search

UCL / STIR

Software for Tomographic Image Reconstruction
http://stir.sourceforge.net/
Other
108 stars 90 forks source link

SPECTUB:Sensitivity appears inverted along x axis #452

Open danieldeidda opened 4 years ago

danieldeidda commented 4 years ago

I came across a repeatable issue. The sensitivity appears inverted along the x-axis when reconstructing SPECT data. As a consequence attenuation correction can be wrong. This happens in both master and in #445 but with inverted axis.

KrisThielemans commented 4 years ago

this is in #445?

KrisThielemans commented 4 years ago

can you please clarify if this was a new bug introduced in #445, or an existing one.

danieldeidda commented 4 years ago

can you please clarify if this was a new bug introduced in #445, or an existing one.

This happens in both master and in #445 but with inverted axis.

KrisThielemans commented 4 years ago

ouch. @RebeccaGillen worth checking with @danieldeidda what's happening.

Let's try to merge #445 soon. I think there was 1 thing outstanding, but can't remember... continue the conversation on that there.

danieldeidda commented 4 years ago

ouch. @RebeccaGillen worth checking with @danieldeidda what's happening.

Let's try to merge #445 soon. I think there was 1 thing outstanding, but can't remember... continue the conversation on that there.

I did not notice at the beginning because the phantoms I was using were quite symmetrical but as soon as I tried something slightly more complex it was visible

danieldeidda commented 4 years ago

I double checked with the master and I can reproduce the sensitivity flip along x axis.

image

danieldeidda commented 3 years ago

Unfortunately I came across the same problem again: I reconstructed data from GE OPTIMA and the attenuation appears inverted along x in all the patients

danieldeidda commented 3 years ago

Unfortunately I came across the same problem again: I reconstructed data from GE OPTIMA and the attenuation appears inverted along x in all the patients

I mean in the sensitivity image. but can definitely change how the reconstructed image looks like especially where there is air (lungs)

KrisThielemans commented 3 years ago

how do you get the attenuation? what file format?

I presume this means that the SPECT image (with atten) comes out x-flipped. This is what we'd need to fix. But let's be 100% sure about the attenuation.

danieldeidda commented 3 years ago

how do you get the attenuation? what file format?

I presume this means that the SPECT image (with atten) comes out x-flipped. This is what we'd need to fix. But let's be 100% sure about the attenuation.

The attenuation is estimated from the CT (dicom) using the STIR utility. The SPECT image is not flipped but the lung area is I noticed this when comparing vendor and STIR images. You can see below that the lesions and the liver are in the same place but the lungs shape appears inverted. when looking at the umap image (.v) used for the reconstruction it is inverted compared to the sensitivity and the reconstructed image

image

RebeccaGillen commented 1 year ago

Note that this issue is still present in the current version of STIR (v5.1.0). Tested on a Windows system with an asymmetric attenuation dataset: image

The sensitivity data are flipped in the x-direction: image