Automatic image quality quantification and mapping with an edge-preserving mask-filtering algorithm

Area of Science:

• Medical Imaging
• Radiology
• Image Processing

Background:

• Digital radiology generates vast datasets requiring validated image quality.
• Diagnostic accuracy relies on consistent and reliable image quality assessment.
• Current methods may not be efficient for large-scale data analysis.

Purpose of the Study:

• To develop an automated method for quantifying image quality in digital radiology.
• To create a reliable and efficient tool for image quality assessment.
• To enable objective validation of diagnostic imaging operations.

Main Methods:

• A filtering algorithm with a moving square mask was employed to generate local intensity and noise maps.
• Image quality scores (IQs) were computed from filtered image data.
• The method was validated using technical and anthropomorphic phantoms with varied radiation dose, field of view (FOV), and image content, as well as a clinical CT brain image.

Main Results:

• Image quality scores (IQs) positively correlated with radiation dose (CTDIvol), increasing from 0.51 to 0.82 as dose rose from 9.2 to 74.3 mGy.
• High correlation (R²=0.99 and R²=0.98) was observed between IQs and pixel noise across varied conditions.
• Automatic tube current modulation improved image quality score consistency by approximately 60% compared to fixed tube current.

Conclusions:

• The developed image quality score (IQs) offers an efficient automated tool for image quality quantification.
• The method generates a 2D image quality map, facilitating detailed image analysis.
• This automated approach aids in validating diagnostic operations in digital radiology.