# Data-driven arterial input functions

[This article was first published on

Want to share your content on R-bloggers? click here if you have a blog, or here if you don't.

I haven’t devoted any time/effort/code to the extraction and parameter estimation of arterial input functions (AIFs) for DCE-MRI in either the vignette or the submission to **dcemri**, and kindly contributed to R-bloggers]. (You can report issue about the content on this page here)Want to share your content on R-bloggers? click here if you have a blog, or here if you don't.

*JSS*. Frankly, I wanted to communicate the main features of the

**R**package

**dcemriS4**and data-driven AIFs just haven’t made the list… up ’til now. So let’s assume one wants to fit a parametric model to observed contrast-agent concentration time curves (CTCs) obtained from a DCE-MRI acquisition.

The function extract.aif() requires a seed voxel to perform a three-dimensional region-growing algorithm based on correlation. I use FSLView, but feel free to select the voxel using your favorite piece of software. The code below uses the RIDER Neuro MRI data from the NBIA (National Biomedical Imaging Archive) that has been (a) converted from DICOM to NIfTI, (b) brain extracted by applying BET on the 15-degree multiple flip-angle acquisition and (c) converted from signal intensity to contrast agent concentration in all brain voxels. The main steps in estimating a parametric AIF are

- Read in the contrast-agent CTCs (a four-dimensional NIfTI file)
- Read in the acquisition times (obtained from the DICOM data)
- Use extract.aif() with the seed voxel and correlation threshold
- Fit a parametric model (Orton et al. 2008) to the average CTC from the seed-growing algorithm

The thin black lines correspond to individual CTCs from the voxels, the red line is the average CTC used to fit the parametric model and the green curve is the fitted model. Note, in order to fit the entire time series a compromise was struck between fitting the first pass of the contrast versus the tail. It may be necessary to modify the range of acquisition times in order to emphasize the beginning or the end of the AIF, depending upon which aspect of the dynamic acquisition is more important.

Contrast-agent concentration time curves from voxel (128,75,9). |

The four parameters that characterize the AIF model may be obtained via

To

**leave a comment**for the author, please follow the link and comment on their blog:**dcemri**.R-bloggers.com offers

**daily e-mail updates**about R news and tutorials about learning R and many other topics. Click here if you're looking to post or find an R/data-science job.

Want to share your content on R-bloggers? click here if you have a blog, or here if you don't.