Upcoming Events
Log In
Pricing
Free Trial

Introduction to Brain MR Spectroscopy, Dr. Doris Lin (9-15-26)

HIDE
PrevNext

0:02

Hello, and welcome to Noon Conference hosted by Medality.

0:06

Noon Conference connects the global radiology community through free live

0:10

educational webinars that are accessible for all, and is an opportunity to learn

0:14

alongside top radiologists from around the world.

0:18

Today, we are honored to welcome Dr.

0:20

Doris Lin for a lecture entitled "Introduction to Clinical Brain Proton

0:25

Spectroscopy." Dr. Lin is a professor of radiology at the John Hopkins University

0:30

of Medicine and an attending neuroradiologist at the Johns Hopkins Hospital, where

0:36

she specializes in diagnostic neuroimaging.

0:39

She's been involved in the clinical and research use of advanced MRI techniques,

0:44

including spectroscopy imaging. At the end of the lecture, please join Dr.

0:49

Lin in a Q&A session where she will address questions you may have on today's

0:53

topic. Please remember to use the Q&A feature to submit your questions so we can

0:57

get to as many as we can before our time is up.

1:00

With that, we are ready to begin today's lecture. Dr.

1:03

Lin, please take it from here.

1:06

Well, first of all, welcome. This is a little bit of an esoteric topic on

1:11

spectroscopy, but hopefully in the next 45 to 50 minutes, I could give

1:17

you a glimpse of what we use in clinical spectroscopy,

1:23

particularly in the central nervous system in the brain.

1:27

So clinical MR spectroscopy actually has been used for a long time.

1:31

It was first approved by FDA in 1992.

1:37

Three years later, there was an automated spectroscopy that was implemented on the

1:42

commercial scanner, the GE at the time, and that has the name of a probe.

1:47

And in 1998, there was an approved CPT diagnostic code

1:53

that allows this procedure to be billed and reimbursed.

1:58

And that actually gets covered by Medicare and a number of insurance companies.

2:04

But fast-forward to 2001, that was 20-plus years

2:09

ago, American College of Radiology and CMS actually

2:14

put a commission together and performed a study of our literature review and

2:18

determined that spectroscopy

2:22

could only be used as an investigational tool, which does not meet

2:27

evidence-based medicine criteria, which is quite a blow.

2:31

And in 2004, Medicare and a number of insurance companies followed suit and just

2:37

declared that this is an investigational tool, and it will not be covered.

2:43

Not until very recently, 2021, and everything's sort of kept it quite

2:48

silent. But there was actually a retirement of this declaration of a national

2:54

coverage determination for spectroscopy. What does that really mean?

2:57

It means that in certain cases, it actually,

3:03

indeed, whenever it's performed for a clinical indication, you actually could

3:09

get reimbursed by a number of insurance companies.

3:15

During throughout this time, the past 20 years when it was not reimbursed,

3:22

actually, spectroscopy was very alive and well.

3:26

There are a lot of research developments, and it also used clinically, and that is

3:31

actually demonstrated on this very nicely written review article in

3:37

radiology. And I want to highlight and draw your attention to this.

3:41

They conclude, abstract summary, saying that the usefulness is demonstrated

3:47

and very well established for a number of indications, including brain neoplasms, a

3:53

number of neonatal and pediatric disorders, particularly HIE, inherited

3:58

metabolic disorders, and other

4:03

conditions, particularly differentiating neoplasm from non-neoplastic inflammatory

4:09

lesions. And this is actually the recent

4:15

page available from the website published by Blue Cross Blue Shield and as a

4:20

medical policy. I want to highlight that here actually shows that they

4:25

are considering spectroscopy as clinically appropriate when

4:31

it actually could be used for a number of indications.

4:35

This actually expands to quite a few areas, but it's very similar to what we have

4:40

known about in terms of neoplasm, distinguishing recurrent brain tumor from

4:45

radiation necrosis, and inborn error in metabolism we often see in

4:51

white matter disorders, pediatric central nervous system abnormalities.

4:56

So I think that is always considered a high-yield application for

5:02

spectroscopy. But also, just in general evaluation of neoplasms, grading,

5:08

and distinguishing from other infectious or inflammatory etiologies and even

5:13

epilepsy. Those are actually deemed appropriate and can be covered by insurance.

5:20

With all that said, so what is an MR spectroscopy?

5:24

Well, this is something that we use the same MRI scanner.

5:29

But MRI, MR imaging, what we do is to actually

5:35

get the proton signal that is spatially

5:42

mapped into imaging by the variations of the frequency and phase.

5:48

Spectroscopy, on the other hand,

5:51

is something that produces sort of squiggly lines such as this, that we call

5:55

spectrum, MR spectroscopy, that has a lot of peaks and values,

6:01

and we actually assign those peaks into different metabolites.

6:05

So practically, it gives you a biochemical composition or

6:10

information of that area that you have selected from the voxel area of

6:16

interest. So this is totally different from MRI, but it actually offers some very

6:21

unique information that is not afforded by your routine MR imaging.

6:29

So on the x-axis, we have this ppm, parts per million scale.

6:35

That is used to identify a peak location, and that is also--

6:40

this is basically by the difference of two frequencies of two different peaks

6:46

divided by the operating scanner frequency in hertz.

6:51

And here by definition, we have NAA, which is an N-acetyl

6:57

aspartate defined at 2.02 parts per million.

7:02

Know that we use this as a reference because if you remember from your chemistry,

7:07

we do do MR.

7:10

Then that is a spectrum of a solution in test tube, and usually the reference is

7:18

methyl tetrazoline. But that is, of course, not present in the

7:24

normal brain or any of the tissue that we have.

7:28

So this is to serve as a reference and just give you a sense of what kind

7:34

of

7:36

quantity or concentration that we see.

