Deep Medicine: How AI will improve self-care

4 min read
artificial intelligence healthcare glucometer
Image credit: Depositphotos

Welcome to TechTalks’  AI book reviews, a series of posts that explore the latest literature on AI. This post is the first part of a two-part interview with Dr. Eric Topol about the impact of artificial intelligence on health care and medicine.

In the last part of our interview with Dr. Eric Topol, we discussed how artificial intelligence algorithms can return the gift of time to doctors and help them have more human interactions with their patients. This is a subject that Dr. Topol discusses early on in his latest book “Deep Medicine: How Artificial Intelligence Can Make Healthcare Human Again.”

Another topic Dr. Topol discussed about the role of artificial intelligence in healthcare was giving every person more insight and control on their own health. This is one of the areas where deep learning algorithms have made great inroads.

Less visits to the doctor

Thanks to innovations happening in the field of internet of things (IoT) and wearables, there are plenty of tools that patients can carry around with them and regularly collect vital signs and other health-related data. This information can then be fed into neural networks, which will then draw pertinent insights and conclusions.

For patients, it might mean that they will need less visits to the doctor.

“For the patients, AI gives them more charge to let them do many things that are common diagnoses, without the need for a doctor, and also tracking their own data that they generate, to guide their help.

There are now many personal health devices that can collect important data at short intervals and run them by AI algorithms to obtain important information about patients’ health without the need to visit a doctor. Complementing that are several AI-powered health assistants that provide users with basic-but-effective diagnosis that would have otherwise required going to a clinic and waiting for your turn.

The AI-powered virtual medical coach

But the future is broad health support, Dr. Topol envisions. So, for instance, if you’re at risk for a condition known from your genome such as a heart disease, then AI-powered health care devices will accrue all your data that moment to moment seamlessly and feed it into a deep learning algorithm that has been tailored to your health condition. The AI will also improve its performance with up-to-date world medical literature and data that is relevant to your condition.

“The AI has everything about your prior medical history, your labs, your scans, your genome, sensors you’re wearing, your environmental sensors, medical literature. All this data is just continually being assessed to give you feedback, whether that’s through an avatar or speech or text, you can pick that,” Dr. Topol says.

For the moment, AI assistance exists for limited conditions, like managing diabetes. But as the technology develops, it will expand to other areas as the pieces come together. As data collection and analysis becomes easier and more affordable, AI algorithms will be able to train on deep data about thousands and millions of patients and provide the best advice and coaching.

“The way it works today is there’s this one-off appointment with the doctor. In the future, you will have a coach that is with you all the time as long as you want to have this coaching operation,” Dr. Topol says. “We have this unique capability of bringing in massive amounts of data about a given individual, and processing the data with AI algorithms and giving feedback, and that is another way to decompress the role of the doctor whereby the person who is getting this feedback could contact this doctor when there’s need. The virtual medical coach is starting to get legs. Eventually it will be a way, a path to prevention, which is fulfilling a dream we never really actualized before.”

Who will own health data?

wearable fitness tracker

The privacy implications of data-hungry AI algorithms are already being widely discussed by scientists and thought leaders. Who will own the massive data that patients generate? Will we be able to trust large tech companies with our health profiles? If not, should governments take care of them? There are many ways any of those scenarios can turn into privacy disasters, especially as AI algorithms become increasingly efficient at influencing people in inconspicuous ways.

For the moment, patient data is scattered across the archives of different hospitals and health organizations and the servers of tech companies. There’s no single store where patients can access their data.

Dr. Topol believes that in the future, data will belong exclusively to the patient, a point he expands on in Deep Medicine. “Undoubtedly, there’s going to be progressive democratization. Once the data is imminently portable, you can’t withhold it from people. People should own their data,” he says. “There is no home today for owing data, only parts of it are in electronic records, and even that’s dispersed quite widely. When you’re generating data, more and more people are going to be using medicalized sensors, wearable biosensors, and you will have our genome assessed and gut microbiome assessed, and all these other data—they don’t sit anywhere.”

In addition to the privacy implications of patient data being at the mercy of other entities, the scattered structure of health data makes it hard to train efficient AI algorithms. “If you have all your data, you can use it as input for any deep learning, AI algorithm. Today, no one has all that data properly aggregated. So if you don’t have good input, you’re not going to have good output from AI algorithms,” Dr. Topol remarks.

The perils of having too much health data

While we explore the benefits of data- and AI-driven self-care, we must also discover the possible negative effects too much exposure to health data. Studies show that tracking symptoms too closely can have an adverse effect on the health of patients, creating a feedback loop in which seeing negative symptoms lead to anxiety, which further affects the symptoms, and so on.

Dr. Topol acknowledges that collecting data at high frequency and returning it to the patient can make things worse by generating false positives and causing anxiety. That’s why there’s no one-size-fits-all approach that everyone should adopt.

“It has to be used in a very special, specific way, for the right person, at the right time, rather than just broadly distributed or marketed,” Dr. Topol warns.

As an example, Dr. Topol mentions the Apple Watch 4, which has a built-in functionality that uses AI to detect atrial fibrillation, a disease that causes abnormal heart rhythms and can lead to severe heart failures. There have been several cases where Apple Watch 4 has been able to save the lives of patients by detecting early symptoms. But studies also show that it can show false positives for the people who don’t have heart problems.

“Apple Watch is being marketed as a way to monitor your heart rhythm for everyone. That’s not good, because most people, especially people who are young, less than age 50, their risk of having atrial fibrillation is exceptionally low,” Dr. Topol says, adding that the feature can cause a false alarm, pushing the user down a tunnel that could wind up with all sorts of negative things.

“So what we have are great powerful tools that have to be applied with great discrimination, that is with very careful, judicious use, which isn’t likely to happen by accident. It’s going to take a lot of work,” Dr. Topol says.

Advertisements

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.