DCE analysis options

More options relating to the DCE analysis can be accessed from the <Options> button at the bottom of the window.

The following menu will appear:

Kinetic modelling

Choose between the following kinetic models (see 'Theory' section for more details):

  • Extended Tofts (3-parameter model)
  • Tofts-Kermode (2-parameter model)
  • Patlak (2-parameter model)
  • Two-compartment exchange (TCx) (4-parameter model)
  • Incremental model

The incremental model will use the best adapted model at any point. You can also choose to generate a model selection map, showing which model was most optimal on a pixelwise basis. The following code is used: dark blue= Vp only (no leakage); light blue= Patlak; yellow = Kermode (no Vp); green= Extended Tofts; red: two-compartment exchange (TCx).

Use iterative NLLSQ model: By default, the convolution integral forming the basis of DCE kinetic modelling (see 'Theory' section) is solved by matrix inversion, by re-casting the convolution integral to a matrix equation, yielding a fast and efficient way of estimating the kinetic parameters (see Murase, K Magn Reson Med 51(4) 2004). The convolution integral can, however, also be solved by standard iterative non-linear least squares (NLLSQ) methods whereby the sum of squares error between the data and the fitted function is iteratively minimized. Under certain conditions, the NLLSQ approach may provide more correct parameter estimates, but at the cost of much longer processing times. This option is therefore mainly suggested for use in the 'interactive ROI mode' and not for parametric mapping.

Descriptive parameter estimation:

Use this to set the time-range used for the AUC calculation. When Normalize parameters to AIF is selected, the values of descriptive parameters like AUC, upslope and washout are normalized to the corresponding values of the AIF.

Remove spikes in output maps: when this option is checked, outlier values (spikes) are removed from the output maps. Such spikes e.g. can occur due to unstable curve fitting conditions with noisy input signal.  You can further specify if outlier pixels should be set to zero or to the maximum pixel value in the image (after outliers have been removed). The Histogram cutoff value specifies the percent cutoff for outlier definition. if this value is set to e.g. 98% then the upper 2% of the pixel values in the output data is treated as outliers.

% enhancement threshold:  specifies the minimum amount of enhancement (signal difference between baseline and maximum value) required for a given pixel to be included in the analysis. If set to 0% then all pixels (above general noise threshold, if applied) are analyzed.

AIF-tissue delay:

The kinetic models used for DCE-analysis inherently assume that the AIF and tissue response contrast onset occur simultaneously. In practice, however, there may be a significant delay between the AIF onset and the tissue onset, and this delay me also vary between tissue types and between normal tissue and pathology.  nordicICE implements different strategies to correct for possible AIF-tissue delays:

Auto-detect: when selected, a global correction is applied to align the AIF with the global (average time-intensity curve) for the entire image volume.

Fixed delay: when selected, the specified fixed delay (in terms of number of image time-points) is used for the global AIF-tissue delay correction

Additional pixel-wise correction: when selected, additional pixel-wise delay correction is applied. Note that this option will increase processing time.

AIF options

Hematocrit correction factor: Specifies the factor used to estimate plasma concentration based on the whole blood signal response. This value is typically set equal to the average hematocrit in the study population.

Auto-AIF detection

The cluster technique used for automatic AIF detection can be optimized for various parameter settings, which can be specified here.

Population based AIF

Two different shaped population based AIFs are included. When 'Population based AIF' is selected in the main menu of the DCE module, the application will use the AIF specified here. A ‘population based AIF’ as used here is a standard AIF derived from a multi-compartment kinetic model of the expected shape of a contrast bolus in the brain following a rapid injection in a ‘standard’ person. The kinetic model used to obtain the standard AIF is described in van Osch MJ et al. Magn Reson Med.2003 Sep; 50(3): 614-22. The two different pre-defined AIFs differ in sharpness and maximum amplitude.  

Additionally, you can define your own AIF as a binary file and use this in the analysis. If specifying your own AIF from a binary file, the number of entries in the file must match the number of dynamic time-points in the raw data. You also need to specify the temporal resolution of the supplied AIF. If different from that of the dynamic data, the provided AIF will be re-sampled to match the resolution of the imaging data. The amplitude values of the AIF entries are assumed to reflect the concentration of the contrast agent at each time-point (in units of mM ).

You can also define a bi-exponential AIF where all parameters can be defined manually.

For all predefined AIFs as defined above, the relaxivity in blood (in units ofmM-1s-1) of the contrast agent used need to be specified since the AIFs here are represented in units of contrast agent concentration (mM )and need to be converted in the relaxation rates (1/R1 or 1/R2,R2*, depending on acquisition type). Default relaxivity values are provided based on ‘standard’ gadolinium chelates at 3.0 T. For T2/T2* based perfusion imaging, the actual in vivo blood relaxivity is more complex, and some studies suggest a quadratic dose-response. An option of model a quadratic dose-response (as described in vanOsch et al Magn Reson Med 49(6): 1067-1076).

 

Related topics:

Kinetic modeling theory

 

Logo Description automatically generated