Development of a Global Noise Index for the Assessment of Image Quality In Clinical Radiography
Abstract
Purpose
General radiography lacks a clinically relevant, vendor-agnostic metric to indicate for-presentation image quality across anatomical views. While exposure index is commonly used to indicate exposure appropriateness, it is biased by vendor-specific region-of-interest (ROI) selection and does not consider the influence of post-processing. As a first step towards automated assessment of radiographic quality, we have developed a Global Noise Index (GNI) for general radiography to quantify for-presentation image noise in clinical radiography images.
Methods
Anthropomorphic hand and elbow phantom radiographs were acquired across a range of techniques (50–60 kVp, 0.4–5 mAs) and post-processing parameters. Images were segmented into soft tissue and bone masks. Laplacian decomposition was applied to generate five frequency sub-bands per image. Relative noise in each tissue mask was calculated as the standard deviation of the high frequency sub-band divided by the mean of the low frequency sub-band. Soft tissue and bone relative noise were combined using a weighted sum to produce the GNI. To establish a ground-truth noise reference, five acquisitions were repeated at the highest mAs setting for the given kVp and post-processing setting, summed to generate a low-noise reference image, and scaled to match the intensity profile of the image being analyzed. Ground truth noise was estimated by subtracting the corresponding low-noise image from each acquired image.
Results
GNI demonstrated strong correlation (ρ = 0.894) with independently measured true noise across all evaluated conditions. GNI values decreased concordant with increasing mAs and kVp. GNI across all variations in post-processing parameters with correlation factors of ρ = 0.945-0.981.
Conclusion
This study demonstrates the feasibility of a vendor-agnostic GNI for routine assessment of noise in for-presentation radiographic images. Pending clinical validation, radiographic GNI provides a potentially reliable metric for image quality evaluation, optimization, and assurance beyond conventional metrics.