n o r d i c B r a i n E x
Tutorial – DSC Module
Please note that this tutorial is for the latest released nordicBrainEx. If you are using an older version please upgrade.
NordicNeuroLab AS
Møllendalsveien 1
N-5009 Bergen, Norway
Phone: +47 55 70 70 95
Email: [email protected]
m a k i n g f u n c t i o n a l M R I e a s y
nordicBrainEx Tutorial – DSC module
Date modified: 2019-02-20 Page 2 of 17
Contents 1 Loading of perfusion data................................................................................................................................... 3
2 Coregistration ..................................................................................................................................................... 4
3 Interacting with the results ................................................................................................................................ 4
3.1 Bolus arrival time ........................................................................................................................................ 5
3.2 General use ................................................................................................................................................. 5
3.3 Merge CBV for export to neuronavigation ................................................................................................. 7
3.4 Saving DSC data .......................................................................................................................................... 8
3.5 Spike detection ......................................................................................................................................... 11
4 Parametric maps............................................................................................................................................... 12
4.1 Calculation of parametric maps ............................................................................................................... 12
4.1.1 Leakage correction ........................................................................................................................... 14
5 Perfusion settings ............................................................................................................................................. 15
5.1 DSC settings .............................................................................................................................................. 15
5.1.1 Motion correction ............................................................................................................................ 15
5.1.2 Leakage correction ........................................................................................................................... 15
5.1.3 Vessel removal.................................................................................................................................. 16
5.1.4 Normalization ................................................................................................................................... 16
5.1.5 Auto-detect noise threshold ............................................................................................................ 16
5.2 Output maps ............................................................................................................................................. 17
6 References ........................................................................................................................................................ 17
nordicBrainEx Tutorial – DSC module
Date modified: 2019-02-20 Page 3 of 17
1 Loading of perfusion data To start the perfusion analysis, on structural and one DSC series have to be chosen. If the DSC type is unknown,
right-click to set the type (Figure 1). It is possible to select several structural series, but only one DSC dataset for
each analysis. To open DSC settings, right-click and select Edit DSC Settings. For more details about the DSC
settings, see section 5 Perfusion settings on page 15.
Hit Next to start analysis.
Figure 1: To start perfusion analysis, select one structural and one DSC series. To set the type of DSC, right-click and select the correct type from the list. To open DSC settings, select Edit DSC Settings. Hit Next to start analysis.
nordicBrainEx Tutorial – DSC module
Date modified: 2019-02-20 Page 4 of 17
Figure 2: If you select series from different studies (time points), you will get a warning. Select yes, if you do want to load series from different studies.
If you select series from different studies (time points) when loading data, you will get a warning (Figure 2). This is
to assure that any loading of series from different studies is only done if intended.
2 Coregistration An important step in the analysis is coregistration of the DSC series to the structural series. This will ensure that
perfusion maps used as overlays are displayed in the correct position. Please see ‘Tutorial –Visualization and
Interaction’ for details about the coregistration.
3 Interacting with the results
Figure 3: Perfusion analysis is done and the normalized BV map is added as an overlay. VOIs are added in the tumor tissue the contralateral normal appearing tissue, and the mean normalized CBV value is measured. The intensity curve and histogram of the VOI is displayed.
nordicBrainEx Tutorial – DSC module
Date modified: 2019-02-20 Page 5 of 17
When the perfusion analysis is completed, the perfusion maps are generated and shown as thumbnail in the left
side of the panel. Add any map as overlay or underlay to investigate the output (Figure 3).
3.1 Bolus arrival time
The bolus arrival time is calculated automatically by the program, but can be edited by sliding the red line along
the time curve. Click to use the new time position (Figure 4).
Figure 4: The bolus arrival time is automatically calculated, but can be changed by sliding the orange line along the time curve. Use recalculate to update with the new position.
3.2 General use
• Display perfusion output maps as overlays or underlays:
- Right-click on the thumbnail
• Using the Volume-of-Interest (VOI) tool to:
- Measure statistical values on the parametric maps (Figure 5).
- Display histogram and intensity curves (Figure 5).
• VOI statistics, intensity curves and histograms are displayed by selecting the icons next to the VOI name
in the VOI list (Figure 5).
• Change CBV overlay threshold (Figure 6) to remove low CBV values from the visualized overlay.
nordicBrainEx Tutorial – DSC module
Date modified: 2019-02-20 Page 6 of 17
Figure 5: Two VOIs are added, one in tumor tissue and one in normal appearing tissue. The statistical values are measured in the two regions, and the intensity curves and histograms for the two VOIs are displayed.
nordicBrainEx Tutorial – DSC module
Date modified: 2019-02-20 Page 7 of 17
3.3 Merge CBV for export to neuronavigation
Similar to for BOLD, DTI and DCE results, it is possible to save the CBV result in a format that can be used in
neuronavigation systems using Merge. Settings for merge export can be defined when choosing Settings and then
Export results as... in the upper left corner of the Visualization interface. The BOLD activation maps can then be
exported in a format compatible with viewing on certain neuronavigation systems.