7:38

Most of the metabolites, those chemicals that we see in the brain,

7:44

is actually in a very, very small quantity, ranging from 0.5 to 10

7:50

millimolar, that kind of concentration.

7:53

And if you compare to water, water has two protons, but then water is

7:59

in the concentration of 110 molar range,

8:05

and that is sort of at least 10 to the 4th or 10 to the 5th

8:11

magnitude greater than any of the metabolites that we're actually interested in

8:17

looking at.

8:20

Okay, so with that, you might say, "Well,

8:25

so how do we do this?" That is something that it is very important in spectroscopy,

8:28

is that water resonance is actually at 4.7 parts per

8:34

million. In addition to during the acquisition we suppress, we actually saturate

8:39

the water signal. We also just cut off the display beyond 4.5 parts per

8:45

million so that we could actually see those tiny amounts of metabolites in this

8:51

spectrum and display here

8:54

And as a reference, in terms of sensitivity of any of the chemicals that we could

9:00

detect using spectroscopy, I list the sensitivity that you might be able

9:06

to get from SPECT or PET. Of course, those are actually very, very

9:12

sensitive and really the true molecular imaging.

9:16

And then you also use radio-labeled ligands and receptors and the tracers to

9:21

actually detect that. But the beauty of spectroscopy is that you could perform in

9:25

the same scanner without any radiation, without any additive agents, and then you

9:31

get this kind of information from the scan.

9:35

The information that you could get, proton spectroscopy,

9:40

is simple here. Listed here, you have three major

9:44

metabolites that we see: NAA, creatine and choline.

9:50

NAA, or N-acetyl aspartate, is heavily present in axons, neurons, the

9:56

cell bodies, and neuronal processes.

9:58

So that is a great marker for neuronal quantity and health.

10:03

So if there's anything that destroys the neurons, whether it's atrophy or

10:09

ischemia and/or the dysfunction, you actually see decrease of

10:15

NAA. So this is a wonderful marker.

10:18

Creatine is a combination of creatine and phosphocreatine, and those

10:24

are the molecules that you know that

10:28

represent energy metabolites and that is also present.

10:31

A lot of time we consider it somewhat constant within the brain tissue, or at least

10:36

in the

10:39

voxel selection that we use, and that is actually present at 3.0 parts per

10:45

million.

10:46

And there's a sort of a downstream in choline is

10:53

immediately you could see that this is a membrane constituent.

10:58

So any process due to abnormal rapid membrane

11:04

turnover, cell turnover because of neoplasm or inflammation with the

11:09

destruction of the membrane, you could actually expect to see the elevation of

11:14

choline. So glial cells and demyelination and those are things that are seen.

11:20

And also in pathologic conditions, you could also detect lactate.

11:24

Lactate has a classic appearance. This is a rabbit ear.

11:29

Rabbit is sort of due to J-coupling of this

11:34

proton, and they split into a distinct distance that is varied by seven hertz,

11:41

and that is present at 1.33 parts per million.

11:46

Whenever you have anaerobic glycolysis, ischemia change,

11:51

or bioenergetic failure, you can actually see that the lactate is

11:58

elevated.

11:59

So that's simple enough. And then you have additional metabolites that we're

12:04

calling minor metabolites we could actually see in the short echo time.

12:09

And that is when you obtain at the TE less than 35

12:15

milliseconds. That includes myo-inositol, which is present in glial cells and

12:20

astrogliosis, glial proliferation, and more of that.

12:24

You also have osmolyte within the brain.

12:31

Here to the

12:34

downside of the shoulder of the NAA, you have a combination of glutamate, which is

12:39

an excitatory neurotransmitter, and glutamine, which is amino acid, as a storage

12:45

form of glutamate, and that also can be determined.

12:51

And finally, in this range of 0.5 to 1.3 parts per million, there are also lipids.

12:57

So this is sort of additional small metabolites in addition to the major ones that

13:02

you could see when you actually select a different echo time.

13:07

Now, this lists a number of other compounds that you actually could identify,

13:13

including the ones that if you look very closely, particularly if you could see at

13:19

high magnetic field strength that have a better separation.

13:23

And the very small number of compounds, including very

13:28

great interest of GABA, if you do editing techniques, you can actually see that.

13:33

And then a number of things are listed here in pathologic conditions such as

13:39

ketogenic diet, ketosis, or

13:44

phenylketonuria. And finally, I'm going to also touch upon 2-HG is

13:50

onco-metabolite that we actually could identify using spectroscopy.

13:54

Finally, exogenously administered dilantin

14:00

byproduct and alcohol you could also identify.

14:06

So how it is done. Let's move on to that.

14:10

So this, as I said, is done in the same MRI scanner whenever you have the

14:16

commercial scanners, but then it requires a license to have an MRI spectroscopy

14:21

package available. And it could certainly be done in

14:25

any of the clinical strengths of 1.5 tesla or

14:31

3 tesla.

14:34

The two major localization techniques include a single voxel, which is easy to

14:39

perform, and also spectroscopy imaging that sometimes you might hear that's

14:45

multi-voxel. You could also call it MRI spectroscopic imaging or

14:51

chemical shift imaging.

14:54

Single voxel is widely available, easy to perform.

14:58

You could use a short echo time, and then you get lots of compounds that could

15:02

offer you some indication. Usually, you get a voxel size of four to

15:08

eight, ideally eight

15:11

cubic centimeter. The drawback is that only a few brain regions can be interrogated

15:16

at each time.

15:18

But this is simple enough to do that.

15:21

And the typical protocol we use a PRESS, point resolved spectroscopy.

15:25

Spectroscopic imaging, I'm not going to show today.

15:28

I would say that it is less available, but certainly can be done in many of the

15:32

commercial scanners. The processing is much more intensive, but

15:38

the advantage is that it provides you lots of different areas to interrogate.

