Transforming Ophthalmology With AI

Join us as we explore Medical Frontiers!

Globally, at least 2.2 billion people have a vision impairment or blindness, of whom at least 1 billion have a vision impairment that could have been prevented or has yet to be addressed.

Early detection is critically important for eye diseases.

To obtain efficient and accurate diagnoses, there is a move towards making greater use of artificial intelligence.

Efforts to introduce AI into the operating room are also beginning, and this is expected to lead to error-free and safer surgery.

So today we are going to look at the potential advantages of a collaboration between ophthalmology and AI.

Eye diseases often progress without obvious symptoms.

Accurate diagnoses by doctors are essential for early detection.

Ophthalmologists rely on ocular fundus images, or images of the back of the eye, to diagnose diseases.

But detecting subtle changes in the early stages of disease is difficult.

An accurate diagnosis requires time.

Ophthalmologist Takahashi Hidenori aims to solve such problems with new technology.

He developed the world's only imaging diagnosis support system that can detect around 100 types of eye diseases using artificial intelligence, or AI.

I used to spend nearly an hour a day
interpreting images from some 30 patients.

At times, I would overlook
abnormalities such as bleeding.

I developed the system to get help
from AI to solve these problems.

Takahashi had AI study 500,000 fundus images the hospital accumulated over the past decade, along with abnormalities and diagnoses for each one.

He has continued to update the system by providing it with detailed data on blood vessel irregularities and changes in blood flow.

AI almost instantly analyzes any image it is given.

If it finds abnormalities, it will present potential disease names.

The system helps to reduce the time required by doctors to reach a diagnosis.

There are high hopes that the new technology will improve the accuracy of diagnoses.

Nice to see you.

Pleased to meet you.

Kakizawa Eiko developed glaucoma in her forties.

I'm completely blind in my left eye.

In my right eye, I have vision in one third
of the visual field, around the center.

Based on her experiences, Kakizawa has been trying to spread the word about the serious consequences of the missed signs of eye disease and the importance of early detection.

I used to have good eyesight.

But while driving, I noticed
I had lost my depth perception.

I'd run onto the curb or cross the centerline.
I could have died.

But I never thought my eyes were to blame.

A friend urged her to visit an eye doctor.

By the time she finally went, she had advanced glaucoma and had already lost vision in her left eye.

I was told I could go totally blind.
I was devasted.

It was out of the blue.
I knew nothing about glaucoma.

The optic nerve sends visual information from the eye to the brain.

In glaucoma, pressure builds on the nerve for various reasons, causing irreversible damage and killing the nerve little by little.

This leads to a gradual loss of the visual field.

It can ultimately cause blindness.

Glaucoma is the leading cause of acquired blindness in Japan.

It's important to get your eyes checked
so that glaucoma can be detected early.

You should remember that no treatment
can reverse glaucoma.

Let's take a look at how AI can help with the early detection of glaucoma.

Fundus images are used for AI diagnoses.

The focus is on the optic nerve.

The images are captured from the front of the eyes.

This is a fundus image of a healthy eye.

The roundish part is the optic nerve.

The nerve typically looks whiter in a glaucomatous eye.

The whiter it is, the more serious the damage.

In these darker areas, the eye's inner lining
is exposed because no nerve fibers remain.

Even if light hits this part, there's no way
for this to be conveyed to the brain.

Vision is lost in this part of the eye.

AI analyzed this image and indicated the possibility of early-stage glaucoma.

It's difficult to tell this image apart from that of a healthy eye.

Finding abnormalities is extremely challenging, even for doctors.

AI made its diagnosis based on minor damage it detected to the upper part of the optic nerve and early signs of the loss of surrounding nerve fibers.

It's really hard to tell.

The changes are so subtle that even eye doctors
would miss them unless they look carefully.

AI can identify patterns in the images we provide.
It will not miss patterns it has seen before.

- If we can detect abnormalities at this stage?
- Your eyes will be OK for the rest of your life.

There's no cure for glaucoma.

But if caught in its early stages, its progress can be delayed with eye drops or surgery.

The AI system is now being used in health checkups at a medical facility affiliated with the hospital.

It is helping with the early detection of eye diseases.

Come in.

Excuse me.

A double check by the doctor and AI improves the accuracy of screenings and helps maintain the health of the patients' eyes.

I'm amazed at the progress of the technology.

It's great that you can get
help from AI to detect diseases.

Erica asked Takahashi how the introduction of AI will affect doctors and diagnoses.

What key advantages do you think the AI offers other doctors？ Thanks to the AI system,

the time required to diagnose diseases
is now a third of what it used to be.

