IEEE Future Directions
Related Articles
I remember a talk I listened to at TTM 2016 looking into wearable devices as tools for improving healthcare. The speaker said that devices like the Apple Watch (meaning all consumer grade devices) where simply too weak in terms of data capture to be considered seriously as medical devices. Notice that he was not just referring to the quality of the sensors used, he was actually more focussing on the way they were used (you wear the watch one way now, than it rotates a bit on your wrist, then it goes over the cuff of your sleeve…).
Now a study from the University of California at San Francisco and Cardiogram, a health care start up, tells a different story. The data gathered from those sensors can be processed by an AI software resulting in accurate detection of abnormal heart rhythm, 97% accuracy, detection of sleep apnea, 90%, and hypertension, 82%. Don’t be fooled by these numbers: you cannot expect 100% and actually those numbers are better than what you would get from a sporadical use of medical devices. Indeed, the upside of wearables, such a watch, is that it is basically always with you!
The important advantage is that a wearable like a watch can detect an abnormal situation in a person that is not aware of it. In the US it is estimated that 80% o people with moderate to severe sleep apnea are not aware of their condition and likewise many people suffering from hypertension are not aware of it. Wearing a watch that can detect the condition may turn out to be a life saver, since appropriate cure can be initiated.
Of course, in order to be useful the watch shall be “credible”. If the detection is not reliable people will disregard its result. This is what AI is achieving: turning sensors data analyses into reliable information.
Cardiogram and the University of California developed a software, DeepHeart, and trained it on 70% of the 6000+ participants in the study and applied on the other 30% participants that were not used in the training phase.
This is the third study so far using deep learning technology applied to health care after the one announced by Google for detecting diabetic eye disease and the one from Stanford detecting skin cancer.