• Color export
This will export the BOLD activation maps as separate DICOM image series that can be viewed together
with the structural volume on various neuronavigation workstations.
• White pixels on greyscale
This will export the BOLD activation maps as DICOM image series where the activated regions are shown
as white pixels in the structural dataset.
Figure 6: CBV threshold can be changed in the DSC tab. Here, one can also use VOI filtering and merge to neuronavigation.
In order to do so, adapt your workflow to the following steps:
1. Do perfusion analysis and add the CBV map as overlay.
2. Change the threshold of the overlay by sliding on the CBV threshold slider (Figure 6).
3. Locate the area(s) of interest and add VOIs
nordicBrainEx Tutorial – DSC module
Date modified: 2019-02-20 Page 8 of 17
4. Select VOI filtering and use include/exclude on VOI(s) to locate the region relevant for merge (it is possible
to skip this step)
5. Verify the output in real time in the MPR.
6. Select Merge to create a new series.
7. Stop VOI filtering to continue.
While in VOI filtering mode, some of the application’s functionality is blocked. To access all functionality, VOI
filtering mode must be stopped (hit ‘Stop’, see).
Figure 7: Merge CBV for neuronavigation: Step 1: add the CBV map as overlay. Step 2: Threshold the CBV. Step 3: Locate the area of interest and create a VOI. Step 4: Select VOI filtering to include/exclude regions. Step 5: Verify the output in real in the MPR. Step 6: Select Merge to create new series.
Note that multiple VOIs can be included or excluded to include/exclude multiple regions from the merged series.
3.4 Saving DSC data
There are several different ways of saving the DSC data.
• Save the derived maps to the database:
- Right-click on the thumbnail and select Save
- Right-click on the thumbnail and select Save and send to send it to a remote entity (e.g. PACS)
• Create a new DICOM series with perfusion maps as overlay (see Figure 8):
- Add a perfusion map (e.g. blood volume) as an overlay
- Right-click in MPR and select Create snapshot, Slice selection or Slice all.
- This will create a new secondary capture slice package that can be saved to the database and sent to
PACS.
• If choosing Merge in the Interaction panel, the selected CBV overlaid on the structural datasets can be
merged into a new DICOM series.
• By right clicking on any of the three planes of the current volume in the MPR, choosing
– Create snapshot will open the Slice editor window of the current slice.
– Create snapshot of MPR will open the Slice editor window of the three current planes of the MPR,
as well as the visualized activations in the 3D viewer.
nordicBrainEx Tutorial – DSC module
Date modified: 2019-02-20 Page 9 of 17
– Copy will copy the current slice to the clipboard, so it can be pasted into other programs (like
word etc.).
See Figure 8.
• In the Slice editor window (See ‘Tutorial – Visualization and Interaction’), accessed by right clicking on any
of the three planes in the MPR and choosing Slice selection (and clicking on Slice) or Slice all, slices can be:
– Saved to the database
– Saved and sent to a remote entity (for example PACS)
– Added to report (Select ‘View slices’ by right-clicking on the new thumbnail after saving to
database)
– Saved as AVI-file (Select ‘View slices’ by right-clicking on the new thumbnail after saving to
database)
• Send, accessed by right clicking on series’ thumbnail in the Data panel will send the series to a remote
entity (for example PACS). Read more about saving sessions in ‘Saving data to database and PACS’ in
‘Tutorial – Handling Image data’.
• Add to report (bottom left of any plane in the MPR) can be chosen to add the current slice to Report (see
‘Report’ in ‘Tutorial – Visualization and Interaction’).
• Saving the session can be done by choosing File -> Save Session. This will save the entire session (all loaded and acquired datasets) to the database. The session name that you entered will appear in the session table on the Select patient data window. Read more about saving sessions in ‘Saving, loading and sending a Session’ in ‘Tutorial – Handling Image data’.
nordicBrainEx Tutorial – DSC module
Date modified: 2019-02-20 Page 10 of 17
Figure 8: Slice selection from the axial MPR has been selecting, displaying the VOIs and the VOI statistics, as well as the histogram for the CBV and the intensity curves for the DSC dynamic series for the two VOIs. This frame can be send to PACS by selecting 'Save and send'.
nordicBrainEx Tutorial – DSC module
Date modified: 2019-02-20 Page 11 of 17
3.5 Spike detection
Spike detection is a quality check by visual inspection of the dynamic series, which can be performed by the user.