It's been a great help.

Every day, doctors check images of eyes, ranging from healthy ones to those showing serious damage.

They must do this quickly and accurately.

Takahashi has programmed the system to present images in order from higher risk to lower risk.

We can focus on what has been pointed out by AI.

This means that if we have an hour
to interpret fundus images, we can focus on the high-risk ones while we have
energy, then look at the rest as we grow tired.

The process is less stressful and more efficient.

Also, it's not unusual for doctors
to miss abnormalities.

I'm hoping that AI can reduce
such unfortunate occurrences.

Image diagnoses by AI are effective in detecting diabetic retinopathy, a diabetes complication that causes poor vision.

Like glaucoma, it can progress without obvious symptoms, making early detection a challenge.

Blood containing a large amount of sugar can damage the capillaries at the back of the eye.

This causes bleeding, leaving red spots.

Lipids leak from ruptured blood vessels, appearing white.

Controlling blood sugar levels is a key to treatment.

In some cases, a laser is used to prevent further bleeding or other damage.

If left untreated, it can cause sudden massive bleeding, which obstructs the visual field.

Blindness may occur in the advanced stages.

AI helps doctors detect abnormalities early on.

There are white spots that may be diabetes-related.

A close look also reveals
several dozen spots of bleeding.

So with the introduction of artificial intelligence that you can pick these things up.

Yes. Even if I miss the bleeding, the AI diagnosis
will prompt me to look for signs of diabetes.

The AI-based image diagnostic system is 83 percent accurate.

Takahashi plans to improve its accuracy by having the system study more fundus images and abnormalities.

Are there any disadvantages or are there any weaknesses with the program?

What AI is really doing is just sorting out images,
a task no machine could perform before.

Humans need to teach it how to interpret images.

For example, it needs to be
told what bleeding looks like.

It sometimes fails to detect
something humans would never miss.

Humans should double-check diagnoses by AI.
That way, we can prevent oversights.

What are your hopes in terms of it that it's used to be more sort of widespread?

Developing countries face a
severe shortage of eye doctors.

The system can make a
huge difference in such places.

Unlike humans, it will keep working
tirelessly as long as electricity is available.

I'm certain it will drastically reduce blindness.

The application of AI has also begun in the ophthalmology operating room.

This general hospital with 35 ophthalmologists is a regional healthcare hub.

It performs around 8,000 eye surgeries a year.

It has developed an AI system to ensure error-free surgery.

This footage shows the entrance to the ophthalmology operating room.

A patient heads toward the operating room from the door on the right.

Another patient whose surgery is finished exits from the left.

Ophthalmic surgery takes less time than other types of surgery.

It is performed in far larger numbers.

Frequent arrivals and departures of patients put a strain on the staff and sometimes lead to major errors.

The American Ophthalmological Society analyzed 106 cases of surgical confusions in the US.

63 percent involved implanting the wrong lenses during cataract surgery.

Procedures on the wrong eye accounted for 27 percent.

Surgery on the wrong patient or erroneous surgical procedures made up around 10 percent.

The hospital has developed an AI system that is used on a tablet computer to ensure error-free surgery.

Ophthalmologist Tabuchi Hitoshi leads the development team.

He launched a start-up in 2017 and is working with a company that has technology which can detect counterfeit banknotes and recognize faces.

Humans make mistakes,
no matter how important the task.

The more people there are, the more
dependent they become on each other.

A group of five people is more
prone to errors than a team of two.

AI is not influenced by anyone or anything
and it relies only on its algorithms.

It has neither peer pressure nor emotions,
which could affect its performance.

In developing the system, Tabuchi had AI learn three types of pre-operative errors that should never occur.

Let's look at how the system works in a cataract surgery.

The first error that should never happen is a patient mix-up.

Good morning, Miki-san.

An outpatient nurse confirms the patient's name and takes photos of her face.

The pictures are uploaded to the system.

The nurse takes the patient to the entrance of the surgery center, where an operating room nurse takes over.

This is where errors can happen.

Let me check your wristband.
On which eye are you operating today?

- Both eyes.
- Alright. Let me take photos.

The system compares the photos with those taken beforehand and confirms they are of the same person.

Facial verification, cleared.

Face masks don't disrupt the recognition process.

The nurse checks the lens to be implanted in the patient's eye.

Using the wrong lens is the most common error.

A cataract is a clouding of the eye's lens that leads to a decrease in vision.

The cloudy lens is removed surgically and replaced with an artificial one.

This is where lenses are stored.

