Potential utility of postmortem CT-based classification as a practical pre-autopsy assessment tool for hepatic steatosis

Abstract

Abstract This study aimed to establish postmortem computed tomography (PMCT)-specific Hounsfield unit (HU) criteria and developed an HU-based assessment tool for pre-autopsy triage of hepatic steatosis. Overall, 166 deceased individuals underwent whole-body PMCT followed by autopsy with liver histopathology. Hepatic attenuation was measured as a volumetric mean HU of the entire liver and as a four-point mean HU. No significant difference was noted between the volumetric and four-point mean HU ( p  = 0.183). However, the four-point mean HU differed significantly across histological steatosis grades ( p  < 0.001) and decreased with increasing severity, although the difference between mild and moderate steatosis was non-significant ( p  = 0.981). Receiver operating characteristic analysis indicated an optimal cutoff of 42.59 HU (sensitivity 60.7%, specificity 86.1%), from which three HU-based categories were defined (≥ 44.5 HU: normal; 42.0–44.4 HU: borderline; <42.0 HU: suspected steatosis). In the validation cohort ( n  = 104), the tool achieved an accuracy of 0.865, sensitivity of 0.756, specificity of 0.949, precision of 0.919, and F1-score of 0.829. This three-category PMCT HU-based classification system provides a practical tool for the triage of hepatic steatosis. The high specificity and precision support its use as a rule-in aid to prioritize suspected steatosis cases and improve autopsy workflow efficiency.