The user can exclude those slices or volumes from the analysis which contain spikes. The spike detection is
normally done if there are "outliers", i.e. images/slices where the signal-to–noise ratio is low or when the results
of the DTI analysis have not induced the expected results. In that case you can have a closer look at the dynamic
series and exclude spike slices or volumes.
Figure 9: The spike detection interface for DSC in nordicBrainEx.
The spike detection interface (see Figure 9) can be opened by right-clicking on the DSC series’ thumbnail in the
Data panel of the Visualization interface and selecting Spike detection.
The DSC series can be displayed dynamically (all volumes for one selected slice or as a volume (all slices for a
chosen volume)). Use the slider to easily change the slice/volume. The curve in the upper right corner shows the
maximum pixel value of all the images to the left. If you have a slice/volume with high pixel values, this will show
on the curve. If you have detected a spike slice, click on the slice in order to mark it, and then choose Discard slice,
or Discard volume if you want to discard the entire volume at the specific time point. The following border colors
are used in the Spike detection interface:
nordicBrainEx Tutorial – DSC module
Date modified: 2019-02-20 Page 12 of 17
• A green border indicates the current slice/volume.
• A yellow border indicates that a slice/volume is marked to be discarded.
• A red border indicates that this slice/volume has been saved as discarded in a previous session.
Slices/volumes can be restored by choosing Restore slice or Restore volume (please note that after you re-run the
analysis, the slices/volumes cannot be restored without re-loading the data). When you have marked all
slices/volumes you would like to discard, click on Save. You will first be prompted with a warning telling you that
discarding slices will modify the pixel values, and then a question if you want to re-run the analysis. If answering
yes, the analysis will be redone without the discarded slices. All marked slices will then be replaced by the average
of the corresponding slice from the two neighboring volumes.
4 Parametric maps
Figure 10: The parametric maps are displayed under derived data on the left side of the visualization screen. Right-click to interact (view as underlay / overlay, save to database, view slices etc.).
4.1 Calculation of parametric maps
The resulting parametric maps appear as icons on the left side of the screen under “Derived Data” (Figure 10).
nordicBrainEx Tutorial – DSC module
Date modified: 2019-02-20 Page 13 of 17
The parametric maps that are obtained are calculated using standard formulas and assumptions. The first step is
to obtain the contrast agent concentration as a function of time. It is obtained using the assumption that the
change in relaxation rate is proportional to the concentration of contrast agent: CArR =*
2 . Combining this
with the relation between MR signal and relaxation rates,
,)(
ln
=)(
0
*
2TE
tSI
SI
tR
the concentrations can be found. Then the following parametric maps are obtained:
• Normalized blood volume (nCBV)
The pixel values are initially set equal to dttRrCBV )(= *
2 . If leakage correction is turned on, the pixel
values are then leakage corrected (see explanation of leakage correction below). Finally, the map is
normalized.
• Relative blood volume (rCBV)
Same as nCBV, but not normalized.
• Uncorrected nCBV
This map displays the normalized blood volume without leakage correction.
• Uncorrected rCBV
This map displays the relative blood volume without leakage correction.
• Leakage map
This map displays the degree and type of leakage. Pixels close to zero (black) represent low degree of
leakage. Positive (red) or negative (blue) pixel values represent T1-dominant or T2-dominant leakage,
respectively.
• Mean transit time (rMTT)
The relative MTT values are calculated from the following relation: dtR
dtRtrMTT
*
2
*
2=
• Normalized Blood Flow (nCBF)
The blood flow values are calculated based on this relation, followed by normalization.
nordicBrainEx Tutorial – DSC module
Date modified: 2019-02-20 Page 14 of 17
rMTT
rCBVrCBF =
• Relative Blood Flow (rCBF)
Same as nCBF, only without normalization.
• Time to peak (rTTP)
Relative time of maximum concentration expressed in seconds.
• Vessel Map
Displays the vessels found during the vessel removal.
4.1.1 Leakage correction
The method used in nBX for leakage correction is described in the article by Bjørnerud et al (2011). In brief, the
method uses the tail of the residual function to estimate the leakage. The residual function is defined through the
equation
𝐶(𝑡) = 𝐴𝐼𝐹(𝑡)⨂𝑅(𝑡)
where 𝐶 is the measured concentration, 𝐴𝐼𝐹 is the artierial input function and 𝑅 is the residual function. The
residual function describes the fraction of contrast agent that is left in a voxel at the time t after a pulse of contrast
agent enters the voxel. The concentration is thereby equal to the convolution between the input concentration,
the arterial input function, and the residual function.
