[Intl_DxMedPhys] MUGA EF discrepancies?

Tue May 20 18:44:01 EDT 2025

Thank you everyone.

I got my answer. And being assimilated into the hive🚀, I’ll share what I learned with you here:

The original question I received from a Radiologist was: If we use the numbers in the red box, why don’t we arrive at the EF that Siemens software gives (orange box)?

[cid:image005.png at 01DBC9AB.773898B0]

The formula they were using was EF = (EDC-ESC)/(EDC-bkg)

which makes (as Bill pointed out) too many assumptions. So what is the formula Siemens uses? Looking up the Help Menu on the scanner gave me:

[cid:image007.png at 01DBC9AD.5A51D840] (I put the parentheses for clarification)

which is the same formula Bill mentioned.

If we plug those numbers in the red box into this formula, we get the 84.2% that the software gets.

However the story doesn’t end there: 84% was a bit too high for my liking.

So I redid the ROI placements myself. I specifically moved the BKG and SYSTOLIC ROIs,… and voila:

[cid:image008.png at 01DBC9AD.5A51D840]

Which is much closer to the Echo value the patient previously got (68%).

Subjective? Perhaps. But I think the BKG ROI the techs had picked was not a true background. The BKG has to be (consistently) placed near but not overlapping LV or lung edge, in a low-noise area (e.g., lateral chest wall).

Another point of interest:

On the same page where you see the TAC, Siemens also gives you a Table of counts. Those are the CORRECTED counts, which means you can directly take the MAX value from the table and subtract the MIN value from the table, and use the formula EF=ED-ES/ED to directly get the EF number (so no messing around with pixel areas and such). The CORRECTED is the bottom yellow curve seen above.

OK.

Live long and prosper y’all.
[cid:image006.png at 01DBC9AE.C34CBBC0]

Nima

[cid:image001.png at 01DBC9AA.7712DDB0]
Nima Kasraie, PhD, MSc, DABR, DABSNM
Lead Physicist, Children’s Medical Center Dallas<https://urldefense.com/v3/__https://www.childrens.com/__;!!KGKeukY!0VU2W9u4Y6LZypxVsSmNuEhJODJbZFEtKies1u_q-pE0qiWTT8YVcCtFA6h2u0g-s-LNka9vvT3mHqNKqLTEkTd4v8T3e-GehPO6jeApMokyKyGGzjsHZw$ >
Associate Professor, Department of Radiology, UT Southwestern<https://urldefense.com/v3/__https://profiles.utsouthwestern.edu/profile/187443/nima-kasraie.html__;!!KGKeukY!0VU2W9u4Y6LZypxVsSmNuEhJODJbZFEtKies1u_q-pE0qiWTT8YVcCtFA6h2u0g-s-LNka9vvT3mHqNKqLTEkTd4v8T3e-GehPO6jeApMokyKyG1AiUAGQ$ >
Phone: 214-648-7978
[cid:image002.png at 01DBC9AA.7712DDB0]

From: Bill Erwin <ervster57 at gmail.com>
Sent: Tuesday, May 20, 2025 3:35 PM
To: Nima Kasraie <Nima.Kasraie at UTSouthwestern.edu>
Cc: intl_dxmedphys_wd_osu_list at lists.osu.edu
Subject: Re: [Intl_DxMedPhys] MUGA EF discrepancies?

Nima, The EDbkgdC and ESbkgdC values cancel out only if the ED and ES ROI areas are identical, which should not happen. If you move the bkgd ROI around, it would not surprise me that the EF is affected due to a change in the average bkgd count

Nima,

The EDbkgdC and ESbkgdC values cancel out only if the ED and ES ROI areas are identical, which should not happen.

If you move the bkgd ROI around, it would not surprise me that the EF is affected due to a change in the average bkgd count value (bkgd ROI total counts / bkgd ROI area).

If you calculate both EDbkgdC and ESbkgdC (using Siemens displayed values for average background count and ED and ES ROI areas), and then use them in the formula as I have stated above, what do you obtain?

On Tue, May 20, 2025 at 2:04 PM Nima Kasraie <Nima.Kasraie at utsouthwestern.edu<mailto:Nima.Kasraie at utsouthwestern.edu>> wrote:
Thanks Bill.

Assuming the same bkg (as you say) simplifies the formula to the green formula I put up earlier. (the bkg values in the numerator cancel out. No?)