15:45

The nominal voxel size is much smaller, meaning that you have less of volume

15:50

averaging artifact. It gives you much better brain coverage, so anything that is

15:56

heterogeneous, such as a glioblastoma with the enhancing tumors and

16:02

necrosis in the surrounding infiltrating tumors, this is actually the best

16:08

technique if you could afford doing that.

16:12

This is the pulse diagram, just to show that the two

16:17

major protocols for single voxel localization of the PRESS, point resolved

16:21

spectroscopy, and also STEAM, stimulated echo.

16:26

This is by 90, 180, and 180 degree pulses.

16:31

And in general, PRESS is much frequently used because this produces a

16:37

signal-to-noise ratio almost twice as much compared to STEAM.

16:41

And here, just to give you a sense what multi-voxel is like.

16:45

So here, instead of a tiny area of a voxel area interest, you get a volume

16:52

of interest here. We start with the chemically selective saturation

16:57

pulse to get rid of the water signal, and then localization in three different

17:02

gradients. And then here we actually produce lots of little grids of each of those

17:07

voxels. You can actually put a cursor and get a spectrum.

17:11

So you get a lot of information from this volume interest, but of course, it takes

17:15

a longer time, and as I said, it is a lot more demanding for the

17:20

technologist to perform and also for anybody to process it.

17:26

So our role as a radiologist,

17:30

how do we actually do that? Of course, we wanted to see

17:36

when it is indicated and what kind of information we want to

17:42

extract from that, right? And then in terms of technique and how the major

17:47

decisions come into determining, first of all, whether you want to do single voxel

17:53

versus multi-voxel ... selection of a TE, the echo time, the

17:59

short one versus intermediate 140, or twice of that at 280

18:04

milliseconds. And as I alluded to, you get a lot more information,

18:10

minor metabolites from the short TE exam.

18:14

But long TE exam, you actually get very clean-cut peaks and actually

18:20

get some essential information if that is all you needed.

18:24

And then you just need to choose where you want to interrogate.

18:28

If there's a lesion, you should always get a contralateral normal brain as a

18:32

control. And in the voxel size, ideally you get large

18:38

enough voxels so that you get enough of the time, enough the signal to actually

18:43

produce a meaningful spectrum that you can interpret.

18:48

In terms of analysis, all the scanners actually provide post-processing, and then I

18:53

think the technologists actually produce that and then send it to your scan so you

18:58

can actually do that. A lot of time when we're reading this, we actually just do

19:03

qualitatively. In research field, and any time you wanted to be more precise or see

19:09

subtle changes, yes, there are some quantitative softwares that you could do the

19:14

post-processing, but you doing clinical use is very similar to reading a MRI

19:20

scan that I interpret mostly qualitatively, but also in

19:25

conjunction with the MRI itself and the clinical information.

19:30

That is important to note. This is an example that if you have a focal

19:35

lesion, for example, this in the right middle cerebral artery, you have a FLAIR

19:39

hyperintensity, a little expansile. You put a voxel there, and that is your lesion.

19:45

You get a spectrum, and then put a contralateral voxel, the

19:51

same volume, and then just put it over there, and then you get another

19:57

spectrum. So that is easy to compare with a single voxel.

20:01

Whenever there's no focal lesion, I still choose two different locations just to

20:06

get representation from different gray matter versus white matter.

20:11

One is the thalamus, deep gray. Sometimes you could also include the basal

20:16

ganglia. And another voxel is placed in the largest area that

20:22

you could find a white matter that is without getting into the volume averaging

20:27

artifact. So here is placed in the central semiovalie.

20:31

You could do the left or right because there's no focal lesion that's identified.

20:36

So that is my approach.

20:39

But this is actually sort of an exception, special scenario.

20:44

If you are concerned about hepaticity, you know that this is a reversible delirium

20:49

process. Yes, you very simply place a single voxel, but you have to

20:55

use a TE of a 35, 30 milliseconds, a short TE.

21:01

And you use the mesial paraxial gray matter and then another voxel in the white

21:06

matter to get a different sampling of the gray versus the white.

21:11

Why do we do this? Because this is the scenario that you have accumulation

21:17

of a number of neurotoxins from hepatic failure, and then

21:23

we are relying on short TE to see the elevation of a glutamine. Okay.

21:30

This is some tips. If you cannot find the lesion very well, yes, it is okay

21:36

to add the spectroscopy after you have performed the MRI

21:42

and have given contrast. It does not really affect metabolic levels.

21:47

And in general, if somebody who does not have the kind of patience to hold still

21:53

for that duration, that takes quite a few minutes for spectroscopy interrogation.

21:59

Single voxel is much faster and less motion sensitive, so go with that.

22:04

If major metabolites are the primary concern, you don't really need to get a short

22:09

TE. So I think that the long TE press protocol works very well.

22:15

A lot of time, I love the intermediate 140 milliseconds TE because in

22:21

that

22:25

you see the lactate being inverted and then that's distinguished from many other

22:29

metabolites that could be there, such as lipid. However, this is a note of caution.

22:34

At 3 Tesla, which is very widely available these days as a clinical scanner.

22:40

However, because the chemical shift dispersion, the TE 140,

22:46

intermediate TE, actually could now produce a pitfall because the lactate

22:52

gets sort of obscured. So if that is the primary concern, primary question,

22:58

avoid 140.

23:01

Another pause. I think this is quite a bit, but I think this is just a technical

23:06

aspect. So my default protocol that I do clinically at

23:11

Hopkins is a single voxel using PRESS, two by two by two centimeter ideally,

23:18

because if it is smaller, you actually get a lot of noise.

23:22

And use the two TEs, short and long, but know that I bypass 140 milliseconds to

23:28

go immediately to 280, and in two locations. Okay.

23:36

So I hope that you get from this latter half of this talk,

23:42

you see that how it is applied in clinical practice.

23:45

I think I want to convince you that it can be very useful in the diagnosis,

23:49

prognosis, and treatment response monitoring.

