How AI can make health care better

Summary

The video discusses the growing role of artificial intelligence (AI) in healthcare, particularly in addressing the shortage of doctors and improving patient care. AI is shown to rapidly analyze medical data, such as retinal scans, helping to diagnose diseases more efficiently. The script highlights success stories, such as saving eyesight and heart operations, while also addressing concerns around patient privacy and the 'black box' nature of AI decision-making. AI is seen as transformative, potentially speeding up medical procedures, trials, and innovations, but challenges remain in ensuring transparency and accountability.

Takeaways

• 🤖 AI has the potential to revolutionize healthcare by transforming diagnosis and treatment processes.
• 🧑‍⚕️ There is a global shortage of doctors, leading to delays in patient care, particularly in fields like ophthalmology.
• 👁️ AI systems can diagnose over 50 types of eye diseases quickly, which could prevent many patients from going blind due to treatment delays.
• 🌍 By 2050, vision impairments are expected to increase by 50%, making the role of AI in diagnostics more crucial.
• 🔬 AI's ability to process and analyze medical data faster than humans could improve diagnoses across multiple fields of medicine.
• ⚠️ Privacy concerns remain a challenge, as illustrated by legal actions against DeepMind over inappropriate use of NHS patient data.
• 🔒 New AI technologies like Bitfount aim to improve data privacy while enhancing patient care by allowing data to stay within hospitals.
• 🏥 AI can potentially speed up the testing of medical procedures, such as simulating heart valve replacements before surgery.
• 💡 Empowering clinicians to develop their own AI models could lead to new discoveries in disease patterns and more personalized treatments.
• 💻 Virtual trials using AI can significantly reduce the time and cost of clinical trials, accelerating the development of new medical devices.

Q & A

What is the main medical problem highlighted in the script?

The main medical problem is the growing number of patients and a shortage of doctors to treat them, leading to delays and inadequate care in healthcare systems worldwide.

How can AI potentially revolutionize healthcare?

AI can transform healthcare by diagnosing and treating patients more efficiently, making medical procedures and trials faster and more effective, and handling large volumes of patient data that human experts cannot manage in a timely manner.

What specific eye condition is discussed in the script, and how has AI been involved?

The script discusses age-related macular degeneration, the most common cause of blindness in the UK and US. AI has been developed to diagnose over 50 types of eye diseases quickly, significantly improving diagnosis speed compared to human doctors.

What are some of the risks associated with delayed eye care as mentioned in the script?

Delays in eye care can result in irreversible blindness, as some patients are not seen and treated promptly due to overwhelming numbers of appointments, as highlighted by cases of patients losing their vision while waiting for care.

How does the AI system improve the diagnosis of eye diseases?

The AI system can analyze retinal scans within seconds, identifying various disease features much faster than human experts, who may take hours or even days to complete the same task.

What are some concerns about using AI in healthcare, as discussed in the script?

Concerns include threats to patient privacy, especially in cases where AI companies like Google DeepMind have mishandled sensitive medical data. There are also worries about accountability and interpretability when using AI models, as they often operate as 'black boxes' with complex, opaque decision-making processes.

How does the collaboration with machine learning startup Bitfount aim to address data privacy issues in AI healthcare applications?

Bitfount's technology acts as a secure switchboard, allowing queries to be sent to data without the data leaving its home location, such as a hospital. This helps maintain patient privacy and reduces the risk of data misuse while still enabling AI-driven analysis.

What potential benefits does AI offer in the development and approval of new medical treatments?

AI can speed up the testing and approval of new treatments by using privacy-preserving techniques, automating data analysis, and reducing the time required to conduct trials, ultimately bringing better medical solutions to patients faster.

How is AI being used to recognize patterns that are invisible to human experts?

AI has been developed to recognize patterns in retinal scans, such as identifying a person’s gender, which is something human experts cannot do. This ability could lead to further discoveries in disease patterns and biomarkers.

What are virtual trials, and how do they benefit the testing of new medical devices?

Virtual trials use AI to create digital replicas of patients, allowing multiple variations of new procedures or technologies to be tested without posing any risk to real patients. This approach speeds up trials, reduces costs, and helps identify the best treatments more efficiently.

Outlines

00:00

🩺 AI in Healthcare: A Solution to a Growing Medical Crisis

The global healthcare system is struggling with an increasing number of patients and a shortage of doctors. Artificial intelligence (AI) is emerging as a potential solution, with the ability to revolutionize diagnostics and treatment. AI can expedite the development and testing of medical procedures, offering significant benefits. Elaine Manor, a patient with age-related macular degeneration, shares her experience of regaining vision through successful treatment. However, many patients face delays in treatment, leading to severe consequences. Dr. Keen and his team have developed AI systems capable of diagnosing over 50 eye diseases swiftly, providing a much-needed answer to the overwhelming patient load. AI’s potential extends beyond ophthalmology, as it can process vast amounts of data rapidly, promising improvements in various medical fields.

