Detection of fatty liver using virtual non-contrast dual-energy CT

Abstract

Purpose Determine whether liver attenuation measured on dual-energy CT (DECT) virtual non-contrast examinations predicts the presence of fatty liver. Methods Single-institution retrospective review from 2016 to 2020 found patients with DECT and proton density fat fraction MRI (MRI PDFF) within 30 days. MRI PDFF was the reference standard for determining hepatic steatosis. Attenuation measurements from VNC and mixed 120 kVp-like images were compared to MRI PDFF in the right and left lobes. Performance of VNC was compared to measurement of the liver-spleen attenuation difference (LSAD). Results 128 patients were included (69 men, 59 women) with mean age 51.6 years (range 14–98 years). > 90% of patients received CT and MRI in the emergency department or as inpatients. Median interval between DECT and MRI PDFF was 2 days (range 0–28 days). Prevalence of fatty liver using the reference standard (MRI PDFF > 6%) was 24%. Pearson correlation coefficient between VNC and MRI- DFF was -0.64 (right) and -0.68 (left, both p < 0.0001). For LSAD, correlation was − 0.43 in both lobes (p < 0.0001). Considering MRI PDFF > 6% as diagnostic of steatosis, area under the receiver operator characteristic curve (AUC) was 0.834 and 0.872 in the right and left hepatic lobes, with an optimal threshold of 54.8 HU (right) and 52.5 HU (left), yielding sensitivity/specificity of 57%/93.9% (right) and 67.9%/90% (left). For LSAD, AUC was 0.808 (right) and 0.767 (left) with optimal sensitivity/specificity of 93.3%/57.1% (right) and 78.6%/68% (left). Conclusion Attenuation measured at VNC CT was moderately correlated with liver fat content and had > 90% specificity for diagnosis of fatty liver. Graphical abstract

Spectral CT: Current Liver Applications

“…For VNC images, a liver attenuation of <40 Hounsflied units (HU) is highly specific and positively predictive for moderate to severe steatosis using unenhanced SECT criteria, supporting their reliability despite the mild overestimation of liver attenuation compared to true unenhanced images (by 5.4 UH) [72]. A recent study performed using proton density fat fraction MRI as the reference standard found only a moderate correlation between VNC attenuation values and liver fat content with 57-68% sensitivity, but discovered high (≥90%) specificity for diagnosis of steatosis [73]. Others found comparable diagnostic performance of CT parameters measured in VNC and true unenhanced images for diagnosing liver steatosis [74][75][76]; however, further studies are needed if we are to be able rely on VNC values for the diagnosis of liver steatosis.…”

“…Workflow limitations related to the increased reconstruction time have been improved with multiple strategies, including the use of lighter workstations or integration of postprocessing software in remote workstations [6]. However, large amounts of data require capacious storage systems and demand more interpretation time [73].…”

Non-invasive imaging biomarkers for liver steatosis in non-alcoholic fatty liver disease: present and future

“…Optimal thresholds were 54.8 hounsfield unit (HU) (right) and 52.5 HU (left), with sensitivities/specificities of 57%/93.9% (right) and 67.9%/90% (left). For the hepatosplenic weight loss difference, the AUROCs were 0.808 (right) and 0.767 (left), with optimal sensitivities/specificities of 93.3%/57.1% (right) and 78.6%/68% (left) [ 72 ].…”

Dual-Energy CT in Oncologic Imaging

“…Notably, recent advancements have highlighted its significant advantages in the evaluation of oncology patients [ 20 , 21 , 22 ]. Key among these applications are virtual non-contrast (VNC) and iodine maps, which effectively highlight tissue vascularity [ 23 , 24 , 25 , 26 , 27 , 28 , 29 , 30 , 31 , 32 , 33 , 34 ] in both primary tumors and metastases [ 35 , 36 , 37 , 38 , 39 , 40 , 41 , 42 , 43 ]. Furthermore, enhanced visualization of vascular structures enables the optimized assessment of vascular involvement [ 44 ].…”

“…Among these, in addition to primary hepatocellular liver tumors, there are most hypovascularized liver metastases, which are among the most frequent in daily practice, affecting patients with colon, breast, and lung cancer. Non-contrast dual-energy acquisition could also be beneficial in improving the performance of applications focused on studying bone, liver, and calcifications, which may be associated with pancreatic or obstructive diseases [ 29 , 41 ]. These acquisitions are also the most commonly used and likely more reliable for opportunistic assessment of vertebral density, which can be utilized as an alternative to conventional bone densitometry for identifying osteoporosis.…”