[CCPPETMR/SIRF] error in STIR when using an attenuation image in the AcquisitionSensitivityModel (#103)

0 views
Skip to first unread message

Kris Thielemans

unread,
Feb 8, 2018, 12:33:39 PM2/8/18
to CCPPETMR/SIRF, Subscribed

when using an attenuation image

    asm = AcquisitionSensitivityModel(attn_image, am)
    acq_model = AcquisitionModelUsingRayTracingMatrix()
    acq_model.set_normalization(asm)

and using this for reconstruction, we seem to get

'ForwardProjectByBin: forward_project called with incorrect related_viewgrams. Problem with symmetries!\n' exception caught at line 579 of /Users/rich/Documents/OneDrive-UCL/Code/SIRF/Source/src/xSTIR/cSTIR/cstir.cpp

This is likely a STIR bug.

—
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHub, or mute the thread.

evgueni-ovtchinnikov

unread,
Feb 8, 2018, 1:35:08 PM2/8/18
to CCPPETMR/SIRF, Subscribed

I suspect you have not done am.set_up, have you?

Kris Thielemans

unread,
Feb 8, 2018, 4:17:24 PM2/8/18
to CCPPETMR/SIRF, Subscribed

that's not it I'm afraid

    am = AcquisitionModelUsingRayTracingMatrix()
    am.set_up(template, attn_image)
    asm = AcquisitionSensitivityModel(attn_image,am)
    asm.set_up(acq_data)

    # select acquisition model that implements the geometric
    # forward projection by a ray tracing matrix multiplication
    acq_model = AcquisitionModelUsingRayTracingMatrix()
    acq_model.set_normalization(asm)
    #acq_model.set_background_term(randoms)
    acq_model.set_up(acq_data, image)

by the way, the am.set_up shouldn't necessary I think (it should be called by the acq_model.set_up via asm.set_up if it all works ok).

evgueni-ovtchinnikov

unread,
Feb 9, 2018, 11:06:22 AM2/9/18
to CCPPETMR/SIRF, Subscribed

Closed #103.

Kris Thielemans

unread,
Feb 9, 2018, 6:52:55 PM2/9/18
to CCPPETMR/SIRF, Subscribed

Reopened #103.

Kris Thielemans

unread,
Feb 9, 2018, 7:07:48 PM2/9/18
to CCPPETMR/SIRF, Subscribed

I still have the same problem (with SIRF/devel and STIR/master)

Kris Thielemans

unread,
Feb 12, 2018, 11:46:47 AM2/12/18
to CCPPETMR/SIRF, Subscribed

I found the problem. It is due to a limitation in STIR which I will try to fix. The direct cause is that the mu_map.hv is slightly shifted w.r.t. the origin. This probably came about with the zooming etc. Clearly, STIR should be able to handle this, but at this point in time it doesn't.

An ugly work-around for now is to remove the offset keywords in mu_map.hv. Then everything runs ok. Let's hang on a bit before we do that though.

By the way, the clue was always in the warn.txt:

WARNING: Disabling symmetries in transaxial plane as image is shifted

it's just that we don't see these (as STIR writes a bit too many warnings at present)

Kris Thielemans

unread,
Feb 13, 2018, 3:59:04 AM2/13/18
to CCPPETMR/SIRF, Subscribed

Closed #103.

Kris Thielemans

unread,
Feb 13, 2018, 3:59:05 AM2/13/18
to CCPPETMR/SIRF, Subscribed

This is a STIR problem, so closing here

Reply all
Reply to author
Forward
0 new messages