23:53

And I will show some examples, including ischemic

23:58

conditions, metabolic disorders, particularly in pediatric population, and

24:04

also the one that we actually could take more advantage of

24:10

is the tumor characterization. Not so much for primary diagnosis, but distinguish

24:15

neoplasm from non-neoplastic processes, and also radiation necrosis versus a

24:21

residual tumor. And also finally, the new onco-metabolite.

24:25

I think spectroscopy is going to make a difference in terms of how we practice.

24:31

So here we have a spectrum in a full-term newborn presenting with neonatal

24:37

seizures.

24:39

This is a T1-weighted image that's not going to show a lot that you might say, "Oh,

24:43

well, it's most watery." But putting a voxel over there, and this is a metabolic

24:48

map, you could see that there is elevation of lactate.

24:51

The spectrum shows this beautifully inverted doublet that's a classic lactate that

24:56

you see in TE of 140 milliseconds.

25:00

This is a choline. It is high, and NAA may be low, but know that in

25:06

a newborn, choline is actually dominant peak.

25:11

And we will revisit that. But this actually by assignment based on the

25:16

location parts per million, you can actually know what they are.

25:21

And this is indeed hypoxic ischemic injury that they're presenting with this, and

25:26

the lactate is a classic signature.

25:29

A number of studies have been performed quite a while ago, and it demonstrated that

25:34

MR spectroscopy is very, very sensitive to metabolic changes such as a hypoxic

25:40

ischemic injury. These days, we heavily rely on diffusion-weighted image for those

25:44

to identify those abnormalities, but know that in neonates,

25:50

the peak of HII or diffusion changes may not be apparent until

25:56

three to five days after the initial insult, and then it rapidly

26:02

reverses. But spectroscopy is actually sensitive immediately within 24 hours and

26:07

also

26:09

a lot of time you undergo after the acute event, you have a secondary energy

26:14

failure after reperfusion, and that is another time that you actually capture

26:19

marked elevation of lactate, which becomes the best signature.

26:26

With a prolonged hypoxia or ischemia, we could also gradually see NAA decreases.

26:32

And this study, this paper presented 20 years ago, actually showed that

26:38

the NAA concentration as well as the ratio of lactate to NAA could actually predict

26:43

outcomes. Here, I want to draw your attention again to remember that a normal

26:48

infant choline is at the highest peak.

26:51

And with this baby, with a small amount of lactate, actually turned out to have a

26:57

mild outcome and then could do favorably.

27:00

However, if you have the huge whopping amount of lactate and a decrease in NAA,

27:04

that typically corresponds to a severe outcome

27:08

Here is a five-year-old with a complicated history of adrenal insufficiency,

27:13

actually rely on hydrocortisone every night to sort of boost up the adrenal

27:18

function. But after missed one dose, presented something like this, a change of

27:24

mental status. FLAIR shows wide entire left hemisphere, but very

27:29

much isolated to the gray matter. FLAIR hyperintense, reduced ADC,

27:36

but DSC perfusion mean transit time is actually you have a fast arrival

27:42

of a blood to this area, and that also confirmed that you have increased cerebral

27:46

blood flow to the left hemisphere on ASL perfusion. So what is happening?

27:52

Now, spectroscopy actually shows that, wow, this is a

27:57

140 TE and lactate inverted beautifully shown, and then at a

28:03

short TE of 35 or 30, you actually operate here.

28:08

There's definitely indication of a bioenergetic failure.

28:12

So this is actually most likely a phenomenon due to

28:18

glucose being low, and you may consider metabolic stroke.

28:23

Or another scenario is that this is sort of a part of the routine response to

28:29

metabolic anomaly. You actually have seizure activities, and it's showing this kind

28:34

of appearance.

28:37

In metabolic disorders, a lot of important area of metabolism, you could expect

28:42

that spectroscopy could be very helpful, and it sort of has a distinct signature.

28:47

This is a three

28:49

and a half month old infant. Has some symptoms that are like seizures.

28:55

Pediatrician also noted that it's a pretty accelerated head circumference

29:01

growth, so big head. Here on MRI T2-weighted image, of course, everything's still

29:06

very watery, but here, definitely a lot of hyperintensity in the occipital white

29:12

matter and thalamus. Thalamus definitely abnormal signal.

29:17

So here, a voxel placed in the thalamus.

29:21

On the first glance, you may say, oh, the Hunter angle and the AA creatine

29:27

choline looks great, right? But remember that this is a young infant.

29:32

You do not expect to have such a high NA.

29:36

So this is indeed a signature, and it makes a beautiful diagnosis of

29:41

Aicardi-Goutières disease. Rare disease, but

29:46

very poor prognosis, and this actually nails the diagnosis very nicely by

29:52

in vivo spectroscopy.

29:55

There's another rare case.

29:58

Note that spectrum-- well, this is probably short TE.

30:02

You have a lot of ammonium metabolites, but here you have NA.

30:07

But here we seems to be missing something.

30:10

So after treatment, then you actually see the peak of creatine.

30:15

So this is a rare disease of agynine, creatine deficiency.

30:20

The enzyme deficiency could be treated by supplement.

30:27

So that is actually first demonstrated in the paper.

30:31

This is a graph from this review article in Radiology, but there are additional

30:36

references that actually talk about the identification by using spectroscopy, the

30:41

mechanism, and also the treatment response, and a very beautiful way

30:47

that to non-invasively and longitudinally follow therapy, and then in addition to

30:53

making a diagnosis.

30:56

Here I have a... This is one of the clinical cases I've encountered.

31:00

Four-year-old with sort of a scaly, like a fish scale, that kind of a skin changes,

31:05

very rough skin, development delay, and spasticity.

31:08

So almost like a classic clinical triad.

31:11

Has some white matter signal abnormalities, very non-specific.

