# A comparative evaluation of CT global noise calculation methods for clinical image quality assessment

## TL;DR

This study compares different methods for calculating global noise in CT scans to assess their impact on clinical image quality standards.

## Contribution

The study provides a comparative evaluation of global noise calculation methods in clinical CT protocols, highlighting variability and implications for standardization.

## Key findings

- Significant differences were found across global noise metrics in all CT protocols.
- Wisconsin_tissue_mean consistently produced the highest global noise values.
- Mode-based metrics showed higher agreement, suggesting dose dependency.

## Abstract

The recently introduced CMS quality measure for computed tomography (CT) requires compliance with two key metrics, radiation dose and image quality, across 18 exam categories. However, the method for calculating global noise (GN) remains undefined, with references to the “Duke method” and “Wisconsin method” as possible options. This lack of clarity raises concerns regarding standardization, compliance, and clinical relevance.

To compare GN calculation methods proposed in the Duke and Wisconsin papers and evaluate their variability across clinical CT protocols.

A retrospective analysis of 719 CT exams was performed across seven protocols, including five abdominal and two chest categories. One protocol (chest PE) included exams reconstructed with both smooth and sharp kernels. Five GN metrics were evaluated: Duke_tissue_mode, Wisconsin_tissue_mean, Wisconsin_tissue_mode, Wisconsin_air_mean, and Wisconsin_air_mode. Statistical differences were assessed using the Friedman test with pairwise Wilcoxon signed‐rank tests, and Pearson correlation matrices were used to evaluate agreement.

Significant differences (p < 0.05) were observed across GN metrics in all protocols, with Wisconsin_tissue_mean consistently producing the highest values. Correlation analysis showed strong agreement (r > 0.7) for renal stone, chest w/o, and abdomen/pelvis protocols, but weaker correlations for urogram, renal mass, and enterography. Mode‐based metrics showed higher agreement (r > 0.9), suggesting dose dependency. In the chest PE protocol, the smooth kernel yielded GN values well below the CMS threshold, while the sharp kernel exceeded the threshold in tissue‐based metrics.

Significant variability across GN metrics highlights the need for a standardized, clinically relevant method. Without clear definitions, the CMS measure's effectiveness in ensuring image quality and dose management may be compromised, an issue also raised by the AAPM‐commissioned panel.