Volume of interest and tumor segmentation

The selected VOI can either be the whole image volume, or a volume selected by seed growing or manually selecting voxels to be included in the analysis.
Whole image analysis: From the <Perfusion> tab, selected <Analysis scope><Whole image>. This is the default mode, and the entire perfusion image volume is analysed. The noise threshold in the data is then automatically determined as part of the perfusion analysis.

It is recommended to do VAI analysis for the entire volume. The segemented lesion mask can also be used for measuring statistical parameters on the maps.

Lesion segmentation is initialized by pressing  on main menu bar. When activated, the <Lesion segmentation > tab is shown in the right panel (figure 1). The functionality here is based on the <Pixel Edit > option already available is nordicICE, but has been extended /improved as follows:

• Full MPR interaction: seed points for seed growing can be placed in any of the orthogonal MPR views. Same for ‘manual’ placement of mask voxels (using the <add/erase> voxels option)
• Seed growing limited by selected VOI
• Range of additional morphologic operations:
• Open/close for removing ‘speckled’ pixels and filling in ‘holes’ in lesion mask (e.g. central necrotic regions)
• Erode/dilate options to expand/shrink final lesion mask

When applying seed growing based segmentation, the segmentation range is limited by the VOI selection, combined with the Segmentation range setting. VOI is defined, either by the icon on main toolbar, or by checking the <Set target VOI> button within in the Lesion segmentation tab. This defined the volume of interest which limits the range of the seed growing expansion in all three dimensions. The VOI rectangles will appear in the MPR views (figure 2) and can be adjusted in size position in any of the MPR windows and will be adjusted accordingly the orthogonal views.

Figure 1: Lesion segmentation / seed growing menu options.

Lesion segmentation is most performed using the following steps:

2. Place a seed point within tumor region in any of the three MPR planes using default settings. The resulting (3D) defined selected volume will be shown in the three MPR views (figure 2). If the resulting segmented volume is too large/small, try adjusting the <sensitivity > and <segmentation range > parameter

3. Erase/add voxels to mask in current plane using the add/erase pixel option and ‘draw’ voxels to be added/removed. Use morphometric operations, if required, to fill (close) in necrotic regions or expand/ shrink (dilate/erode) mask.

4. Note that multiple VOIs can be defined; i.e. seed growing can be defined within one VOI and the VOI can then be moved to a new position and a second seed growing operations can be done within this region.

Figure 2: Sample result of initial 3D seed growing from placing a single seed point within tumor region. The VOI sets the limitation of the seed growing extension

Once the lesion mask is defined and edited, it can be saved to file by pressing the <Save mask> option.

Loading lesion mask from file: A predefined lesion mask (a binary Nifti file) can be loaded from file using the <Load lesion from file> option. Once loaded, remember to press <Register> .  Also, a drawn lesion mask can be saved to file using the <Save> option.

NOTE: if image co-registration is performed, this should be done before a lesion mask is loaded from file to ensure correct positioning of the mask. The mask file is assumed to already be co-registered with the structural series prior to loading.

Saving lesion mask to Dicom database: The active lesion mask can be saved to the nordicICE DB as a binary Dicom object by pressing <Save mask to DB>. The mask will then be added as a new series to the patient series containing the scan from which the VAI module was started.