31:15

What do we see on spectroscopy? So here, we actually single voxel in a

31:20

non-focal lesion matter. So one placed in the thalamus, the other one in the

31:25

central semiovalley. But both actually show that there is abnormal peak

31:31

at 1.3 parts per million without a doublet. So what is it?

31:36

This is classic location of a methylene group of a lipid,

31:42

and that confirms the clinical diagnosis of

31:46

Sjögren-Larsson syndrome.

31:50

Now, this is another rare inherited autosomal recessive genetic condition.

31:55

Of course, the diagnosis could also be made by genetic

31:59

testing. But here it actually just beautifully show that because an

32:05

important area of lipid metabolism of microsomal enzymes, it has a deficient fatty

32:11

aldehyde dehydrogenase producing those mobile lipid we could actually

32:16

identify from

32:19

just putting a voxel on the clinical exam.

32:24

Finally, an older person, 56-year-old, no other medical conditions,

32:30

although a remote history of some kind of a transient loss of consciousness.

32:36

Now presenting with vision loss

32:39

for almost eight days. Did not present until eight days later

32:45

with this kind of MRI flair hyperintensity in the right occipital lobe, definitely

32:51

involving the visual cortex, and very subtle

32:56

area of decreased diffusivity. But again, this is a sort of strange pattern for

33:04

infarct, right? And it's also isolated.

33:06

This is not your classic PCA infarct, but it certainly acts a little like that.

33:12

So a little bit four months later, presenting again, the right side of the

33:18

abnormality is evolving to atrophy, encephalitis, and then there's a new

33:24

abnormality extending into the left occipital as well

33:30

as the parietal lobes. And this probably corresponding to the focal motor

33:36

seizure affecting the right arm, the clinical presentation.

33:39

So this is the imaging appearance.

33:43

Immediately, you're talking waxing, waning, shifting that kind of signal

33:47

abnormality involving the cortex, right?

33:50

And looking like a seizure or a stroke-like thing. So what is it?

33:55

Spectroscopy is performed at the time.

33:58

What we do see is much elevated lactate. Note that doublet.

34:04

This is a

34:08

long TE, this is short TE exam, but both actually showed that very high

34:13

lactate, and that is classically seen in conditions like this of

34:19

mitochondrial disorder. It took another three months.

34:23

The person come back again with multiple other shifting areas, without

34:30

necessarily any restricted fusion.

34:33

Finally, deltoid muscle biopsy showed ragged red

34:39

fibers, which nails the final diagnosis of MELAS.

34:44

But from the clinical presentation, MRI imaging, and spectroscopy, you could

34:49

actually have made that brilliant diagnosis as well.

34:53

This is a mitochondrial dysfunction and causing this kind of a metabolic stroke, if

34:58

you like to call that.

35:03

The next

35:06

part, I'm actually going to touch upon tumor.

35:08

This is a big field, and it's quite dense, so I have to speed up a little bit.

35:13

So spectroscopy and tumor.

35:16

This is a nice diagram

35:19

put in this nature review. Choline is a product that

35:25

we actually major contribution from our diet.

35:30

But in the presence of a tumor, there's also-- well, and then this actually get

35:36

converted into a lot of membrane constituents, phosphatidylcholine, and

35:41

lysophosphatidycholine, et cetera.

35:45

But in cancer

35:48

pathology, there are a number of activation enzymes that shifts to

35:54

have elevated production of a number of products, including choline,

36:00

phosphocholine, and glycerol phosphocholine.

36:04

So those actually become a very useful marker for neoplasm, particularly

36:11

more malignant, very cellular ones.

36:14

If you have a tumor cell hypercellularity, here is a plot

36:20

of cell density actually nicely have a very nicely linear

36:26

proportion to the

36:29

the concentration of choline. And that is deriving from lots of different

36:34

pathologies, including oligodendro, sort of high-grade

36:39

glioma, low-grade oligodendroglioma, astrocytoma, and GBM.

36:45

GBM, remember that a lot of the time the cell actually becomes less dense because

36:49

of synchronicosis after stripping the normal blood supply, et cetera.

36:56

Here is one example of signal abnormality. A little bit faint contrast enhancement.

37:02

Diffusion trace image actually shows some areas of hypercellularity as indicated

37:08

by sort of decreased ADC values. On MR spectroscopy, you can

37:14

sort of see that in this lesion,

37:18

the marker of choline is much elevated. NAA is low.

37:24

And of course, we always wanted to have a contralateral site as a control.

37:30

You could see that as a normal display of NAA being the highest, if you want to

37:34

call it a hunter angle or not, and this is completely reversed, right?

37:38

So this is

37:41

basically the classic appearance of a neoplasm.

37:46

And this actually is still a low grade, but it's an oligodendroglioma,

37:52

very hypercellular, and that is reflected nicely on this case.

37:58

We don't really use spectroscopy as a primary means of making a

38:04

diagnosis of brain tumor. However, it has a lot of signatures that are worth

38:09

reviewing and understanding. These days, of course, all the tumors get biopsied,

38:14

then resected, and you have final pathology.

38:17

But I think this also helps us to determine what we might be able to glean from

38:23

spectroscopy as added information.

38:25

So here I have the spectrum of the parietal white matter.

38:28

This all come from this study by Hao et al.

38:32

NAA being at top of this is a short echo time, so you get a lot of metabolites as

38:37

well. Know that grade two astrocytoma, it is characterized by

38:42

higher choline level, lower NAA, but most notably you have elevation

38:48

of myelin oso. Remember that this is actually your glial marker, astroglial

38:54

proliferation that is most abundant in even low grade astrocytoma.

39:00

As the grade shift should be higher to the anaplastic, this may still be elevated,

39:06

but you actually start to see more of a lactate and lipid.

39:10

And then finally shifting to glioblastoma when you have synchronicrosis and normal

39:15

modular peripheral enhancement.

