Latest Posts

Latest Comments


The beating heart of innovation: how machine learning could drive patient health care


Posted by |

Innovation is key if we’re to continue to drive public services forward, but innovation needs to be supported by technology if it’s to fulfill its potential. Technology innovation was very much at the core of the development of the AliveCor Kardia™ Mobile ECG product. AliveCor’s founder, Dr David Albert MD, used his understanding of hospital-grade, heart monitoring medical devices and applied it to that universal access device – the smart phone. Dr Albert recognised that technology could help patients to monitor their own heart rates using a small device that links to an app on their smart phone. The data can then be sent to medical practitioners or the patient can be connected directly to the doctor, who can then review and advise the patient.

The ability to self-monitor and control your own medical data has been growing exponentially since the advent of the smart phone. Yet truly the industry is still in its infancy; as is its potential to expand into additional areas, such as machine learning.

The ability of machine learning to replace human learning and to speed up the analysis of data will be key in the delivery of patient services in the future but it has a long way to go. Like babies, it’s still learning and growing in its understanding. The child analogy works well here.

Take AliveCor as the example. Our machine learning capabilities are currently at level 1 – or infancy. We’re able to interpret an ECG to see if the heart rhythm is normal or if the most common abnormality is present, atrial fibrillation. We’re able to look at data generated by our ECG devices and app on an individual basis and recently added the ability to integrate a users Apple Health app data (if they use Apple and agree to share).

As our systems’ machine learning abilities grow a little and take its first steps, it can begin to apply what it knows. We already have well over six million ECG readings on our cloud-based servers growing exponentially. These have been reviewed and analysed and sent on where requested. We now hold more readings than a Cardiology Consultant could review in a lifetime of work. The system is looking at this data and may begin to generate insights at a population level. Backed up by a depth of data, the ideal goal of the system would be to advise patients on their potential risk levels and recognise early warnings of a possible heart issue.

Then as the system begins to grow it could continue to analyse the data of millions of patients globally in more detail. Theoretically, it may be able to recognise patterns, and to make assumptions based on what it’s learnt from gigabytes of real data. The holy grail would be to start to make predictions. For example, if analysis is able to identify hot spots of cardiac activity, it could seek to correlate or identify causes. We would love to be able to point to the likelihood of cardiac arrest within the population. That way, doctors and healthcare providers could make decisions around possible preemptive care rather than having to treat the outcomes: reducing an entire population’s risk.

While for many machine learning overtaking human learning seems a long way in the future, its application is ready right now. And for us working in the field, we’re able to see the benefits in treating patients, now and as its capabilities grow into teenagers and adulthood.



Post a comment

Comment submitted! Comments needs approval before being displayed.