Skip to content

Naming Acquisitions for Automated Classification


This document details best practice for proper classification.

Instruction Steps

One of the first gears that is run when new DICOMs are loaded into Flywheel is the Scitran: DICOM MR Classifier gear. This gear extracts header information from the DICOMs to determine how to classify the image data. Proper classification has many benefits to how data can be used in downstream workflows and processes.

While it may seem like magic when it works, the classifier gear is only as good as the data it receives as input. One of the best ways to ensure proper classification is to name the acquisitions on the scanner in a descriptive manner. The SeriesDescription DICOM header field is where the acquisition name is stored in the DICOM file.

The table below shows some of the key strings that the classifier will look for in the SeriesDescription DICOM field to associate those data with the appropriate classification scheme.

T1 T2 Functional Diffusion
t1 t2 fmri dti
t1w func dwi
spgr bold diffusion
mprage resting

If one of those strings is not found within the SeriesDescription DICOM header field, the classifier will do its best with the context given, but with no guarantees of success. In those cases you should consider manual classification of your data. See Data-Classification for more information.