0:00

So in this presentation, we're gonna talk

0:02

about image quality and artifacts in DBT.

0:07

So, um, like traditional screening mammography,

0:10

uh, DBT quality, of course, is highly dependent

0:12

on excellent positioning and good compression.

0:15

It's true of all mammography, so the expectations

0:18

are the same in the sense that you'll have

0:21

good compression in both your CC and MLO view.

0:24

On the CC view or MLO view, we

0:26

expect to see nipple profile.

0:28

Um.

0:29

We expect that the posterior nipple line will be

0:33

appropriate, uh, comparing between those two views.

0:36

Um, that there's adequate, uh, visualization

0:39

of the posterior tissues, including the

0:40

pectoralis muscle on the MLO view, and

0:43

that it has, um, a convex outward margin.

0:47

Of course, uh, positioning, just like with 2D

0:49

mammography, is also difficult in DBT and highly

0:52

dependent on highly trained and quality technologists.

0:57

In terms of the actual physics of DBT, there

0:59

are, uh, multiple QC tests as in 2D mammography,

1:03

and you can see here in this table, um, some

1:05

of those, the nuances of those are dependent on

1:07

the particular system that your practice uses.

1:11

Um, but it's the same kind

1:12

of ones we'd expect with 2D.

1:16

There's a mandated training for all, uh,

1:18

radiologists, technologists, and physicists in

1:21

terms of DBT, um, just like 2D mammography as well.

1:27

There are a couple particular artifacts which

1:30

are germane to DBT and, uh, in practice it's

1:33

helpful to be able to identify these artifacts.

1:36

Now, over time, imaging, uh, modality, uh, companies

1:42

have been working to improve their algorithms.

1:44

So if you looked at DBT exams from the earliest

1:49

stages of DBT, the reconstruction address.

1:53

Some of the artifacts, a lot of

1:54

those have been addressed over time.

1:56

For example, um, there used to be a lot of

1:58

problems with metal in the breast

2:01

or biopsy clips or surgical clips, something

2:02

like that with a kind of a large halo effect

2:04

around them that has been largely mitigated

2:07

over time by better reconstruction algorithms.

2:09

So really isn't a problem today anymore.

2:12

Um, I'll just go through a few kind

2:15

of important artifacts that are still, um,

2:17

I wouldn't say challenging, but it's

2:19

at least good to be aware of them.

2:21

The first is the S blurring or ripple artifact.

2:24

Um, it's related to a limited number

2:25

of acquired projections, right?

2:27

So in most systems you get about 15 projection images.

2:30

You can see this perpendicular to

2:31

the X-ray tube sweep direction.

2:34

So on this CC view here, you might see

2:36

it's in the medial part of the breast.

2:37

Those red arrows are the edges of the

2:40

skin may appear falsely thickened.

2:42

I don't know if this would really

2:43

lead you to believe that the patient

2:46

has skin thickening when they don't.

2:47

But it might at least draw your attention to it.

2:51

Um, coarse calcifications and clips will elongate,

2:53

giving kind of a slinky sort of appearance.

2:55

In some systems, this can be a little

2:58

bit more prominent than in others.

3:00

Um, for the most part, you know what it is.

3:04

Um, but if you don't have a particularly

3:06

high-resolution system, then identifying that

3:10

thing as a clip versus, let's say, a coarse

3:12

calcification can be a little bit more difficult.

3:15

They'll still both have this kind of slinky

3:16

appearance, um, and it can be a little bit distracting

3:20

or, you know, maybe potentially obscure some

3:23

very small calcifications or something nearby.

3:26

Um, as I mentioned, this appearance may be

3:28

uh, mitigated by processing algorithms, which—

3:34

Um, the stair-step and bright-line artifacts, um,

3:37

are types of truncation artifacts, meaning that

3:41

at the extremes of the, um, image acquisition,

3:45

there might be something that is in the way

3:47

of that projection image that is then therefore

3:50

projected into the tomosynthesis dataset.

3:54

And what I'm showing here is the stair-step artifact.

3:57

Um, and this is the patient's chin.

3:59

Uh, you can see the sort of rounded curvilinear

4:02

kind of line on the superior part of the image.

4:04

Um, you can see it just on the DBT slices.

