DeepHeart, an AI for cardiac health, is using consumer wearables to reduce the number of people with hypertension and sleep apnea who go undetected.
According to the World Health Organisation, 1.1 billion people have chronic high blood pressure — known as hypertension. Of these individuals, one in five are undiagnosed.
Meanwhile, sleep apnea — where disturbances are caused by having one or more pauses in breathing or shallow breaths while you sleep — is a problem with around 24 percent of men and 9 percent of women. A shocking 80 percent of people with the condition do not realise they have it.
There’s a clear issue with the detection of these conditions and — within the US alone — hypertension and sleep apnea are expected to cost $46bn and $150bn, respectively, in direct medical spend, lost productivity, and accidents.
Cardiogram, who developed the DeepHeart AI, is hoping to step in and solve the problem.
Waking up to the issue
Cardiogram wants to take advantage of the wearables many consumers already own — such as the Apple Watch, and Android Wear, Garmin, and Fitbit devices. The data collected of users from these devices can be used by DeepHeart to detect potential conditions with impressive accuracy.
For sleep apnea, DeepHeart achieved an accuracy of 90 percent. For hypertension, the accuracy is a still impressive 82 percent. Over time, it’s likely this accuracy will increase further.
DeepHeart has already displayed its expertise in detecting conditions. Back in May, Cardiogram released a study which showed DeepHeart could detect atrial fibrillation — the most common abnormal heart rhythm — with 97 percent accuracy.
According to Cardiogram, the heart rate sensors in many consumer wearables can detect conditions not typically associated with the heart — such as sleep apnea — due to the body’s autonomic nervous system connecting the heart with the brain, liver, stomach, intestines, pancreas, esophagus, and blood vessels.
In order to train its AI to recognise changes which can indicate the various conditions, Cardiogram recruited 6,115 users of its Android and iOS apps to participate in an online study with the UCSF.
The next step for Cardiogram is to have its studies undergo peer-reviewed clinical research; including a previous study where the AI was trained to look for the signs of a stroke. Further down the line, Cardiogram wants to train its AI to detect undiagnosed diabetes.
DeepHeart is showing us the potential for AI in healthcare. Some undiagnosed ailments could be detected and dealt with earlier to improve the outlook and quality of life for patients. The results may even be adopted into digital health services like GP at Hand to provide detailed information on an ongoing basis to doctors, and prevent wasting their precious time with multiple appointments.
There’s almost limitless potential here, and we’re excited to see how AIs like DeepHeart impact healthcare in the coming years.
What impact do you think AI will have on healthcare in the coming years? Let us know in the comments.