ironmap
Last updated
Ironmap is a script that receives preprocessed 3D+Time fMRI data and outputs one volume, where the value of each voxel is the inverse of the normalized T2* measurement. 1/T2* can be used as a proxy for intracellular iron (ferritin) and is particularly useful in the study of dopaminergic systems in the brain.
You can find the ironmap bash script (ironmap.sh) either in the BNC scripts directory on Oscar (/oscar/data/bnc/scripts
) or on github.
bash ironmap.sh -i <input> [optional flags]
-i <input>
: REQUIRED. Input one fMRI 3d+time NIFTI file.
-m <mask>
: OPTIONAL. Input a brain mask. An MNI brain mask is recommended. If none is provided, one will be created using afni's 3dAutomask.
-o <output>
: OPTIONAL. Output file suffix. Will be attached to the end of the input filename. Default is "ironmap".
-a
: OPTIONAL. Take the average/mean of each voxel across time rather than the median.
-v
: Print script version.
-h
: Print this help text.
Ironmap differs from other T2* mapping techniques (QSM, ME-EPI) in that it quantifies the inverse normalized T2* (1/nT2*) with a single-echo EPI time series. For background on this method, please refer to: Larsen & Luna, 2015; Sonnenschein et al., 2022. When designing an imaging sequence, it is important to choose an echo time that optimizes the T2*-weighted signal of your regions of interest. Although resting state scans are recommended, ferritin levels remain stable over time (unlike the BOLD signal) and can be quantified using task-based data as well. If both resting-state and task-based runs are available, test-retest reliability can be measured across different conditions (Price et al 2021). Additionally, the 1/nT2* ironmap can be created by aggregating data from all EPI runs/conditions.
Time averaging is an essential step in targeting the aspects of the T2*-weighted signal which do not change over time. However, there are multiple approaches to creating a single volume T2* map from a multi volume time series. Acquiring the median for each voxel across time is recommended, as it can reduce the impact of outlier volumes in the time series (Parr et al., 2022). It is also acceptable to take the average/mean over time. This script takes the median by default, but the -a flag can easily be used to take the average instead. The most important consideration is that the method is consistent across runs and participants.
First, at every individual TR, ironmap.sh normalizes the value of each voxel to the mean of the entire volume. Normalization of the T2*-weighted signal mitigates the effect of run to run variability within and between participants/sessions. In the final step, the inverse is calculated for the ease of interpretation; intracellular iron deposition and T2* have a negative relationship (time-averaged T2* values decrease as the amount of intracellular iron increases, Bartzokis et al., 2022). Some groups report their results without taking the inverse. In these instances, lower values reflect increases in intracellular iron.
Typically, the whole-brain ironmap calculation is followed by an ROI analysis that examines the mean ironmap values within specific regions of interest. Voxels across the brain should have an ironmap value of roughly ~1, with higher values associated with higher intracellular iron.