Intraocular lenses come from different makers and in various types and strengths.

The hospital stores about 7,000 kinds.

Many patients require different lenses for each eye, making errors even more likely.

Several staff members confirm the lenses before AI does another check.

Lens verification, cleared.

An X appears on the screen, if they have the wrong lenses.

Let me cover you with a cloth.
Please keep your hands away from your face.

Surgery on the wrong eye must also be prevented.

People tend to confuse left and right.

During ophthalmic surgery, the patient's entire face is covered except for the eye, making it hard to notice an error.

A card is placed on the opposite eye, which enables AI to tell left from right.

Verification is complete.

Eye verification, cleared.

When operating on both eyes, this step is carried out twice.

After they have completed all three steps, the surgeon enters the room.

Let's do a final check.

- AI verification is complete.
- Good.

- You're Miki Keiko, right?
- Yes.

We'll operate on your right eye.

After AI verification, the team does a final visual and verbal confirmation.

The surgery then finally begins.

Tabuchi says the AI system is 100 percent accurate.

It enables the members of the surgical team to do their job confidently.

It's finished. You did great.

It went well.

We tend to believe people around us
are keeping an eye on things.

But that's not always true, so confirmation
by both humans and AI is important.

Rather than relying entirely on AI,
we need to also take action ourselves.

Tabuchi holds feedback meetings regularly with the staff to share problems.

Detailed records are kept on AI verifications for all surgeries.

Close calls occur once or twice in every 1,000 surgeries.

But AI has managed to prevent any incidents.

Let's continue to share any incidents.
They should at least be shared between the nurses.

I'm determined to make surgery error-free.

You might have heard about futuristic surgeries
using state-of-the-art technology.

But there are many basic things we should do to
avoid mistakes, and they should be the priority.

My ultimate goal is to completely automate
the verification process for surgeries.

Understanding the strengths of AI and implementing it widely can result in major advantages for both doctors and patients.

And AI has great potential, not just in ophthalmology, but also in other fields of medicine.

I very much hope the strengths of both humans and AI can be used to realize a future where more accurate and safe examinations and treatments are possible.

Smartphones and computers are part of our daily lives.

But by looking at them for long periods, we overwork our eyes.

Takahashi will share tips to care for tired eyes and keep them healthy.

Aging and various types of stress can lead to a decline in visual function.

Check how many of the following 10 signs of unhealthy eyes you might have.

Your eyes tire more easily than before.

Your vision sometimes worsens in the evening.

You find it hard to read newspapers or books for long periods of time.

You sometimes spill food or drink during meals.

Your vision is blurry, even with glasses.

Your eyes tend to be sensitive to light.

You sometimes need to blink to see clearly.

Straight lines sometimes appear wavy.

You have felt unsure of your footing on the stairs.

You have missed traffic signals or road signs.

How many did you have?

Oh, six.

Six? I have five. I have five!

My eye is tied more than easily than before, number one.

Sometimes difficult to see in the evening a little bit.

I find that also with reading books and probably little bit less and often I feel it's too bright.

I don't get enough sleep.

I spend more time looking at the
computer screen than at my patients.

I'm afraid I'm straining my eyes.

But what other issues? I mean, and one of the diseases is it related to?

Easily tired eyes or poor evening vision
can be a result of dry eye, which is caused by a drop in tear production.

Light sensitivity can occur with cataracts
or dry eyes, both of which are age-related.

How many of the items applied to you?

If there were two or more, visit an eye doctor and get your eyes checked.

Takahashi will share an easy massage and exercise to relieve eye strain.

First is the massage.

Gently press down on the eye area with your fingers just enough to feel comfortable, while avoiding the eyeballs.

These parts are also connected to the eyes.

Use your thumbs to massage your temples
and your other fingers for the scalp.

The muscles around the eyes extend to the temples and the scalp.

Massaging them helps to relieve tension in the whole region and ease tired eyes.

I can feel my scalp nice and tight.

Do this massage for about five minutes in the evening or later, when your eyes feel tired.

Next is an eye exercise.

Move the muscles around the eyes to relax them.

Our eye muscles become less flexible as we
age and use them less frequently.

But we can ease stiffness by rolling our eyes.
Try to roll them outward as much as possible.

Slowly rotate your eyes clockwise and counterclockwise, two times each.

Because I think it's so much a quality of life, really comes from the eye, I think we take it for granted.

Vision loss makes life extremely difficult.

If you notice any abnormalities,
it's important to go see an eye doctor.

Thank you so much Associate Professor Takahashi.

Thank you for having me.