One thing is for sure: When I moved the BKG ROI inside the Siemens scanner workflow, the EF changed considerably. (I brought it down to 75%).

I’ve ruled out other possible suspects:

  *   Temporal smoothing or interpolation. The TAC looks nice and curved, not flat. And no noise.
  *   No gating issues
  *   No Arrhythmia.
  *   Patient had a strong squeeze

That leaves count distortions due to detector saturation or patient size (attenuation effects), and poor ROI definitions as suspect.

But I’m still not convinced. A 15% discrepancy is kinda high. (or is it?)

[cid:image001.png at 01DBC9AA.7712DDB0]
Nima Kasraie, PhD, MSc, DABR, DABSNM
Lead Physicist, Children’s Medical Center Dallas<https://urldefense.com/v3/__https:/www.childrens.com/__;!!MznTZTSvDXGV0Co!EUTO1PWtiSPY9N2XJxPk5380gWmaPWcdlr-9fQzfRWoZp7CJmuMhAP-SMkVPUp7zW5J0LKBoT9CRDgTGy-6b4p3CuuGV$>
Associate Professor, Department of Radiology, UT Southwestern<https://urldefense.com/v3/__https://profiles.utsouthwestern.edu/profile/187443/nima-kasraie.html__;!!KGKeukY!0VU2W9u4Y6LZypxVsSmNuEhJODJbZFEtKies1u_q-pE0qiWTT8YVcCtFA6h2u0g-s-LNka9vvT3mHqNKqLTEkTd4v8T3e-GehPO6jeApMokyKyG1AiUAGQ$ >
Phone: 214-648-7978
[cid:image002.png at 01DBC9AA.7712DDB0]

From: Bill Erwin <ervster57 at gmail.com<mailto:ervster57 at gmail.com>>
Sent: Tuesday, May 20, 2025 1:25 PM
To: Nima Kasraie <Nima.Kasraie at UTSouthwestern.edu<mailto:Nima.Kasraie at UTSouthwestern.edu>>
Cc: intl_dxmedphys_wd_osu_list at lists.osu.edu<mailto:intl_dxmedphys_wd_osu_list at lists.osu.edu>
Subject: Re: [Intl_DxMedPhys] MUGA EF discrepancies?

Nima, The correct formula for planar NM MUGA EF is supposed to be: EF(%) = 100 x (EDC-EDbkgdC - ESC-ESbkdgC) / (EDC-EDbkgdC) EDbkgdC = bkgd ROI total count x ED ROI area / bkgd ROI area ESbkgdC = bkgd ROI total count x ES ROI area / bkgd ROI
Nima,

The correct formula for planar NM MUGA EF is supposed to be:

EF(%) = 100 x (EDC-EDbkgdC - ESC-ESbkdgC) / (EDC-EDbkgdC)

EDbkgdC = bkgd ROI total count x ED ROI area / bkgd ROI area
ESbkgdC = bkgd ROI total count x ES ROI area / bkgd ROI area

Where the background ROI is typically drawn near the LV in the end systole image (assuming that area represents the projection of extra-LV tissue "background" activity), and the same bkgd total counts from that ROI at ES used for both ED and ES bkgd correction.  That is how it is supposed to be drawn on Siemens (either automatically or manually).  Automated method results are illustrated in Fig. 18-23 in Bushberg 4th Ed and Fig. 18-12 in the study guide (for those who have either one of those).  The quantitative data needed for the calculation are in the figure (for anyone wanting to verify the calculation).

I recommend comparing all the ROIs defined by the automated software with those defined manually.  Any substantial differences would certainly result in such a discrepancy.  (It would be a NM physician's call as to which ROIs are correct.)

Regards,
Bill

On Tue, May 20, 2025 at 12:26 PM Nima Kasraie via Intl_dxmedphys_wd_osu_list <intl_dxmedphys_wd_osu_list at lists.osu.edu<mailto:intl_dxmedphys_wd_osu_list at lists.osu.edu>> wrote:
Hello group, Need some collective wisdom from the hive: What are possible causes of the automated software overestimating EF, compared to manual EF calculation in MUGA studies? Does anyone have any experience with this? I have a case of discrepancy
Hello group,

Need some collective wisdom from the hive:

What are possible causes of the automated software overestimating EF, compared to manual EF calculation in MUGA studies? Does anyone have any experience with this?