05:01

🔒 Enhancing Patient Privacy While Integrating AI

The challenge of integrating healthcare data across different medical institutions hinders comprehensive patient care. Dr. Keen’s collaboration with Bitfont, a machine learning startup, aims to improve data connectivity and patient privacy. Bitfont’s technology functions as a switchboard, allowing data to remain at its source while enabling necessary inquiries, thus ensuring data security. This innovation could accelerate the approval of new treatments, addressing current delays in bringing ideal medical solutions to market. The healthcare sector anticipates significant growth in AI’s value, potentially empowering clinicians to independently develop AI models, fostering new discoveries in disease patterns and biomarkers. By 2027, the healthcare AI market is expected to be eight times larger than in 2020.

10:01

🩻 Virtual Trials: Revolutionizing Medical Testing

AI is transforming the testing of new medical devices through virtual trials, reducing the time and cost required for clinical research. Dr. Blackman and Professor Frangie at the University of Leeds have developed a method to create 3D digital replicas of patients, allowing multiple treatment scenarios to be tested on the same virtual model. This approach has proven to be as effective as traditional trials, but significantly more efficient. The team's virtual trials produced results comparable to three clinical trials in just three months, at a fraction of the cost. While virtual trials cannot replace human trials, they offer a powerful tool for accelerating the development and testing of new medical technologies.

Keywords

💡Artificial Intelligence (AI)

Artificial intelligence (AI) refers to the simulation of human intelligence by machines. In the video, AI is portrayed as a potential solution to the growing global healthcare crisis, where there are more patients than doctors. AI can diagnose diseases, analyze data, and even aid in the development of new treatments faster than humans, as seen with its application in retinal scans for eye disease.

💡Retinal Scans

Retinal scans involve imaging the retina, the light-sensitive layer at the back of the eye. In the video, AI is used to analyze these scans to diagnose over 50 types of eye diseases, providing results within seconds. The process is highlighted as a major advancement in healthcare, especially for conditions like macular degeneration, which can lead to blindness.

💡Macular Degeneration

Macular degeneration is a common eye condition that affects the central part of the retina (the macula), leading to vision loss. In the video, Elaine Manor, a patient with this condition, shares her experience of nearly going blind but was saved by medical intervention. The use of AI in diagnosing and treating eye diseases like macular degeneration is highlighted as a critical advancement.

💡Patient Data Privacy

Patient data privacy refers to the protection of personal medical information from unauthorized access. The video raises concerns about AI companies, like Google DeepMind, using patient data without proper consent, leading to legal challenges. It emphasizes the need for privacy-preserving techniques to ensure patient data is not compromised while still benefiting from AI technologies.

💡Virtual Trials

Virtual trials are simulations of medical procedures or treatments using digital models of a patient’s anatomy. In the video, virtual trials are shown to speed up the testing of new medical devices, such as heart valves, without putting real patients at risk. These trials can simulate multiple treatment scenarios, providing faster and more cost-effective solutions compared to traditional human trials.

💡Healthcare Ecosystem

The healthcare ecosystem refers to the interconnected network of hospitals, doctors, patients, and technologies involved in delivering medical care. The video discusses how AI can integrate and streamline various aspects of this system, such as connecting patient data across different healthcare providers, leading to more efficient and personalized treatments.

💡Deep Learning

Deep learning is a subset of AI that involves neural networks with multiple layers to analyze complex data. In the video, deep learning models are used to process retinal scans, enabling AI to diagnose diseases more efficiently than human experts. It also plays a role in recognizing patterns, such as identifying a patient's gender from retinal images, which humans cannot do.

💡AI in Medicine

AI in medicine refers to the use of artificial intelligence technologies to improve healthcare. In the video, AI is shown to revolutionize the field by diagnosing diseases, analyzing massive amounts of patient data, and assisting in the development of new medical devices. AI is expected to play a crucial role in addressing the shortage of healthcare professionals and increasing the efficiency of medical treatments.

💡Bitfount

Bitfount is a startup mentioned in the video that collaborates with hospitals to facilitate secure data sharing without compromising patient privacy. The company uses a system that ensures patient data remains in its original location while allowing researchers or doctors to query it. This innovation aims to improve the governance of medical data while addressing privacy concerns.

