MR Image Loading and Denoising

To start parcellation of a brain PET series, select the Brain Parcellation tab of PNEURO and activate the MR button in the lower right for arriving at the MR page.

The MR page allows loading the T₁-weighted brain MR image of the same patient using the Load MR button. The image should be T₁-weighted, cover the entire brain and have a pixel size of about 1mm. After loading, the image should appear in radiological HFS orientation.

MR Image Cropping

The MR image is the basis for the parcellation. Experience has shown that problems may occur if the MR field-of-view is much larger than the brain as occurs for instance with sagittal MR acquisitions. Therefore, please use the Crop facility for reducing the MR data set to the relevant portion with skull and brain, but without the neck in the same way as the PET image is cropped.

MR Image Segmentation

The MR image will be segmented into gray matter (GM), white matter (WM) and cerebrospinal fluid (CSF). The algorithm uses four parameters:

Denoising strength	Denoising of the MR image may improve the segmentation of gray matter, white matter and CSF. If a Denoising strength other than None is selected, a non-local means denoising algorithm is applied which preserves structure boundaries unless the strength is too high.
Sampling	Density of pixels considered in the calculation.
Bias Regularization	Serves for compensating modulations of the image intensity across the field-of-view. Depending on the degree of the modulation, a corresponding setting can be selected from the list. The parameter to the right indicates the FWHM [mm] to be applied. The larger the FWHM, the smoother the variation is assumed.
Cleanup threshold	Procedure for rectifying the segmentation along the boundaries.

It is recommended to use the default settings and only experiment with other parameter values if the segmentation fails. The default settings can be recovered by the Button Default Settings Extras Tool button.

The actual segmentation is started with the Segment MR action button. Note that the denoising and segmentation process may take several minutes.