TL;DR
This study investigates the reproducibility of a multi-material decomposition (MMD) algorithm for measuring liver fat volume fraction (FVF) in patients with nonalcoholic fatty liver disease (NAFLD). The research highlights that while MMD is generally reproducible across different liver lobes, its accuracy is significantly influenced by the presence of blood vessels and the use of iodine contrast agents during CT scans.
What Problem Does This Paper Address?
Nonalcoholic fatty liver disease (NAFLD) is a growing global health concern, and accurate, non-invasive methods for quantifying liver fat are crucial for diagnosis and monitoring. While spectral CT-derived multi-material decomposition (MMD) shows promise as an imaging biomarker for hepatic steatosis, there's a lack of standardization in its measurement. Different studies report varying results, and the impact of key factors like blood vessel interference, anatomical location within the liver, and the presence of iodine contrast agents on FVF measurements remains unclear. This paper aims to address this gap by systematically evaluating these influences on MMD-based FVF measurements in NAFLD patients.
Methods
The researchers conducted a retrospective study on 101 NAFLD patients who underwent spectral CT scanning. This included a non-contrast phase (NCP) and two contrast-enhanced phases: late arterial phase (LAP) and portal vein phase (PVP). The core of their methodology involved applying the MMD algorithm to generate liver fat images. Regions of interest (ROIs) were strategically placed in three different liver lobes: the right posterior lobe (RPL), right anterior lobe (RAL), and left lateral lobe (LLL).
Crucially, they compared FVF values obtained under two main conditions:
- Without the reference of enhanced CT images: ROIs were placed on non-contrast images, and FVF was measured.
- With the reference of enhanced CT images: ROIs were placed on contrast-enhanced (specifically, PVP) images, and then the corresponding FVF was measured on non-contrast (NCP) images, with adjustments made to avoid visible blood vessels.
Statistical analyses, including the Friedman test and Bland-Altman analysis, were used to compare FVF values across different lobes and scanning phases, and to assess consistency.
Datasets
The study utilized CT scan data from 101 patients diagnosed with NAFLD. These patients underwent both non-contrast and contrast-enhanced spectral CT scans. The paper does not introduce a new benchmark dataset but rather uses existing patient data for analysis.
Key Insights and Ideas
- Vessel Interference is Significant: The presence of blood vessels in the liver significantly impacts FVF measurements. When ROIs are placed without avoiding vessels, the measured FVF is lower, likely because blood (and its contrast agent) has a different composition than liver tissue and fat, leading to an underestimation of fat content in those voxels.
- Iodine Contrast Alters FVF: The study found that FVF values obtained after contrast injection were consistently higher than those from non-contrast scans. This is attributed to the MMD algorithm's process of generating virtual un-enhancement (VUE) images, which, while aiming to remove contrast, may not perfectly replicate true non-contrast conditions, leading to differences in FVF.
- Location Matters Less Than Expected: Within the context of this study (which focused on patients with relatively homogeneous fatty liver), the anatomical location of the ROI within the three chosen lobes (RPL, RAL, LLL) did not significantly affect FVF measurements when using enhanced CT images as a reference. This suggests that for homogeneous NAFLD, lobe selection might be less critical than other factors.
- Reproducibility is Achievable with Care: The MMD algorithm demonstrates reproducibility across different liver lobes, but achieving accurate and consistent results hinges on careful methodology, particularly in avoiding vascular structures and accounting for contrast enhancement.
Key Results
- FVF values measured with the reference of enhanced CT images (and thus avoiding vessels) were significantly higher than those measured without this reference (P < 0.05 for all lobes).
- There was no significant difference in FVF values obtained from NCP images with the reference of enhanced CT images between any of the three lobes (RPL, RAL, LLL) (P > 0.05).
- FVF values showed a significant increase across the scanning phases: PVP > LAP > NCP (all P < 0.001 for each lobe).
- Bland-Altman analysis revealed poor consistency of FVF values between any two scanning phases (all P < 0.05), indicating that the phase of scanning significantly influences the measurement.
Limitations and Future Work
The authors acknowledge several limitations:
- Contrast Administration Rate: The rate of contrast injection might not have been sufficient to eliminate all enhancement differences caused by varying circulation times among patients.
- ROI Size: The influence of ROI size on the results was not specifically evaluated.
- Lack of Gold Standard: The study did not have access to histopathology (gold standard) for direct comparison.
- Patient Selection: Patients with significantly decreased liver CT density (indicating mild NAFLD) might have been excluded, potentially limiting generalizability.
- Comparison with MRI/US: The study did not compare its MMD CT findings with MRI or ultrasound, which are other common methods for assessing liver fat.
Future work could involve optimizing contrast administration protocols, investigating the impact of ROI size, and conducting comparative studies with MRI and ultrasound to further validate the MMD approach. Exploring methods to better account for iodine contrast in MMD algorithms is also a promising avenue.
Summary updated