17【原声】Chapter 6 | Self-Tracking and Self-Improvement

Part 1: Epic shift in collecting health data

00:00/13:15 

Hello, listeners of Himalaya, and welcome to this episode in our series "AI and Us". This time and for the next three episodes we will take a look at how data-driven AI changes medicine and human health care and we begin by taking a look at the recent epic shift in collecting health data.

So let's start with a simple case. Suppose you caught a really nasty stomach bug and fell sick. Because the bug hasn't gone away after two days, you decide to consult a doctor. The moment the doctor sees you, you are a blank slate to her. Of course, it might not be the first visit you ever made to this doctor. You may have been visiting the doctor before, for example, when you badly sprained your ankle two years ago, or with an ear infection last winter. 

The doctor may know your name, and your age, and quickly scan the patient record to remind herself of your past visits. But that only will tell her about a sprained ankle or an ear infection. Then you tell the doctor about your stomach problem. It's the first real piece of information. You add that you think it's a stomach bug. The doctor may nod, because stomach bugs are quite common. In her mind, she'll think of the standard stomach bug symptoms, and may ask you whether you had any of them. 

She is collecting additional pieces of data. Then she may think of the two or three most likely alternative illnesses with similar symptoms and based on all this decide to collect more specific data, by taking your temperature, or measuring your heart rate, or drawing blood to get it analysed.

If the limited number of data points the doctor is collecting this way conform with the standard symptoms of a stomach bug, but not with some of the alternatives she is considering, she'll likely agree that it's a stomach bug, and based on what treatment an average patient would get when suffering from a stomach bug, she'll write up a few medicines for you to take and off you go. 

The whole affair is over in a couple of minutes. At best, the doctor may tell you to come back in a few days if you aren't getting better. That kind of approach is based on data - your temperature was taken, your blood was analysed, you were interviewed. But when you look more closely at it you realize that the whole process of medical diagnosis and treatment is pretty "data-impoverished". Only a few data points were collected and analysed.

Why does your doctor not collect more data? It's the same reason you do not come to your doctor with much of the necessary data already collected. Because collecting health data has long been difficult, time-consuming and costly. And because hardly anybody trusts health data that has been collected by somebody else. 

That's why when doctors send patients to hospitals, there all the data is collected yet again: blood pressure is measured, and heart rate, temperature is taken, blood is drawn, CTs and MRIs are ordered, even if all of the data had already been collected by the doctor before.

Data management in health care feels outdated because it is. We collect too little data, we manage it badly, and we don't share it. It feels as if half a century or more of digital technologies, networks and data-driven analysis just went by the medical professions with little impact.

Part 2: Digital devices help data collection

03:58/13:15 

So are we stuck forever in an outdated data mindset in health care? No, not at all. Because there is an astonishing data revolution under way, and we are already in the midst of it. We just haven't fully realized it. It’s driven by the availability of cheap and good sensors, and by increasing use of AI and machine learning to make sense of all the data these sensors collect.

Just think about it: a typical smartphone captures movement and acceleration in three axis left/right, up/down, and forward/backward. That way we can play games on our smartphones by turning, twisting and moving the phone. But the sensor data can also be used to capture the way we walk. Our gait is quite individual – every person walks slightly differently, depending on weight, size, height, muscle strength and many other factors – and so we can use data capturing our gait as a security feature, much like a fingerprint or a retina scan. 

But, far more interesting in the health context, is the fact that our smartphone can also measure our gait and compare it over time – and alert us when our walking style changes ever so slightly. This may point towards serious illnesses we are developing. And for those suffering from an illness that affects walking, such as Parkinsons, the smartphone can measure progression as well as differentiate between good and not so good days. Similarly, apps on our smartphones can ask for a person with a tremor to hold it in the hand for a minute to measure the intensity of tremor and collect a valuable time-series of tremor data.

