Spatial non-uniformity of liver virtual non-contrast attenuation on dual-layer spectral CT: Impact of body habitus and peripheral right-lobe error.
In larger patients, central liver measurements may be preferred over peripheral measurements when evaluating diffuse liver attenuation using dual-layer CT.
Where it sits
this study against the rest of the survodutide (bi 456906) corpusSummary and findings
This study evaluated the accuracy of virtual non-contrast liver attenuation from dual-layer spectral CT in 99 patients. The median absolute differences between true non-contrast and virtual non-contrast liver attenuation were reported across different phases. The study found that peripheral virtual non-contrast error exceeded central error in all phases, particularly in the portal venous phase.
Abstract
<h4>Purpose</h4>To evaluate the global and spatial accuracy of virtual non-contrast (VNC) liver attenuation derived from dual-layer spectral CT and its association with patient body habitus.<h4>Methods</h4>This retrospective single-center study included 99 patients examined with true non-contrast (TNC) and three post-contrast phases on dual-layer CT. Global TNC-VNC agreement was assessed in liver segments S2, S7, and S8. Detailed center-to-periphery analysis in segments S7 and S8 was performed in 87 patients. Spatial non-uniformity was quantified using the spatial error gradient (SEG), defined as the difference between peripheral and central absolute VNC error.<h4>Results</h4>Median absolute whole-liver TNC-VNC differences were 3.0 HU in the arterial phase, 4.4 HU in the portal venous phase, and 3.1 HU in the delayed phase. Peripheral VNC error exceeded central error in all phases; the largest difference was observed in the portal venous phase (4.4 vs 2.3 HU), with a corresponding SEG of 2.4 HU (p < 0.001). Abdominal effective diameter showed a weak but statistically significant correlation with SEG in the portal venous and delayed phases (both p = 0.004). In an exploratory ROC analysis, effective diameter predicted elevated peripheral error with an AUC of 0.746.<h4>Conclusion</h4>Dual-layer CT-derived VNC images showed good overall agreement with TNC, but exhibited a population-level tendency toward spatial non-uniformity, with peripheral error exceeding central error. Based on these exploratory data, in evaluation of diffuse liver attenuation, central liver measurements may be preferred over peripheral measurements in larger patients.
Background
This paper addresses the accuracy of virtual non-contrast (VNC) liver attenuation derived from dual-layer spectral CT, particularly in relation to patient body habitus. Prior studies have suggested variability in imaging accuracy based on body composition, but specific impacts on liver attenuation measurements had not been extensively characterized. Understanding these relationships is crucial for improving diagnostic accuracy in imaging.
Methods
The study employed a retrospective design involving 99 patients who underwent true non-contrast (TNC) and three post-contrast phases on dual-layer CT. Global TNC-VNC agreement was assessed in liver segments S2, S7, and S8, with a detailed center-to-periphery analysis performed in 87 patients. The primary outcome measure was the median absolute differences in liver attenuation across the various phases.
Results
The primary endpoint revealed median absolute whole-liver TNC-VNC differences of 4.4 HU in the portal venous phase (p < 0.001). Peripheral VNC error exceeded central error in all phases, with the largest discrepancy noted in the portal venous phase (4.4 vs 2.3 HU). The spatial error gradient (SEG) was quantified at 2.4 HU, indicating significant spatial non-uniformity.
Interpretation
The findings indicate that while dual-layer CT-derived VNC images generally agree with TNC measurements, there is a notable tendency for spatial non-uniformity, especially in larger patients. The effect size of the differences, while statistically significant, may not be clinically meaningful without further validation. Limitations such as the small sample size and single-center design may confound the results, suggesting caution in applying these findings broadly.
Key findings
- Median absolute whole-liver TNC-VNC differences were 3.0 HU in the arterial phase, 4.4 HU in the portal venous phase, and 3.1 HU in the delayed phase.
- The largest difference in peripheral vs central error was observed in the portal venous phase (4.4 vs 2.3 HU), with a corresponding SEG of 2.4 HU (p < 0.001).
- Abdominal effective diameter showed a weak but statistically significant correlation with SEG in the portal venous and delayed phases (both p = 0.004).
- In an exploratory ROC analysis, effective diameter predicted elevated peripheral error with an AUC of 0.746.
Limitations
- Retrospective single-center study.
- Small sample size (n=99).
- Weak correlation between effective diameter and SEG.
- Exploratory ROC analysis may not be definitive.