💡Age-Related Vision Impairment

Age-related vision impairment refers to vision loss that occurs as a result of aging, such as macular degeneration. The video discusses how this condition is a leading cause of blindness, particularly in older adults like Elaine Manor. AI's role in diagnosing and treating these impairments more quickly and accurately is emphasized as a significant medical breakthrough.

Highlights

AI has the potential to revolutionize healthcare by transforming the ways patients are diagnosed and treated.

Artificial intelligence can diagnose over 50 types of eye diseases as well as doctors, but in much less time.

Age-related macular degeneration is the most common cause of blindness in the UK and the US, impacting millions globally.

AI can analyze retinal scans in seconds, while human specialists may take hours or days.

An estimated 596 million people had distance vision impairment worldwide in 2020, and by 2050, that number is expected to rise by 50%.

AI's ability to process and analyze large amounts of medical data quickly can lead to more efficient diagnoses across various fields.

AI offers solutions to the growing global challenge of doctor shortages and overwhelming patient numbers, especially in eye care.

There are concerns about patient privacy, particularly after DeepMind, Google’s AI company, faced legal action over its use of NHS patient data.

Bitfount, a new AI platform, aims to improve patient privacy by ensuring medical data never leaves its original location.

Virtual trials powered by AI could significantly reduce the time and cost of testing new medical devices, speeding up the approval of treatments.

AI has been used to create a code-free deep learning model that can identify gender from retinal scans, something human experts cannot do.

AI could potentially empower clinicians to develop their own diagnostic tools, bringing them closer to patient needs.

The growth of AI in healthcare is expected to be eight times bigger by 2027 compared to 2020.

AI 2.0 aims to integrate prior knowledge, such as physics and physiology, with data, leading to more accurate and knowledge-driven healthcare systems.

AI in healthcare will continue to evolve, enhancing both patient care and the development of new medical treatments through more sophisticated models.

Transcripts

the world is facing a big medical

problem

a growing number of patients

and not enough doctors to treat them

so could artificial intelligence be the

cure i feel like i'm working at the

forefront of something that could

potentially be revolutionary to

healthcare in the future ai has the

power to transform the ways patients are

diagnosed and treated when we think

it'll become a game changer and to make

the testing of new medical procedures

more efficient and effective if we can

get devices that can be developed faster

better

and quicker

then there are huge benefits

[Music]

elaine manor is blind in one eye

she is a victim of age-related macular

degeneration

the most common cause of blindness in

the uk and us

when it went to my other eye i was just

terrified i was in bits i was weeping in

the rain and

thinking i don't want to be here

[Music]

for years the threat of losing her other

eye has loomed large

until a successful treatment enabled the

thankful 75 year old to see her way to

some high wire fundraising

i did the high sick wire in europe

i did the sky dive

and then i did the the wing walk

[Music]

