-
Notifications
You must be signed in to change notification settings - Fork 0
Home
This module is intended to analyze series of the MRI ACR large phantom. Series analysis include SNR, geometry (phantom diameter and length), image uniformity, ghosting and B0-homogeneity. These metrics are all calculated from the "plain" water images of the ACR large phantom. The module may be used for other phantoms, including the small ACR phantom, if configuration parameters are adjusted.
A Digital Reference Object (DRO) or synthetic test dataset is available for this module at wadqc/WADQC_MR_ACR_DRO. It is intended to test/validate this MR_ACR analysis module. The repo includes two DRO sample datasets and the Python code that was used to generate them. Also a selector export for this MR_ACR module is included. Note: make sure you install the latest MR_ACR module first, then import the Selector and choose the option to attach it to the already installed module.
Testing of release candidates
None
- numpy
- scipy
- skimage
- matplotlib
- pydicom
The acquisition protocol is prescribed by the ACR manual, which can be downloaded from the ACR website. Basically this is a multi-slice 2D acquisition with 11 slices and 1x1 mm in-plane spatial resolution. Slice thickness equals 5 mm and slices have 5 mm gap. Notes for specific submodules:
- GeometryXY: standard protocol series may be used for GeometryXY (gradient amplitude calibration). Note that these images should be corrected for gradient non-linearity in 2D (not 3D), and parallel imaging may be used. Sufficiently high readout bandwidth should be used to avoid excessive geometrical distortion due to B0 inhomogeneity.
- SNR, ghosting, image uniformity: do not use parallel imaging techniques such as GRAPPA, ASSET or SENSE during acquisition, because the noise is quantified in background ROIs. Also it is recommended to avoid any correction of geometrical distortion, as this would affect noise quantification due to interpolation of pixel intensities, most pronounced in the corners of the image where the ROIs for noise quantification are located.
- SNR of separate (uncombined) coil elements: if the scanner allows, this module may be used to detect disfunctional coil elements. Note that SNR may be highly variable between coil elements if the elements are at larger distance from the imaged slice.
- B0 homogeneity: the submodule is written for single slice series as input. Basically, a dual-echo gradient echo should be scanned, and reconstructed with magnitued and phase images. The B0-scans and storage format of magnitude and phase images in series is highly vendor-specific. Various options are implemented, see info at the config section on the selector page in WAD-QC Admin.
Templates for GE, Siemens and Philips configs are provided with the factory module.
The module is intended for use in dicom_study mode.
The config mainly serves two purposes:
- Series selection is implemented in each submodule and must be configured. Each submodule has a separate section in the config. Depending on the submodule, a variety of criteria may be configured, such as Series Description, number of images in the series, and the image to analyze within that series.
- analysis parameters, such as ROI size and position, type of B0 map etc. These are specific for each submodule and may be tuned to e.g. phantom dimensions. The config files provided with the factory modules for GE, Philips and Siemens scanners are intended for use with the large ACR phantom.
The values in the meta should be adjusted for your specific scanner and scanner software version.
Selection rules should be suitable to select the study (i.e. exam), thus should be non-specific at the series level.