Our smartwatches capture heart rate and our heart muscles’ activity is captured through a simple version of an ECG. You may have heard that in a study with health experts at Stanford University, Apple was able to show that they can detect early signs of heart disease that way, and alert smartwatch wearers to the potential problem before it is too late. It’s an amazing insight, but it could only be gleaned by the ingenious way we capture health data today.

And there is much more: our smartwatches capture fitness data and track our sleep. Thermometers and glucose monitors can send their data to our smartphones. And beyond the smartphone, there are now many labs that provideblood work data digitally and electronically. Our DNA can be sequenced, and our genetic code can be delivered to us.

All of this data is really useful, because it capturesour illnesses,our life quality. And most important, it captures how we change over time. Because that is what an illness is all about: a change over time to the worse. A doctor cannot capture changes over time easily, because he only sees patients at a particular moment, not every day. But our smart devices that are with us all the time do capture our data regularly across time.

All of the useful data, however, is of little value if it is not utilized properly. And that is the domain of AI much more than conventional doctors. That sounds strange, but please let me explain. The data captured is far more complex and multi-facetted than a couple of data points that doctors capture when we visit them. Hence, we need special tools to identify the real insights in these data patterns. Remember? 

That’s precisely what machine learning is good at. And so it has been machine learning that helps identify illnesses from a change of gait and that predicts heart disease from a simple ECG. AI is as indispensable in the health data context as the microscope is indispensable to make sense of what our eyes cannot see.

But teasing out insights from health data that our digital tools collect about us is not the end of my story. In fact, it’s just the start. Because our devices can not only help us see insights in data, they can also help us translate these insights into action. I’ll explain through a personal confession.

Part 3:Digital devices and health

09:04/13:15 

I have been using a smartwatch for years now. At first, I did not much care about the fact that my watch recorded my activities and some of my vital signs, such as my heart rate. I also switched my watch to one with more sensors, and my watches’ software has seen big yearly changes. It’s a strange situation in which your watch gets somehow better every time you update the software. 

But as it records more data about me, and analyses that data, it not only provides me with information, it suggests actions. These are not outright orders, but rather subtle nudges. It tells me that I can still reach my fitness goal for a day, or that I should stand up when I sat for too long in front of the computer.

I did not want to concede it at first, but there is no denying that my watch has changed my behaviour. I now walk up and down stairs rather than take the lift, I look to work out if only a bit regularly. And over the past year and a half my watch has helped me to get sufficient sleep as well, by reminding me to go to bed earlier when I haven’t slept enough the night before. And, yes, I feel fitter and more efficient in my work.

Lots of people have gained weight during the severe lockdown due to the virus pandemic in the first half of 2020. I haven’t. Because I have been exercising regularly and made sure that I would go to bed rather than binge-watch a TV series and snack.

I am engaged in tracking myself. And I do it not because of curiosity (although that’s a part of it), but with a purpose. As I grow older, I want to make sure I lead a healthy lifestyle and get the exercise and sleep my body needs, at least to an extent. Sometimes, of course, I don’t heed my watch’s advice. When it is raining outside, I don’t take my bike for spin or go for a walk. And when I have a lazy day, I disregard the good advice, too. Make no mistake: I still feel very much in control of my life. My smartwatch isn’t my boss; rather it is my trusted adviser. 

Because I know that it has more data than I do, and because I know that it undertakes fairly sophisticated data analysis. I still remember the first time I went on a bike ride with my watch. I switched on exercise mode and off I went. At the end, I looked at my watch and I and was almost blown away. Not only did it record lots of data, but through sophisticated machine learning it saw patterns in it I never would have expected, and gently suggested ways to improve my workouts.

My story, of course, is not unique. Millions and millions of people are tracking themselves, hopefully with the intent to improve their health and life quality. In the process, they collect lots of data about themselves, and subject it to rather sophisticated analysis, gaining true insights from that data. It’s what AI and data does to how we perceive our health – no longer as something that happens to us, but something that we can influence, shape, and improve if we’d like. And in doing so this paves the way for a revolution in health care.

But that’s the story of the next episode. I hope you’ll join me again.