39:18

And this, you could see that a lot of normal brain metabolites

39:24

actually decrease, but then dominated by this lactate and lipid peak.

39:30

This kind of appearance becomes actually virtually indistinguishable from

39:34

metastases. Metastases, of course, you talk about secondary tumor.

39:39

We don't really see,

39:43

as opposed to the primary

39:45

gliomas, though the major metabolite is actually very low, but then it's marked by

39:51

a lot of necrosis and

39:54

lactate and lipid presence. So they actually look very similar.

39:59

And finally, meningioma is a neuroepithelial extra-axial tumor.

40:03

It has sort of something that's a sort of outlier, that kind of appearance, because

40:07

a lot of time people will actually see that the alanine gets elevated, but a lot of

40:13

time it get mixed in this pool of lactate plus lipid.

40:18

So using this, even limited number of different pathologies,

40:24

they were able to sort of plot this by looking at myelin oso to choline

40:30

ratio versus alanine, lactate, and lipid, and segregate into different

40:36

categories. Know that glioblastoma metastases very heavy on the peak of

40:42

lipid and lactate. They're nearly indistinguishable.

40:47

Whereas a primary glio tumors, astrocytoma, whether it's grade two or three, you

40:52

see elevated myelin oso. And of course, anaplastic astrocytoma, higher

40:58

grade, you actually also have more the presence of lipid and lactate.

41:04

And finally, meningioma falls into this lower corner that does not

41:10

really have any of the normal complements of the brain tissue, but

41:16

has elevation of alanine.

41:22

Here is a study of almost a clinical trial using machine learning and deep learning

41:28

from multiple sites in the European country.

41:32

And actually they plot and they start talking about the diagnostic accuracy.

41:36

They have 200 plus

41:39

training cases and 97 test cases. And they found that there was almost 90%

41:44

diagnostic accuracy in distinguishing most of the tumors versus

41:51

whether this is mets versus low grade,

41:55

or meningioma or GBM versus low-grade glioma.

42:00

However, the error rate or the lack of a diagnostic value is

42:05

the distinction between GBM and the mets.

42:10

So keep that in mind. A lot of time they look very similar in spectroscopy, whereas

42:16

the other ones, you can actually often use the spectrum information and then give a

42:22

reasonable diagnosis.

42:25

Here is an article from Dr. Barker.

42:27

It's from our institution, and this is actually comparing

42:33

MRSI and perfusion weight image to

42:38

separate malignant versus non-malignant brain lesions, whether it's a

42:44

low-grade tumor or inflammatory lesions.

42:48

And the ROC curve actually show that the discriminant function analysis show that

42:53

spectroscopy MRSI actually performed superiorly to

42:59

perfusion imaging alone. And that actually really offers a great value.

43:05

And those are a couple of values that use for analysis.

43:09

But of course, if you combine both perfusion and spectroscopy, you actually have

43:14

the superior performance here.

43:21

But in individual cases, how do we evaluate?

43:24

This is a case of a primary sinus lymphoma, edema, some patchy enhancement.

43:30

The CBV a little bit elevated, but choline elevation is a marker of a

43:36

neoplasm. As compared to this, also a lot of edema,

43:43

nondescript patchy enhancement.

43:45

CBV may be a little elevated here, but choline, creatine, NAA, everything's low.

43:51

So this is actually an example of a meningomyelocele.

43:54

It's very useful to distinguish that apart.

43:58

Now, I want to move into something that we have not moved to the clinical practice

44:02

yet, but this is emerging and very useful.

44:05

So 3-HG, as you know, hydroxyglutarate is a marker of IDH

44:11

mutations.

44:13

IDH these days defines a low-grade glioma, and it offers a diagnostic and

44:19

prognostic value, and also is a great therapeutic target.

44:24

A targeted therapy you can actually use to treat those people with a low-grade

44:29

glioma. And how does it, right? IDH is an enzyme that convert in

44:35

the TCA cycle from isocitrate to alpha-ketoglutarate, and that is a sort of a

44:40

non-mutated wild type. You actually have this just normal complements.

44:45

Whenever it's mutated, however, it produce a so-called onco-metabolite that is not

44:51

present in normal brain and not present in people without any

44:57

tumors. So that becomes a great signature. But why is it important?

45:02

Know that this graph from 2015, New England Journal of Medicine, you have a

45:06

survival curve, the Kaplan-Meier plot that of course, low-grade glioma with IDH

45:13

mutation, pretty good survival. And the best is the one with 1p19q

45:18

codeletion, which marks the oligodendrocyte.

45:23

And as opposed to the GBM, very poor prognosis, a very grim

45:29

outcome. But note this, this red line and the triangle, inverted triangle

45:35

is the histologically diagnosed low-grade glioma, but with the wild type,

45:41

without the IDH mutation. This is in the old days, diagnosis made purely by

45:47

histology. However, this survival curve is As grim

45:53

as the classic GBM.

45:56

And that also separate from this, the GBM, but actually these days we're talking

46:01

about IDH-mutated grade four tumor.

46:05

And this basically highlights the

46:10

subsequent revision of WHO classification from 2021

46:15

that by default, any tumor, even though it looks like low grade with

46:21

wild type IDH, automatically it is GBM.

46:27

So this is actually, in fact, developed. This is from Dr.

46:32

Choi from 2013 showing that with optimized spectroscopy using

46:38

TE of 97 milliseconds, you can actually separate this tiny metabolite of

46:44

2-hydroxyglutarate from the glutamate and glutamine.

46:49

And that could have an immediate impact in tumor locally in

46:55

the brainstem where biopsy is not possible or undesirable.

47:00

And here you actually can see. And also a great way to monitor disease

47:07

nowadays that we have approved targeted therapy available.

47:12

So this is a sort of emerging area, not yet widespread in these clinical practice,

47:18

but I do hope that this can be carried out.

