Google believes the future of healthcare is in structured data and AI. If they are right, this will transform the way we interact with health. But the raises the question, should we allow a company who makes money from data closer to our medical records.
Anyone who works in the healthcare knows that the industry is in need of a digital revolution. Companies are aware of this and many are striving to fill the many technical needs of medicine in the 21st century.
One prominent company investing heavily in medical technology is Google. Google is attempting to transform healthcare with their expertise in big data and deep AI. They have a tremendous capability to improve all aspects of medical practice and research. But in a field that is built on the phrase ‘Do no harm’, does Google have the moral code to be trusted with our healthcare data.
In this article we will delve into some of the projects Google are doing, as well as discuss the ethical issues around each. But first…
A doctor is effectively an expert in pattern recognition. They spend years at medical school learning as many patterns as possible, then attempt to see these patterns in patients in order to diagnose them. E.g. a patient with epigastric pain, vomiting and a raised amylase (blood test) = pancreatitis. The pattern is there symptoms, signs and investigations, which all add up to their diagnosis.
This form of pattern recognition lends itself well to artificial intelligence, and seen as though Google is the biggest in this business, it places them in the perfect position to capitalise on this lucrative market.
Google is using a smart approach to tackle some of the world’s most complex diseases. It does this initially by (you guessed it) data generation. It does this through its wearables and analysing imaging, e.g. CT, MRI, etc. It is then attempting to use AI to detect diseases using this data. It is then using this data combined with its wearable tech to help patients make positive lifestyle modifications.
Diabetic retinopathy happens in patients with poorly controlled diabetes. It is the result of damage to the small vessels at the back of the eye and in severe cases can cause blindness. There are characteristic appearance that occur at the back of the eye at various stages of this disease. Google (or more precisely its subsidiary Verily) is attempting to detect these changes using AI. It has partnered with a subsidiary of Nikon, Optos, which produces machines for retinal imaging.
In a paper published by the company, they state that their algorithms are as accurate as opthalmologists at detecting diabetic retinopathy.
Google, again through Verily, are attempting to solve the issue of blood glucose monitoring. Currently, diabetics have to prick there fingers to draw blood in order to directly measure their blood sugar level.
Verily, alongside a medical device company called Dexcom, have produced the Dexcom G6. This device measures the glucose level of interstitial fluid (the fluid between your cells) just beneath the skin. Currently, this device needs a once a day calibration with blood glucose levels, but the hope is the next iteration of this device will not need this, therefore, completely removing the need for diabetics to record blood sugar directly from their blood. News which will be a relief for many needle phobic diabetics.
Google is attempting to develop tools that will allow it to monitor for the development of heart disease. It’s first tool is the Study Watch, which is produced by Verily. The watch includes an ECG and heart rate monitor which can be used to monitor for the signs of heart disease.
The Study Watch is currently the only tool actively in use. However, Google have patented designs for a passive heart monitor. This monitor would image areas of key blood flow and use optical sensors and machine vision to provide a continuous monitor of general heart health. Google have hypothesised that this device could also help detect for signs of stroke and cardiac arrhythmias, which cause characteristic changes in blood flow.
Verily has partnered with teams in the Netherlands to sift through anonymised data of patients who suffer from Parkinson’s disease in an attempt to help identify the disease earlier as well as improve its management. This project has been called the Personalized Parkinson’s Project.
It also acquired a company called Lift Labs, which created the Liftware Spoon that is designed to help those suffering from Parkinson’s to stabilize their food while eating, further indicating that Google has a desire to move into the medical device market.
At the heart of almost everything Google do, there is data. A huge issue with healthcare data is its poor quality, connectivity and accessibility. Data is not stored well, different systems cannot be integrated with one another, and it is difficult to access due to enhanced data privacy regulation in healthcare. In order to tackle this Google have created multiple routes to improve physicians ability to feed healthcare data into its AI systems.
New data pipes
The healthcare industry is beginning to realise the benefits of a fully integrated system, all be it very slowly. FHIR (Faster Healthcare Interoperability Resources) create standard practices for different healthcare elements. This means that companies can build APIs for healthcare data that will allow applications and researchers to access data in a standardised manner.
Importantly, Google purchased Apigee in 2016, an API company that has focussed on building healthcare APIs using the FHIR framework. This company will allow Google to find new ways to ingest this healthcare data.
Google is focusing more energy into its cloud computing platform. Although, it finds itself behind Microsoft Azure and Amazon Web Services in many areas.
However, Google has made an effort to focus on healthcare, attempting to fix some key issues with healthcare platforms and services. One example we have already discussed, Apigee. However, there is also Google’s G Suite providing Google drives, docs, and other tools, all of which can be integrated with Googles HIPAA compliant cloud service, meaning patient information can legally be shared on the platform.
Google may also role out more cloud-based open source solutions to help fix specific healthcare issues. DeepVariant is an open source deep learning tool for genomic analysis. DeepVariant and similar tools could be used as another suite Google publishes in the future, increasing ease of access for researchers outside of Google, further enshrining its position within the healthcare and biochemical research communities.
Third Party Datasets
Rather than just tapping into existing datasets, Google is attempting to build its own through Verily. Verily is working on two large projects:
- All of Us Research: aiming to track 1 million participants using genomic, lifestyle and biomarker data. If they are successful this is likely to produce the world’s most comprehensive healthcare dataset and could have huge implication for healthcare research. The All of Us research project is working alongside the NIH to produce this data.
- Project Baseline: Verily’s own research which is run entirely by Google. Aiming to produce its own dataset from 10,000 participants over 4 years. Participants will monitor their everyday activities using a Study Watch, sleep sensors, and respond to periodic survey questions. Participants also need to visit a participating site 4 times a year for more in depth tests.
It is worth noting that the data for both of these tools is stored in Google’s Cloud infrastructure. Further ingraining the Google Could platform into the health research infrastructure. You may be able to see a pattern emerging?
It is well known that the public perception of Google, especially around issues of privacy, is not good. In some fields this is not so much an issue. But in healthcare privacy and data protection is at the forefront of most peoples minds.
In a recent poll performed by Morning Consult and Politico National Tracking participants were asked ‘How much do you trust Google to keep your personal health data private’. 10% had no opinion, 37% said that they had confidence, whereas a massive 53% said they had very little or none at all. This is a damning statistic for Google, especially if it wants to increase its presence in the healthcare industry.
Although Facebook has become the poster child for internet scandal, Google has known its own issue with data breaches. Just before the EU GDPR regulations came into force, Google found that its Google+ social network had exposed personal data. It found that between 2015 and March 2018 developers from outside the company had potential to access personal profile data due a glitch in their website.
An additional issue is Google’s usual revenue model which is based around advertising. I do not believe advertising would work within the healthcare community as it would likely erode public trust in the company even further. If they are to be successful in this market, they will have to develop novel revenue streams in order to drive profits within the industry.
All this being said, if we look at this from a purely scientific point of view, Google has tremendous capacity to transform our healthcare analytics and platforms for the better. It is undeniable that AI has the ability to expedite the diagnosis and research process. This will lighten the load on a severely stretched healthcare system, as well as provide new insights into research previously incomprehensible using conventional paradigms. There are few companies better positioned than Google, given their AI capabilities, to effect this change.
Google has a tremendous capability to transform healthcare for the better. However, the extent to which they do this is not so much restricted by their technical ability, but the trust the public holds in tech giants with their personal data. In order for Google to capitalise on this new and exciting market, they must first restore our faith in such companies to maintain our privacy and act with responsibility.