0:00

So let's think about the mechanism of perceptual error.

0:04

So once again, these are the errors where you look,

0:06

but don't see where the lights are on nobody's home.

0:10

And, uh, somehow we see, so we, we fail to see something

0:13

that is later obvious, even to ourselves.

0:15

In retrospect, this has a certain rate of occurrence,

0:19

about two or 3%,

0:21

and it has a remarkably stable error prevalence since it was

0:25

first described by Garland in the 1940s.

0:28

And this is despite tremendous advances of technology

0:31

and knowledge being measured across a broad range

0:34

of practice situations, multiple generations of providers.

0:37

I mean, my practice today does not resemble Dr.

0:41

Garland's practice at all.

0:43

I don't think he would recognize what I do,

0:45

and I would think of what he did was quaint

0:48

and, and historic.

0:50

Um, but remarkably, we have the same error rate.

0:53

Um, it really suggests

0:55

that there's an underlying human factor,

0:58

an underlying neurobiological mechanism.

1:01

And, you know, we do have these biological tools

1:04

that have been handed down, you know,

1:06

for millions of years of evolution.

1:09

We operate using the eyes

1:10

and brain that developed

1:12

for very different reasons than sitting in front

1:14

of a monitor all day reading x-rays.

1:17

And, uh, you know, we have some amazing abilities.

1:20

Uh, we know

1:21

that you can detect an abnormality again from Dr.

1:24

Kiki's work, uh, with eye tracking, you know,

1:27

400 milliseconds.

1:28

Most radiologists will find the

1:29

abnormality on an, on an image.

1:31

Um, we know that there's unconscious

1:34

or pre-conscious detection

1:36

and filtering of things that we see.

1:38

We know there's tremendous pattern recognition

1:40

and pattern creation.

1:42

We have all these great superpowers,

1:43

but we also have some limitations.

1:45

And, you know, one of them is we we're, we fall prey

1:48

to certain, you know, classic illusions.

1:52

Like it's very hard to see this as both a duck and a rabbit.

1:56

This could be the rabbit's nose, his mouth and his eyes,

1:58

or this could be the duck's eye and his bill.

2:01

Um, and, uh, you know, the,

2:03

the duck rabbit illusion is very common.

2:05

This was a great paper. This was a few years back.

2:07

Um, Evans et all, uh, they basically got a bunch

2:11

of radiologists to look at mammogram images

2:13

that they flashed on the screen for half a second.

2:16

And, uh, the radiologist could of course, um,

2:19

pick out which ones, you know,

2:21

were from women who had cancer.

2:22

But the interesting thing is they were

2:24

of the opposite breasts, the

2:25

breasts that didn't have the cancer.

2:27

And, uh, so, uh, looking at the mammogram

2:30

of the opposite breasts 80% of the time, uh, uh,

2:33

just about the radiologist could pick it out.

2:35

Looking at that for half a second. It's amazing superpower.

2:39

They cannot tell you how they did it.

2:40

I can't tell you how they did it.

2:42

Um, but yet we still have this error rate.

2:44

And, uh, eye tracking, uh, studies have been very helpful.

2:47

Um, this again, is from Elizabeth Kapinsky from Emory.

2:51

Uh, the, uh, the dots show you where the,

2:53

where the eyes fell, you know, that,

2:55

that the foveal central vision and the, the red bars

2:59

Or the Cades between, uh,

3:01

and, uh, it shows you that the, the GA is tracking of how a,

3:04

how someone looked at a chest x-ray.

3:07

We know that they look at, they'll look at an x-ray, uh,

3:09

differently if they're tired, as opposed to they're fresh.

3:12

Um, so abnormalities are often detected in

3:15

less than a half a second.

3:16

And the gaze also tends

3:18

to dwell on areas, uh, where they miss things.

3:20

So, uh, they'll look at something, something multiple times,

3:23

their gaze will be drawn to it

3:24

and they'll, you know, they'll ultimately not see it

3:27

and afterward not remember, uh, having ever looked at it.

3:30

And, and certainly have no recollection

3:32

of there being anything there.

3:34

And you show them this thing that they missed

3:36

and say, well, you looked at this like eight times.

3:38

Uh, and, um, so, uh, there,

3:40

there is a subconscious process going on

3:43

that doesn't involve working memory.

3:45

And, you know, you know, again,

3:46

the lights are on, but nobody's home.

3:49

And, uh, you know,

3:50

radiologists are no better at

3:51

finding Waldo than anyone else.

3:53

It turns out that our skills, our,

3:55

our amazing superpowers are really only limited to the,

4:00

to our area of expertise.

4:01

And, uh, uh, you know,

4:04

we've all seen solution since childhood.

4:05

We know that the horizontal lines are the same exact

4:08

length, but it sure doesn't look that way.

4:09

Uh, this one's amazing to me.

4:11

This is called the Shepherd's Table illusion.

4:14

Both tabletops were exactly the same.

4:17

And I actually had to get a ruler and measure it

4:19

because I couldn't believe it, but it is true.

4:21

Um, this is a great illusion.

4:23

Uh, this is called the Fraser Spiral.

4:26

It sure looks like a spiral.

4:28

In reality, it's just concentric circles.

4:31

So now you know the secret. I'll take the circle away. Oops.

4:34

Looks like a spiral again.

4:36

Even knowing what it is cannot defeat the illusion.

4:39

This illusion is cognitively impenetrable

4:42

and, uh, you know, how many colleagues are.

4:44

Could you also describe as cognitively impenetrable?

4:49

Um, half the people will see this as moving.

4:53

Half will see it as static. I can't predict.

4:56

Um, here's a functional MRI of my own brain.

4:59

I'm thinking it'd be really great if we could figure out why

5:01

this, why this is happening.

5:03

That's sort of our moonshot. I.