technology. This approach saves huge IT spend and is much more manageable for an IT team to implement and manage.

But how does AI technology fit into the picture? Natural language processing (NLP) is a form of AI technology that has significantly advanced in recent years, such that it is now an ideal tool for the business environment. Able to handle huge volumes of data, NLP can be applied to a database to identify and tag individual entities, such as patients, medications or genes, throughout text. Thanks to these advancements this technology can be utilised by research and development teams and it removes the need for them to invest huge amounts of time scouring records and other databases in an attempt to find something that may not even be stored in that location.

So accurate is the use of NLP that it can be applied to isolated key terms to improve specific terms, for example 'lung cancer' rather than just 'cancer', or phrases such as 'ACL' and 'anterior cruciate ligament'. By tagging this data, whenever it occurs within the database, NLP can break down its occurrence in a sentence, creating and saving these relationships into a graph database.

Although nothing new, graph databases are becoming more relevant than ever as other technologies improve and support their significant potential. Graph databases let users perform a many-to-many data point analysis. Drawing on the relationships they store, these databases are then able to interlink data silos and extract all matching information. By presenting these relationships in a graph format, users can easily observe where the links exist throughout the databases.

Life sciences visual

Practical uses of graph databases
Following the publishing industry, graph database technology has already seen a strong uptake by some of the world's largest healthcare organisations and pharmaceutical companies. With graph databases, medical staff are equipped with a technology that can aid them in making quicker and more accurate diagnoses. For doctors, they can perform a far more rigorous analysis of records when comparing symptoms and find many more cases that share similar symptoms.

In the long term, machine learning technology may even be able to offer initial diagnosis and suggested treatments based on the information provided, however it is still some way off from removing the use of human medical experts.

Graph databases are being used by a major pharmaceutical firm to improve its drug discovery techniques by applying this technology to its wealth of data. Graph databases are also being used to support regulatory compliance efforts. The speed with which companies can report their findings has improved dramatically as graph databases allow for the swift recognition of all documentation that supports any evidence of trials requested by regulators.

Broader developments in pharmaceutical and medical technology, such as personalised medicine, offer further opportunities to apply graph databases. As individual patients offer up specific data regarding their own body such as allergies, previous reactions to medication or other pertinent information, graph databases are the perfect technology to draw out an individual's full background from multiple large databases. The application can be the ideal tool to help identify the right drugs for individual needs, thereby reducing waste and increasing effectiveness.

As AI continues to be the fashionable technology of today, it is important to remain focused on providing a real understanding of what a true definition and practical application can be for certain industries. In the healthcare and pharmaceutical industry, aspects of AI such as NLP and graph databases can create efficiency savings in research and treatment, and in many large healthcare organisations only an extra few percentage points of efficiency add up to billions of dollars' worth of savings that could then be reinvested in R&D.

Article by
Todor Primov

is a life science product manager at Ontotext

21st November 2016

From: Regulatory

Share

Tags

Featured jobs

Subscribe to our email news alerts

PMHub

Add my company
Market Access Transformation

Market Access Transformation (MAT), founded by industry veterans Baiju Aurora and Paul Howard, specializes in developing cutting edge technologies that...

Latest intelligence

Is China ready for a pharmaceutical gold rush?
Some describe doing business in China as akin to the 1990s internet boom – so how stable is its future?...
AstraZeneca’s oncology renaissance
Susan Galbraith played a key role in restoring AstraZeneca’s place in cancer drug development – she talks about the future of oncology and why there’s more to be done to...
Navigating the antibiotic resistance crisis
Blue Latitude Health speaks to Tara DeBoer, PhD, Postdoctoral Researcher and CEO of BioAmp Diagnostics to explore the antimicrobial resistance crisis, and learn how a simple tool could support physicians...

Infographics