MRQy

MRQy quantifies MRI image quality and detects site- and scanner-specific variations to enable correction of batch effects prior to computational model development.

Key Features:

• Generalized Application: Processes MR images from any body region to support diverse medical imaging datasets.
• Quality Measures Extraction: Extracts quality measures including noise ratios and variation metrics, and captures MR image metadata such as voxel resolution and image dimensions.
• Artifact Identification: Identifies site- or scanner-specific variations in image resolution or contrast and common imaging artifacts such as noise and inhomogeneity that require correction.
• Unsupervised Clustering for Site Identification: Employs unsupervised clustering to identify the origin of MRI datasets and facilitate detection of batch effects across sites.
• Outlier Detection: Detects outlier MRI datasets that require correction for common acquisition artifacts.

Scientific Applications:

• Brain MRI evaluation (TCIA): Applied to brain MRIs from TCIA across seven sites, revealing significant site-specific variations that were mitigated after processing.
• Rectal MRI evaluation (local): Applied to rectal MRIs from three local sites, identifying and reducing site-specific variations to support model generalizability.

Methodology:

Analyzes MR imaging datasets in .dcm, .nii, .nii.gz, and .mha formats using Python scripts (QC.py and QCF.py) to generate quality assessment tags and measurements saved in a .tsv file and .png thumbnails for each subject volume.