Wearables can now Detect Early Signs of Diabetes (using Machine Learning, of course)

Wearables have already come leaps and bounds in recent years. From just counting the steps we take to the calories we burn, they can now even monitor our heart rates. Using this latter feature and putting it into a neural network, startup Cardiogram has launched an application that can detect early signs of diabetes.

                                                             Source: Cardiogram

Diabetes has always been a silent killer, but the number of people suffering from it has increased exponentially in the last few decades. A recent study by Cardiogram and the University of California revealed that almost 90% of prediabetics are unaware of what is about to hit them. More often than not, people continue with their normal lifestyle, unaware that their health is at severe risk with each day diabetes goes undetected.

Yes, there are devices that help with diabetes. But most of these help once the disease has already been detected. As it turns out, any wearable device you have (Apple Watch, Android wear, FitBit, etc.) that can monitor heart rate, can now be used to predict the risk of diabetes.

Cardiogram’s method of detecting the risk of diabetes is founded on how the variability of our heart rate can be connected to the likelihood of contracting diabetes. As one can imagine, it’s not possible to gather the potentially millions of data points required to train the neural network (it’s extremely time consuming and expensive). Instead, Cardiogram and the University of California turned to semi-supervised machine learning. Using 33,628 person weeks worth of sensor health data, they trained their deep neural network (called DeepHeart). Once trained, they tested the neural network on another data set, this one consisting of 12,790 person weeks. The result was a super impressive 85% accuracy rate.

                                                            Source: Cardiogram

As of today, the researchers are working on increasing this accuracy rate. Apart from diabetes, the neural network can also detect other heart related illnesses like hypertension and sleep apnea.

Our take on this

The best thing about this application is that you don’t need to spend an extravagant amount of money to reap the benefits. All you need is a smartwatch, and even that can be a relatively cheap one (as long as it monitors your heart rate). Analytics in healthcare has been booming recently and this is just another example of it. Diabetes is a debilitating disease wreaking havoc across the world. Hopefully, Cardiogram’s application will reduce the number of people suffering from it (or at least delay it as much as possible by helping people take precautionary measures).

Pranav Dar

Senior Editor at Analytics Vidhya.Data visualization practitioner who loves reading and delving deeper into the data science and machine learning arts. Always looking for new ways to improve processes using ML and AI.