Basically, if the residual function does not approach 0 after a long time, there is leakage. The leakage may be a T1
type or a T2 type, giving a residual function ending above or below 0. Either way, the leakage correction is done
by finding the mean value of the residual after a preset time 𝑇𝐶 and multiplying it with the time that has passed
after 𝑇𝐶. The value found is subtracted from the initial estimate of the CBV.
In the case leakage correction is done, the flow is recalculated as the peak of the residual function, TTP is set equal
to the time of the residual peak and MTT is recalculated as the ration between blood volume and flow.
nordicBrainEx Tutorial – DSC module
Date modified: 2019-02-20 Page 15 of 17
5 Perfusion settings
Figure 11: In perfusion settings, the user can decide which preprocessing and analysis settings to use, and which output maps to display.
Perfusion settings (Figure 11) can be opened from the loading frame by right-clicking on the DSC series, or from
the visualization interface by right-clicking on the DSC source data thumbnail. In the perfusion settings, the user
can decide which preprocessing and analysis settings to use, and which output maps to display.
5.1 DSC settings
The settings for DSC are global settings, meaning that changes done at one time will be saved and applied to all
future analysis until the settings again are changed.
5.1.1 Motion correction
When checked this turns on motion correction of the DSC image series, meaning that all the image volumes are
coregistered to the first image volume. These parameters are saved internally so that the user can display them at
a later stage by right-clicking on the icon of the perfusion series and choose Motion Correction Results. The
coregistration is performed using an iterative method based on a measure of similarity. Step sizes are calculated
from the gradients of the similarity measure (the mutual information metric), and once the step size of the rotation
and translation parameters is small enough, the iteration process is halted, and the estimated rotation and
translation parameters are applied to the source volume using a cubic interpolation approach. Due to the limited
geometrical resolution of such image series only in-plane rotations and translations for axial slices are done.
5.1.2 Leakage correction
In tissue where contrast agent is leaking out of the vessels into the extra- or intracellular space the blood volume
will not be correctly estimated. The remedy for this phenomenon is what is termed as leakage correction. If not
corrected for, this leakage can lead to either under- or overestimation of the normalized blood volume.
Leakage correction will correct the output maps mean transit time (MTT), time to peak (TTP), blood volume (CBV)
and blood flow (CBF). Additional explanations can be found in Section 4 Parametric maps on page 12.
nordicBrainEx Tutorial – DSC module
Date modified: 2019-02-20 Page 16 of 17
5.1.3 Vessel removal
A patented method for vessel removal has been implemented in the software that identifies pixels that most likely
are affected by the contrast agent in a vessel. These pixels will not be included in the formation of the output
maps. The Vessel Map contains the pixels excluded.
5.1.4 Normalization
Normalization is done by defining the normal tissue in the volume, and dividing all pixels on the value of the normal
tissue. The normal tissue in a normalized CBV map should thus have an average value close to one. The normal
tissue is found by segmenting the brain using a clustering algorithm. First, pixels with values lower than the noise
level threshold are excluded. The remaining pixels are segmented based on their baseline signal intensity, and the
pixels with the highest values are assumed to be lesion and are excluded. To exclude blood vessels from healthy
brain tissue, a second segmentation is performed based on the already calculated BV values, where the highest
values are excluded as blood.
The output maps blood volume (CBV) and blood flow (CBF) are normalized. When normalization is done, the
average CBV value of the normal tissue is equal to 1, while tumor tissue typically has a higher CBV value.
If normalization is used, CBV and CBF output maps are indicated with a ‘n’ prefix for ‘normalized’. All maps that
have not been normalized are indicated with a ‘r’ prefix for ‘relative’.
5.1.5 Auto-detect noise threshold
If checked, an automatick noise threshold is performed, and if unchecked, the user can set the noise threhold
manually using a slider (Figure 12). The noise threshold affects the performance of the normalization and leakage
correction, so a warning will be given when manual noise threshold is selected. If brain tissue is removed by the
nosie threshold, the normalization and leakage correction might not work optimally (i.e. noise threshold should
not be used for segmenting the data).
Figure 12: Noise-threshold can either be detected automatically, or the level can be changed using a slider.
nordicBrainEx Tutorial – DSC module
Date modified: 2019-02-20 Page 17 of 17
5.2 Output maps
In the Output maps tab (Figure 11), the user can select which output maps to display. Details on how these
parametric maps are calculated are given in section 4 Parametric maps.
6 References
Bjornerud A, Sorensen AG, Mouridsen K and Emblem KE (2011) T1- and T2*-dominant extravasation correction in
DSC-MRI: Part I – Theoretical considerations and implications for assessment of tumor hemodynamic properties.
Journal of Cerebral Blood Flow & Metabolism 10:2041-53.