Table 2. Summary table of IQMs.

The 14 IQMs spawn a vector of 64 features per anatomical image on which the classifier is learned and tested.

Measures based on noise measurements
CJV	The coefficient of joint variation of GM and WM was proposed as objective function by Ganzetti et al. [30] for the optimization of INU correction algorithms. Higher values are related to the presence of heavy head motion and large INU artifacts.
CNR	The contrast-to-noise ratio [31] is an extension of the SNR calculation to evaluate how separated the tissue distributions of GM and WM are. Higher values indicate better quality.
SNR	MRIQC includes the the signal-to-nose ratio calculation proposed by Dietrich et al. [32], using the air background as noise reference. Additionally, for images that have undergone some noise reduction processing, or the more complex noise realizations of current parallel acquisitions, a simplified calculation using the within tissue variance is also provided.
QI2	The second quality index of [12] is a calculation of the goodness-of-fit of a χ2 distribution on the air mask, once the artifactual intensities detected for computing the QI1 index have been removed. The description of the QI1 is found below.
Measures based on information theory
EFC	The entropy-focus criterion [33] uses the Shannon entropy of voxel intensities as an indication of ghosting and blurring induced by head motion. Lower values are better.
FBER	The foreground-background energy ratio [14] is calculated as the mean energy of image values within the head relative the mean energy of image values in the air mask. Consequently, higher values are better.
Measures targeting specific artifacts
INU	MRIQC measures the location and spread of the bias field extracted estimated by the inu correction. The smaller spreads located around 1.0 are better.
QI1	The first quality index of [12] measures the amount of artifactual intensities in the air surrounding the head above the nasio-cerebellar axis. The smaller QI1, the better.
WM2MAX	The white-matter to maximum intensity ratio is the median intensity within the WM mask over the 95% percentile of the full intensity distribution, that captures the existence of long tails due to hyper-intensity of the carotid vessels and fat. Values should be around the interval [0.6, 0.8].
Other measures
FWHM	The full-width half-maximum [34] is an estimation of the blurriness of the image using AFNI’s 3dFWHMx. Smaller is better.
ICVs	Estimation of the icv of each tissue calculated on the FSL FAST’s segmentation. Normative values fall around 20%, 45% and 35% for cerebrospinal fluid (CSF), WM and GM, respectively.
rPVE	The residual partial volume effect feature is a tissue-wise sum of partial volumes that fall in the range [5%-95%] of the total volume of a pixel, computed on the partial volume maps generated by FSL FAST. Smaller residual partial volume effects (rPVEs) are better.
SSTATs	Several summary statistics (mean, standard deviation, percentiles 5% and 95%, and kurtosis) are computed within the following regions of interest: background, CSF, WM, and GM.
TPMs	Overlap of tissue probability maps estimated from the image and the corresponding maps from the ICBM nonlinear-asymmetric 2009c template [35].