I have a case of discrepancy between the automated EF value generated by the Siemens analysis software and a manual calculation of the EF, where the auto-generated EF value seems to be much higher than the manually calculated EF obtained using:

Manual EF calculation formula used: EF = (EDC-ESC)/(EDC-bkg)

Using this formula, EF is ~ 67.8%, which falls within normal range and correlates well with the patient's prior echocardiogram showing an LVEF of 67%. However, the software is giving an EF value of 82.7%, which seems high.

Perhaps a software correction I’m not aware of?

Nima

P.S. : Interesting side note:

ChatGPT responded saying a low background count can give elevated EF values:

[cid:image003.png at 01DBC9AA.7712DDB0]

which is actually false: EF goes down with decreased bkg counts (as one can see from the green formula above). And he’s using the same formula.

So there’s that. Ha!😊

And btw, Perplexity said the exact opposite:

[cid:image004.png at 01DBC9AA.7712DDB0]

Moral of the story: your model is only as good as the sources you train it with!

Nima Kasraie, PhD, MSc, DABR (D), DABSNM (I)
Associate Professor of Radiology
Medical Physics Division

Department of Radiology
5323 Harry Hines Blvd., Dallas, Texas 75390-9071
Phone: 214-648-7978
https://urldefense.com/v3/__http://www.utsouthwestern.edu/radiology__;!!KGKeukY!0VU2W9u4Y6LZypxVsSmNuEhJODJbZFEtKies1u_q-pE0qiWTT8YVcCtFA6h2u0g-s-LNka9vvT3mHqNKqLTEkTd4v8T3e-GehPO6jeApMokyKyEcb2ltzg$ <https://urldefense.com/v3/__http:/www.utsouthwestern.edu/radiology__;!!KGKeukY!25bfLAaxVTa-WLkAxQ_8HWNJaRXNj_RcF1yRTDrA0LM3YghSmtAuFFrteQA4M455_O0kbbCf9mCtlRJldwcg_FbJy8-vxMPGswuOyOiVdKzKtGzD2NBPJw$>

[cid:image001.png at 01DBC9AA.7712DDB0]
[cid:image002.png at 01DBC9AA.7712DDB0]

________________________________

UT Southwestern

Medical Center

The future of medicine, today.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.osu.edu/mailman/private/intl_dxmedphys_wd_osu_list/attachments/20250520/948e965b/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image001.png
Type: image/png
Size: 10985 bytes
Desc: image001.png
URL: <https://lists.osu.edu/mailman/private/intl_dxmedphys_wd_osu_list/attachments/20250520/948e965b/attachment.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image002.png
Type: image/png
Size: 31503 bytes
Desc: image002.png
URL: <https://lists.osu.edu/mailman/private/intl_dxmedphys_wd_osu_list/attachments/20250520/948e965b/attachment-0001.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image003.png
Type: image/png
Size: 82436 bytes
Desc: image003.png
URL: <https://lists.osu.edu/mailman/private/intl_dxmedphys_wd_osu_list/attachments/20250520/948e965b/attachment-0002.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image004.png
Type: image/png
Size: 49606 bytes
Desc: image004.png
URL: <https://lists.osu.edu/mailman/private/intl_dxmedphys_wd_osu_list/attachments/20250520/948e965b/attachment-0003.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image005.png
Type: image/png
Size: 163925 bytes
Desc: image005.png
URL: <https://lists.osu.edu/mailman/private/intl_dxmedphys_wd_osu_list/attachments/20250520/948e965b/attachment-0004.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image007.png
Type: image/png
Size: 4894325 bytes
Desc: image007.png
URL: <https://lists.osu.edu/mailman/private/intl_dxmedphys_wd_osu_list/attachments/20250520/948e965b/attachment-0005.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image008.png
Type: image/png
Size: 342448 bytes
Desc: image008.png
URL: <https://lists.osu.edu/mailman/private/intl_dxmedphys_wd_osu_list/attachments/20250520/948e965b/attachment-0006.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image006.png
Type: image/png
Size: 119212 bytes
Desc: image006.png
URL: <https://lists.osu.edu/mailman/private/intl_dxmedphys_wd_osu_list/attachments/20250520/948e965b/attachment-0007.png>