47:24

I think I'm running out of time, but...

47:29

So Ashley, do you want me to go on? Or should

47:35

I sort of

47:37

wrap up?

47:41

You can feel free to wrap up and-

47:43

Okay

47:43

... take some questions if you'd like.

47:45

Sounds good. All right. I will skip this part, that I want to say

47:51

that spectroscopy has also made to the realm of

47:57

modifying the clinical practice and in causing a difference of

48:02

management. This is, you know that a lot of time you're talking about

48:07

diagnosis of a GBM. We use contrast-enhancing area to plan radiation

48:13

therapy. And that perhaps should be altered using

48:19

MS spectroscopic imaging where you could create a map, and the hot zone

48:23

demonstrates elevation of a choline to NAA ratio more than two.

48:28

So this is actually piloted in the study, a three-site study, that they basically

48:33

modify the conventional way of a radiation of this

48:39

clinical target volume of 60 gray and then a five-millimeter margin

48:45

and another five-millimeter margin as the standard care.

48:49

On top of that, they basically use spectroscopy to evaluate, to create a

48:54

new gross tumor volume and

49:00

boost the radiation dose to 75 gray.

49:05

So this was to really see if it's feasible and it's safe and then to

49:10

analyze the survival. So very quickly through this, I want to just go to the bottom

49:15

line, is that, yes, the clinically imaging-defined enhancing area,

49:21

the volume is actually small, average of a 1.58.

49:26

As compared to the new methods, it is a lot larger.

49:32

And with that,

49:34

the result is quite striking. The tumor recurrence, which often comes from

49:40

the resection margins, is actually reduced to 30% versus the standard care of a

49:46

60 to 70%. Of course, you have some sort of progression and some toxicities,

49:52

but it's not

49:55

unexpected, not excessive. And finally, in secondary

50:01

analysis, they show that the patients with a lower post-surgical choline-to-NAA

50:07

ratio has significantly superior survival outcomes.

50:10

Whereas if you just purely judge by the residual enhancement to guide

50:16

therapy, has little correlation or little significance in the progression-free

50:22

survival or overall survival. So this actually has been sort of

50:28

expanded to phase two trials, and I'm really eager to see that get into the

50:34

clinical practice.

50:36

Very fast, tumor necrosis, everything's down.

50:41

Recurrent tumor, you actually see elevation of the choline and the presence of

50:45

lactate. But know that a lot of time we have a mixed histology, and

50:51

however, know that the tumor to separate from necrosis, it is marked by those

50:57

at choline and lipid amount. But on individual cases,

51:03

it could be challenging. So astrocytoma,

51:09

a lot of post-treatment or the tumor cause edema, a little bit

51:15

enhancement. We also see a little bit in the contralateral side.

51:17

Spectroscopy

51:19

actually highlights that these areas are low metabolized.

51:24

That is just treatment change and edema.

51:26

Whereas this little area in the contralateral side, somewhat unexpected, actually

51:32

represents the presence of a tumor.

51:36

And this is actually confirmed aspect on a PET-CT.

51:40

Unfortunately, it was completely missed.

51:43

So those are things that to really keep in mind.

51:47

So finally, I get into the conclusion.

51:50

Hopefully, in this session, I can convince you that spectroscopy really is a

51:56

very useful technique. It allows a non-invasive evaluation of the brain metabolites

52:00

in vivo. It is easy to perform if you know how.

52:06

And then it really provides additional information that is not offered by

52:09

conventional MR by signal changes or even diffusion or perfusion, which

52:15

actually evaluate vascular density.

52:17

This provides a unique information about biochemical composition of the

52:23

tissue and the lesion. There are a number of wide range of clinical applications

52:29

that can be useful for diagnosis, prognosis, and treatment monitoring.

52:33

Hopefully, with this, you have a little understanding of when that may be

52:38

indicated, when you might want to incorporate into your practice, how to protocol,

52:45

and beginning to learn how to

52:48

provide your interpretation, add information, and

52:53

provide better value for clinical diagnosis.

52:58

That said, I want to acknowledge that I worked with Dr. Barker for a long time.

53:03

All this, I learned a great deal from him.

53:07

He does all the clinical processing, and then we have gone through the

53:13

discussion, and I'll provide the many of the slides and the information that I've

53:17

offered here. And I thank you for your attention.

53:22

Thank you for sharing your lecture with us today, Dr. Lin.

53:25

At this time, we will open the floor to any questions from our audience.

53:29

You may submit your questions to the Q&A feature. And Dr.

53:33

Lin, if you want to open up that Q&A box and look through the questions, see what

53:37

you want to answer.

53:39

Okay. So should I-- Okay, I will stop share, and I see in the chat

53:45

or-

53:48

There is a secondary Q&A box. It is hidden a little bit.

53:54

Okay. I got six. Okay. So the first one, why is the quality of a CSI voxel

54:00

degraded at the edge of the field coverage? Great.

54:03

You remember that chemical shift, basically, this is another name for

54:08

MRSI or multi-voxel. You are exciting a much larger area.

54:14

And spectroscopy, you're talking about it is very sensitive to field inhomogeneity.

54:19

You're exciting a lot of things. Whenever you are close to the scalp, that you have

54:25

a lot of lipid, and then that spills over, and then you grade the

54:30

area very prominently. That is also the reason that we meticulously

54:35

apply saturation bands on all the corners of the multi-voxel

54:41

planning. So that's a very good question, and the technical aspect that also

54:47

highlights the fact that it is very clinically challenging, but

54:52

then this is certainly doable.

54:56

From

54:59

Mahmoud Mourad. So why is spectroscopy used instead of other techniques such as a

55:05

biopsy? Well,

55:07

something that's obvious, yes, biopsy is always used for our final

55:13

diagnosis. However, the beauty of spectroscopy is that you get this information and

55:18

providing some diagnostic value without cutting open the brain,

55:24

the skull, and so it's completely non-invasive.

