Entropy-Based CT Radiomics as an Imaging Marker of Hepatic Injury in COVID-19.

Background

Hepatic involvement in COVID-19 is frequently observed, yet conventional CT imaging may fail to detect subtle parenchymal alterations. This study aimed to evaluate whether CT-based radiomic texture analysis can identify liver injury associated with SARS-CoV-2 infection.

Methods

We retrospectively analyzed 41 patients with RT-PCR-confirmed moderate or severe COVID-19 pneumonia who underwent non-contrast thoracoabdominal CT during the acute phase and at follow-up. Liver volume, mean hepatic attenuation, liver-to-spleen attenuation ratio, and radiomic features including first-order and GLCM entropy were extracted using 3D Slicer version 5.6.2 and SlicerRadiomic Revision: 8426cdf. Hepatic injury was defined by elevated serum transaminases. Three additional patients with available liver histopathology were included for correlation with imaging findings.

Results

Patients with biochemical liver injury demonstrated significantly higher hepatic entropy values in the acute phase compared to those without injury (first-order entropy: 1.63 vs. 1.48, p = 0.019; GLCM entropy: 3.12 vs. 2.83, p = 0.013). Entropy metrics were inversely correlated with hepatic attenuation at follow-up (GLCM r = -0.385, p = 0.013; first-order r = -0.346, p = 0.027), indicating possible progression to lower-density states. Ferritin showed a moderate positive correlation with entropy (r = 0.47, p = 0.0017). Histopathological examination revealed steatosis, hepatocellular injury, inflammatory infiltration, and vascular congestion, aligning with radiomic abnormalities.

Conclusions

Entropy-based CT radiomics reflect microstructural liver alterations in COVID-19, supported by both biochemical and histopathological data. This approach may enhance the detection of hepatic injury beyond conventional imaging and could be explored in systemic infections.