Image Quality Metrics (IQMs)¶
Some no-reference IQMs are extracted in the
final stage of all processing workflows run by MRIQC.
A no-reference IQM is a measurement of some aspect
of the actual image which cannot be compared to a reference value for the metric
since there is no ground-truth about what this number should be.
All the computed IQMs corresponding to
an image are saved in a JSON file under the
The IQMs can be grouped in four broad categories, providing a vector of 56 features per anatomical image. Some measures characterize the impact of noise and/or evaluate the fitness of a noise model. A second family of measures use information theory and prescribed masks to evaluate the spatial distribution of information. A third family of measures look for the presence and impact of particular artifacts. Specifically, the INU artifact, and the signal leakage due to rapid motion (e.g. eyes motion or blood vessel pulsation) are identified. Finally, some measures that do not fit within the previous categories characterize the statistical properties of tissue distributions, volume overlap of tissues with respect to the volumes projected from MNI space, the sharpness/blurriness of the images, etc.
Most of the IQMs in this module are adapted, derived or reproduced from the QAP project [QAP]. We particularly thank Steve Giavasis (@sgiavasis) and Krishna Somandepali for their original implementations of the code in this module that we took from the [QAP]. The [QAP] has a very good description of the IQMs in [QAP-measures].
- IQMs for structural images
- IQMs for functional images