4:07

Now, when you reconstruct this into the

4:08

synthesized view, probably won't see it,

4:10

might see it as a little bit of a line.

4:12

Um, that's okay.

4:13

It's pretty easy to figure out and usually it doesn't

4:15

obscure anything, um, to make it impossible to—

4:18

Now with better positioning,

4:19

you wouldn't see it at all.

4:21

Um, but it's okay.

4:21

Sometimes you just can't get the

4:23

best images based on patient factors.

4:27

The bright line artifact is when stair-step artifact

4:29

is reconstructed and appears there's a bright line,

4:32

usually just a solitary one line straight across.

4:35

Um, and again, doesn't usually

4:36

impact your interpretation too much.

4:40

We can see loss of superficial tissue resolution.

4:43

This usually occurs in patients

4:45

with large or dense breasts.

4:47

Um, it creates a need for higher energy X-ray

4:49

beam, and then at the edge of the tissue,

4:51

this causes tissue saturation and image loss.

4:54

We see it both in the DBT slices and SM views.

4:57

Um, if you require those, um, and I wouldn't say this

5:01

really limits our interpretation much, but if it's

5:03

extreme, it can sort of bleed into the image and it

5:05

just doesn't look like a very high-quality image.

5:07

So, for example, in this case here, um,

5:10

this right CC view, we see,

5:12

um, this portion of the nipple.

5:13

There are certain areas here where there's,

5:15

um, complete loss of image, um, with

5:17

this sort of streaky line appearance.

5:18

Occasionally, you'll see it in the edges

5:20

of the skin as sort of lines like that.

5:22

Usually it doesn't go very deep into the breast.

5:24

Again, it probably doesn't really limit

5:25

your interpretation, but it doesn't look

5:27

good and it's not a marker of good quality.

5:30

Motion, I'd say, is probably the most important

5:33

and common, uh, imaging artifact we see in DBT.

5:36

Now, it's also important and

5:38

common in 2D mammography too.

5:40

Um, but it can also be a little more subtle in

5:42

DBT and a little more difficult to recognize.

5:44

A lot of times this motion is related to the

5:46

fact that we have a longer acquisition time

5:47

in DBT, uh, may increase the frequency we

5:50

see motion or the extent of motion artifact.

5:53

Um, this may degrade the conspicuity of calcifications.

5:56

It could be hard to appreciate

5:57

on DBT due to out-of-plane blurring.

6:00

Um, we may see loss of resolution, skin

6:02

line, the axilla, the IMF, all the places

6:04

where we'd expect to see motion artifact.

6:07

And on DBT, we see this probably in two ways.

6:10

One is in the projection images.

6:11

If you send those images to PACS.

6:13

At our institution, we don't do that, so we

6:14

don't typically look at projection images.

6:17

But if you're scrolling through the DBT

6:18

dataset, you can actually see the skin line

6:21

move a little bit as you're scrolling through.

6:23

That's a marker of, uh, motion in

6:25

addition, as in this case here.

6:27

You can see this slinky art.

6:28

We can take advantage of the slinky artifact in this

6:30

case, and what we would expect to see is that if

6:33

the tube is moving across the breast in this plane,

6:38

that the slinky artifact should be straight

6:41

across the breast in that same plane.

6:43

But in this patient, we can see that the

6:44

slinky artifact starts out pretty straight

6:46

and then takes a little bit of an angle and

6:48

moves off, uh, sort of more posteriorly.

6:50

Same thing with this one here. We can

6:52

see that it has sort of a curvilinear

6:53

appearance, and all those mean that the

6:55

patient was moving during the acquisition.

6:57

So, uh, this might be a case that you would say, you

7:01

know what, it wasn't identified by the technologist.

7:03

This is clearly motion artifact.

7:05

I might be missing some calcifications here.

7:07

I'm gonna call them back for a technical

7:09

recall. I'm gonna have those images

7:10

repeated, but this can be difficult to see.

7:14

Um, and a lot of times, you know, when

7:15

technologists are acquiring images, they're

7:17

looking at the images or verifying the images on

7:19

a much lower quality, lower resolution monitor.

7:23

So it can be difficult sometimes to even

7:24

appreciate a subtle motion on those monitors,

7:27

where it becomes more glaringly obvious

7:28

under high-resolution PACS monitors.