Liver density in routine and low-dose computed tomography: the effect of image noise on measurement accuracy

Abstract

The introduction of LDCT-based screening programs into clinical practice allows an additional assessment of the liver. It is known that medium-to-severe steatosis can be detected using LDCT. However, taking into account the increased image noise level and the fact that the liver is always only partly in the scan area, the question arises as to how accurately the liver density can be determined in LDCT relative to routine CT. Purpose. Thus, the following objectives of this study have been established: • To identify the differences between the mean density of the liver as measured by a CT and an LNDT. • To compare the mean density between CT and LDCT in different patient subgroups depending on the mean liver density (<40 HU, 40–50 HU, 50–60 HU, and >60 HU). • To determine the effect of image noise level on the mean liver density values on LDCT compared to CT. Materials and methods. We analyzed 30000 patient records from 2017 to 2019. Inclusion criteria: We included patients with both thoracic non-contrast CT and an LDCT and the time interval between the studies of less than 27 days. Exclusion criteria: The absence of CT-LDCT pair, presence of focal liver lesions, known liver diseases, operated liver, anemia with blood density decrease below 40 HU, hands along the body instead of overhead. Study protocol: LDCT was performed at 135 kV. The routine CT was at 120 kV. Scan length: from lung apex to pleural sinuses. The average radiation dose with LDCT was 0.6–0.8 mSv, and 2.8–4.6 mSv with routine CT. All scans were performed on two 64-detector units from the same manufacturer. We measured liver density with the CTLiverExam software algorithm for automatic liver densitometry. The statistical processing was done using the Stata14 program. Results. We used data from 61 patients for statistical analysis. The average age was 53 years. The ratio of men to women was 23:38. We did not observe statistically significant differences between CT and LDCT. With a breakdown by the initial liver density level, it turned out that in the subgroup 40-50 HU, the differences were statistically significant. No correlation between liver density and standard deviation for CT was revealed (p = 0.338). There was a mild anticorrelation for LDCT with a coefficient of -0.686 (p < 0.0001). Conclusion. Our study shows that liver density measurement in thoracic LDCT is valid. In the context of lung cancer screening programs. An analysis of the image noise/liver density ratio on the LDCT shows an inversely proportional relationship: the higher the noise level, the more significant a “decrease” in liver density. This factor must be taken into account when interpreting the results CT and LDCT.