How AI is Revolutionizing Healthcare: Recent Advances

Physicians examine many charts to perform diagnoses: X-rays, CAT scans, MRIs, and more. But the charts themselves don’t say whether or not you have an illness- they just represent how electromagnetic waves pass through your body.

It’s up to healthcare professionals, then, to examine those charts and decide if something is wrong.

Suppose you get an X-ray because you have symptoms of stomach cancer. There may be abnormal-looking shapes in the chart, but it’s up to a doctor’s expertise to decide if they look malignant. Then they can follow up with more tests.

AI tools can now examine medical charts and look for clues about possible illnesses. They then point out those clues for medical professionals to investigate deeper.

Sometimes, AI can even point out possibilities that human doctors were not aware of.

In a way, this is just a more complex version of something AI has been doing for a long time: image recognition. The main difference is that the salient pieces of information involved are extremely subtle in medical imaging.

The main drawback to these tools is that they are not perfect- false positives and negatives happen. That’s one reason why they can’t be used to make official medical diagnoses. But in the hands of an expert physician, they can sometimes find looming problems that would otherwise go unnoticed.

Personalized Treatment Plans

After diagnosing you with an illness, your doctor will give you a treatment plan. This plan may involve drug prescriptions, physical therapy, or lifestyle change recommendations.

Doctors give out almost exactly the same prescription to hundreds of patients with the same problem, with slight alterations for factors like genetics or allergies.

It makes sense, then, for AI tools to be able to write first drafts of treatment plans for most illnesses. They can take your entire medical history into account, and make a plan similar to other patients that fit your characteristics.

Again, the AI can really only make a first draft. It is up to the doctor to read, make changes, and sign off on the official plan. But this will still save many hours of time spent preparing nearly identical treatment plans.

One limitation here is in dealing with rare diseases. In such cases, with little training data, AI tools may not be able to make good plans on a first draft.

In a recent article, I discussed how AI chatbots are taking the customer service world by storm. Healthcare service is a little bit more complicated than most other industries, but the core is essentially the same.

Generative AI is ideal for chatbots, because they can be trained on niche topics to provide answers to most questions.

There are some legal questions to consider. In the US, HIPAA laws place strong constraints on the privacy of the patient. If AI is not human, what does it mean for it to have access to your medical records? And what if it accidentally leaks your information?

Discovering new drugs is a scientific process, but it is filled with a lot of trial and error. A big part of the work comes down to generating candidates for new drugs to test.

This is a domain where AI is incredibly powerful. In a recent experiment, an AI generated 40,000 new drug candidates in just 6 hours.

There’s one drawback with that particular story: almost all of those chemicals were actually poisons, as the research team was researching biochemical weapons.

But the same technology can be applied for good. It may even be able to suggest ideas for drugs that help rare diseases which otherwise would get little attention.

If you’ve ever worn an Apple Watch for a few days, you already know a thing or two about remote patient monitoring.

Wearables and other IoT devices can gather and synthesize a variety of medical information, from step counts to blood sugar levels. The amount of information they produce can be staggeringly large.

Which is exactly why AI is perfectly suited for making use of it. AI programs can recognize trends and abnormalities, such as sudden blood sugar changes or changes in posture or gait.

Clearly, there are many ways for AI to revolutionize healthcare. Time will tell which of these will turn out to be most useful.

Despite rapid progress in recent years, there are still many roadblocks in the path ahead. AI in healthcare faces many technical, legal, and ethical questions. In my next article, I will take a look at those, and try to make sense of where we are headed.