but the doctor who saved elaine's eyes

says she is one of the lucky ones nearly

10 of all clinic appointments in the nhs

are for eyes that's nearly 10 million

appointments per year so to put it

brutally we're almost drowning in the

number of patients we need to see

and as a result of that there are some

patients unfortunately who go blind

because of delays in being seen and

treated a pregnant mother was left

almost completely blind waiting for care

she has since given birth and has never

seen her daughter's

face dr keen

believes there's an answer

artificial intelligence

he and his partners have developed ais

which can diagnose over 50 types of eye

disease just as well as a doctor

but do so

much much more quickly the ai system can

analyze the retinal scans within seconds

and it can delineate all of the

different disease features on the scans

a human expert would probably take hours

or even days to complete the same task

ai can help to address a growing global

challenge

in 2020

an estimated 596

million people had distance vision

impairment worldwide

of whom 43 million were blind

by 2050 both these figures are set to

increase by approximately 50 percent

a task that can take specialist doctors

hours now being done in seconds through

artificial intelligence and it's not

just eyesight ai's ability to mine and

analyze patient data far more quickly

than humans can mean diagnoses could

improve in many areas of medicine we do

more than a thousand scans per day in

the hospital it's a challenge because

where do we get the the human experts to

be able to review all those scans but

it's also an opportunity because that

huge amount of data is the perfect

substrate for the development of

artificial intelligence systems

but there are concerns

in particular threats to the privacy of

patients

deepmind google's ai company and one of

dr keane's partners has found itself

under fire

google deep mind the search giant's

artificial intelligence arm may have

received the personally identifying

medical records of 1.6 million nhs

patients here at the royal fair hospital

on a legally inappropriate basis

unrelated to its work with dr keane

deepmind is currently facing legal

action over its use of nhs data

yet if data can be better protected

ai has the capacity to make patient care

much better

and more efficient

so we have a world that is essentially

very much connected

but yet healthcare data is siloed

we can order a taxi from almost anywhere

in the world using our smartphones

but yet if we have a patient who comes

to an eye hospital like moorefields but

they're also attending a hospital

because they've got cancer we often

can't easily connect their data

dr keen hopes his latest collaboration

with machine learning startup bitfont

could not only join data dots better but

also improve patient privacy with fount

is is a kind of switchboard all we do is

essentially pass messages between

someone who wants to ask something of

the data set and the owner of the data

the data

never never leaves its home location so

if that data is held by a hospital

no data ever leaves the hospital bitfont

says this technology could have other

benefits like approving new treatments

more quickly and safely patients are

losing out a lot by the fact that ideal

medical treatments for them are not

coming through to to market

with the extra technical guarantees that

privacy preserving techniques like

bitfont can provide there's been a real

feeling around the healthcare ecosystem

that that could speed up a lot of those

governance processes

by 2027

ai's value in the healthcare market is

expected to be eight times bigger than

in 2020 growth could also be boosted if

clinicians reduce their dependence on

coders and start to develop their own ai

systems

this is really exciting because today's

retinal experts have been unable to

identify gender dr kira o'bern is part

of a team of clinicians who have managed

to do what google brain did in 2018

develop ai that can recognize gender

from retinal scans

something no human can do

a member of our research group developed

a code-free deep learning model which

accurately identified gender from

rational images

this is incredibly exciting because by

positioning clinicians

to develop their own models

independently it could really open the

door to further discoveries in both

disease patterns and disease biomarkers

a new generation of doctors believe

empowering clinicians in this way will

bring them closer to patients

generally it's the clinician that's the

healthcare workers working face-to-face

with the patient to understand what the

patients need best

so therefore i believe that if we can

allow them to independently develop

their own tools this will allow the

patient to remain at the very forefront

of everything

it's the world's first hand-held

battery-powered computer able to hold

thousands of data points i think that

we're potentially at a tipping point a

little bit like the tipping point that

we saw in the late 1970s in the

computing industry we had the

introduction of the first personal

computers if you empower people with

this technology even if primitive in the

beginning they will come up with

hundreds or even thousands of

applications that the engineers would

never have thought of

but some are skeptical about pinning

hopes too fast on ai

the issue is that ai models are

essentially black boxes and so what

happens is that when they're working

well they're working well and no

questions are asked but what happens if

a wrong decision is made what happens

when something goes wrong and how do we

really trace that back and ensure

accountability and guarantee

interpretability if we're using these

black box models

while it is early days for ai in

medicine it could also improve the

testing of new medical devices

a long time ago

a million years bc

everything was

absolutely free

in 2021 former part-time singer and

model patricia walker had an artificial

valve inserted into her heart to save

her life i needed a new valve

because the one that i had wasn't

working it was dripping this is why i

was feeling

the pain that i was getting the

exhaustion and it it was getting

worse

although patricia's operation was a

success

it was not without risk and the

cardiologist who inserted the valve says

ai can make new technologies like this

safer for patients

by creating virtual trials

if we can plan a procedure by simulating

that procedure in an individual patient

in a computer-generated model before we

go to the patient then we can get a much

better outcome in the patient without

posing any risk to that patient at all

at the university of leeds dr blackman

is collaborating with professor alex

frangie he is using machine learning to

automate the building of

three-dimensional digital replicas this

is the real one

and this is the fake one

but it's very difficult to the naked eye

to to pick that up

virtual trials mean multiple variations

of a proposed new procedure or

technology can be tried out we can test

different scenarios of treatment on the

same

anatomy and physiology of a given

virtual individual and that's something

which is not again possible to do with

conventional trials

we're also comparing different scenarios

of hypertensive and normal tensive

conditions

virtual trials do not replace human

trials

but they do speed up the time and reduce

the money required to identify the right

devices and humans to test

in 2021 the team at the university of

leeds found their virtual trials

produced the same results as three

clinical trials but much more

efficiently

each of those studies took between six

and eight years to to be undertaken and

they probably cost around 20 to 30

million each of them so what we showed

is that in this computational study that

the execution of it took about three

months so

that's you know less than

20 grand

some working in the medical world

believe a bright future lies ahead as

ais become more sophisticated and more

capable

in other words more intelligent ai 1.0

in my view is the ability to automate

tasks that otherwise

could be very boring or time consuming

or repetitive

ai 2.0

is the one that actually tries to look

at incorporating prior information on

the physics on the physiology in a much

more intimate manner with the data so

it's not just data driven but it's also

knowledge driven for healthcare it's

hard to see a future without ai

i strongly believe that one day

artificial intelligence

will renew this eye

thanks for watching i'm tom standage

deputy editor of the economist to read

more of our coverage of ai please click

on the link don't forget to subscribe

you