The ACR geometric accuracy test was measuring from slice 7 rather than slices 1 and 5 as per the ACR guidance. I have:
fixed this, so that the measurements were coming from the correct slices
combined the separate functions for ACR geometric accuracy slice 1 and 5 (the previous difference was that the slice 5 function was taking diagonal measurements too but the slice 1 function was not). The function now has slice index as an input and if the slice index is 4 then the diagonal measurements are made.
I then realised that the mask was being created incorrectly on slice 1 of the GE data (this was not previously an issue as the slice was being created from slice 7). This was because one of the pixel values outside the phantom was above the masking threshold value. I have changed the threshold value so that it is no longer picking up background noise pixels and including these in the mask. The images on the report look acceptable but some future work could be to test this with more GE data. I then changed three of the unit tests accordingly, as there were very minor variations in the results following this change.
The ACR geometric accuracy test was measuring from slice 7 rather than slices 1 and 5 as per the ACR guidance. I have:
I then realised that the mask was being created incorrectly on slice 1 of the GE data (this was not previously an issue as the slice was being created from slice 7). This was because one of the pixel values outside the phantom was above the masking threshold value. I have changed the threshold value so that it is no longer picking up background noise pixels and including these in the mask. The images on the report look acceptable but some future work could be to test this with more GE data. I then changed three of the unit tests accordingly, as there were very minor variations in the results following this change.