55:28

And if it could provide the genetic information prospectively, that could be very

55:33

useful. And I hope that I conveyed the fact that the 2HG,

55:39

to actually see the IDH-mutated tumor, that

55:45

is particularly important because sometimes they are located in the areas that you

55:49

don't want to biopsy or let alone complete resection. Right?

55:56

The second question is why are only three metabolites measured in spectroscopy?

56:01

Yes. That is when you're talking about a long echo time.

56:06

Those are measured metabolites that you could actually see.

56:11

You might say, "Oh, this is too few," but I showed you that there are many others

56:16

ones. I hope that is clear. It's more than that, but sometimes

56:22

just the three major metabolites and the changes, the ratio, and amounts could

56:26

actually provide you a lot of information.

56:29

And then it gives you a very clean cut spectrum.

56:32

And so I think a lot of time that is just as valuable.

56:36

But yes, you could get additional sort of shorter TE, get other minor metabolites.

56:42

And if you have the tool available shifting to seven Tesla ultra-high field,

56:48

you get even greater complement of metabolites information that

56:54

you could get.

56:56

So what is the predictive value? I think a lot of the things that the predictive

57:01

value, you could see that what I have shown you is very much coming from the

57:07

previous many other studies and group analysis.

57:11

So on the individual cases in clinical MRI, it still takes quite a bit of

57:16

experience and then learn how to sort of spot as a

57:22

qualitative manner. I hope that is also clear and that answer your questions.

57:28

Any role in spectroscopy in neurocognitive disorders?

57:32

Indeed, this is actually a very active field of investigation.

57:36

It has been applied to many different conditions. Alzheimer's is a hot topic.

57:42

Yes, we do see the elevation of some-- well, sometimes if you have

57:48

inflammatory components, glutamate, glutamine, particularly glutamate may be

57:53

elevated and the NA is down. But, I think those are

57:57

maybe subtle changes and then again, it requires sort of a more group, not just

58:02

individual, that kind of diagnostic value.

58:05

Any psychiatric disorders when MRI is completely normal, this actually has been

58:11

applied. Know that glutamate is excitatory

58:16

neurotransmitter. The fact that we can measure it, and it is sort of critical

58:20

during development and plays a role in cognition, memory, and

58:28

learning. So I think that definitely that gets altered.

58:33

It gets decreased in a number of neuropsychiatric disorders such as schizophrenia,

58:39

bipolar disorder, and OCD, et cetera.

58:42

So those are definitely very relevant areas.

58:48

And I think another question, Alzheimer's disease, yes, people do use that to

58:52

evaluate. It probably requires a lot more data to actually pull together to

58:57

actually show on an individual basis, rather difficult.

59:03

So Carla, I have one minute. Spectroscopy of GBM with all

59:09

metabolites low but lactate, and then how could you differentiate from radiation

59:12

necrosis? Yes. A lot of time there's a lot of overlap, and that is a difficulty.

59:19

I hope that from the later group analysis, you could sort of see that indeed

59:24

there's a lot of overlap. But detecting even the slightly higher choline

59:30

and the presence of lactate and lipid, that usually you

59:36

actually do see that it is a residual tumor or neoplasm over there.

59:44

Do you depend on spectrum of edema to differentiate single MET from GBM?

59:50

So that is something that you could utilize,

59:57

the difference between GBM and METS.

60:00

METs actually induce a lot of edema causing flare

60:05

hyperintensity. But GBM is an infiltrating glioma that the

60:10

surrounding flare hyperintensity actually indicates

60:16

tumor infiltration. So very often you could indeed use that to determine.

60:21

But that information is probably better answered if you

60:27

have access to a multi-voxel exam to look at the heterogeneity in that

60:33

larger field.

60:36

Okay. Anything else? What are the conditions where spectroscopy is not useful, for

60:41

example, in areas of bleed and those type of situations? Yes.

60:45

Vikram, thank you for that question.

60:48

I think in this talk, I specifically highlight the utility

60:53

of spectroscopy and where it could shine and it sort of

60:59

becomes a poster child of everything, right?

61:02

It sounds very wonderful, but there are a number of pitfalls.

61:06

Technically, it could be challenging analysis and all that.

61:10

As you said, where there's a lot of blood, blood causes susceptibility artifact,

61:16

distortion of a local homogeneity of the magnetic field.

61:19

Immediately, it's going to affect the way that you are going to get spectroscopy

61:25

done. Shimming is a

61:28

very important factor that you have to get good quality spectrum, and that is from

61:33

technical aspect. But if the lesion is too small in an unfavorable area such as

61:39

skull base, close to fat, scalp, orbit,

61:43

and in the presence of a bleed, a lot of time it could fail.

61:47

And then very often I do make the decision that I'm not going to attempt it because

61:51

it's going to be futile.

61:56

Okay, I think I answered all of them, right? But I also went over time.

62:01

Sorry about that.

62:03

No worries at all. Awesome job getting to all those questions.

62:07

Thank you for that lecture and for answering all those questions.

62:12

Okay. All right. Thank you. And if there's additional questions, if you

62:18

have any that actually you wanted to forward to me, I could also reply afterwards.

62:26

All right, great. Thanks to everyone for participating in our noon conference today

62:31

and asking all those great questions.

62:33

You can access the recording of today's conference and all our previous noon

62:36

conferences by creating a free account.

62:38

We'll also email out a link to the replay later today.

62:43

Be sure to join us next week on Thursday, September 24th at 12:00 PM Eastern, where

62:48

Dr. Pamela Kettwaru will deliver a lecture entitled Fetal MRI.

62:52

You can register for that at medallie.com and follow us on social media for updates

62:57

on future noon conferences. Thanks again, and have a great day.

Report

Tags

Neuroradiology