Applying arterial input function (AIF) deconvolution

To generate semi-quantitative perfusion kinetic maps, the arterial input function (AIF) must be defined. Note that, although the AIF is defined and corrected for, the resulting maps can still not be considered to represent true perfusion (in MRI). This is primarily because the relationship between the measured signal change and the underlying contrast agent concentration is not known. Still, if proper definition - and correction for AIF is performed, kinetic indices may be obtained which should be independent of the AIF in a given patient or given examination allowing for comparison of kinetic values obtained in different patients or at different time-points in the same patient (see Theory section for more information). nordicICE performs deconvolution of the tissue response curves (dynamic curves for all pixels) with the AIF using the method proposed by Murase by default for extended Tofts modelling (Magnetic Resonance in Medicine 51:858–862; 2004). However a full non-linear iterative least square curve fit can also be applied, as described in Options.

Three different types of AIF can be used:

• Population based AIF. Different types of population-based AIFs are defined in Options.
• Automatic detection. This method will use a cluster analysis technique to search for AIF voxels in the entire brain, or a specified area. The number of AIF voxels can be specified. The detected AIF curves can be studied, and deselected, if pressing 'Show AIF curves'.
• From current ROI.

Note that DCE analysis cannot be run without defining an AIF. Thus, if no AIF is specified, the population-based AIF will be used.

Partial volume (PV) correction. This option can be used together with AIF from automatic detection or from current ROI.

Related topics:

DCE Theory