Hello, I have been using pylinac (v2.0) to perform CT scanner CATPHAN QA. I compared the results when the analysis is performed using the web application (AssuranceQA). I have found differences in Slice Thickness and low constrast resolution.
Results from pylinac were verified manually. Does anyone had this differences before?. Assurance QA calculated a slice thickness 2 mm lower than expected one.
Thank you!
Ruben.
Hi Ruben,
The version that Assurance QA uses is much earlier than the source version (v0.6). This is due to the use of scikit learn and scikit image being new requirements for newer versions. Heroku cannot handle what it calls total slug size of >300MB. When you add heavy scientific libraries this surpasses this requirement. Long story short, AQA is behind. The slice thickness is definitely the most finicky of the metrics. This isn’t isolated to pylinac. Use other commercial libraries and they will often give similar results. This is amplified if lower mA is used in the scan or if a larger FOV is used. Originally, I combined several slices together and then averaged to get the results, which indeed gave better and more consistent results, but I’ve been convinced to simply use a single slice, since that seems to be the standard practice.