Hepatobiliary MRI: Signal intensity based assessment of liver function correlated to 13C-Methacetin breath test

Gadoxetic acid (Gd-EOB-DTPA) is a paramagnetic MRI contrast agent with raising popularity and has been used for evaluation of imaging-based liver function in recent years. In order to verify whether liver function as determined by real-time breath analysis using the intravenous administration of 13C-methacetin can be estimated quantitatively from Gd-EOB-DTPA-enhanced MRI using signal intensity (SI) values. 110 patients underwent Gd-EOB-DTPA-enhanced 3-T MRI and, for the evaluation of liver function, a 13C-methacetin breath test (13C-MBT). SI values from before (SIpre) and 20 min after (SIpost) contrast media injection were acquired by T1-weighted volume-interpolated breath-hold examination (VIBE) sequences with fat suppression. The relative enhancement (RE) between the plain and contrast-enhanced SI values was calculated and evaluated in a correlation analysis of 13C-MBT values to SIpost and RE to obtain a SI-based estimation of 13C-MBT values. The simple regression model showed a log-linear correlation of 13C-MBT values with SIpost and RE (p < 0.001). Stratified by 3 different categories of 13C-MBT readouts, there was a constant significant decrease in both SIpost (p ≤ 0.002) and RE (p ≤ 0.033) with increasing liver disease progression as assessed by the 13C-MBT. Liver function as determined using real-time 13C-methacetin breath analysis can be estimated quantitatively from Gd-EOB-DTPA-enhanced MRI using SI-based indices.

To monitor patients with hepatic dysfunction, various liver function tests are used routinely in clinical practice. These tests may evaluate the increasing severity of illness, differ between stages of disease and offer a prediction of therapy outcome [1][2][3] . The routine part of clinical investigations is based on analysis, where the levels of non-volatile compounds, such as proteins and ions, are measured and checked for abnormalities. As a complement to blood parameters, volatile compounds carry information concerning the biochemical status of the individual, which might be examined via breath tests [4][5][6][7][8][9][10] .
Different orally or intravenously administered 13 C-labeled substrates can reflect the function of specific hepatocyte compartments in real time as they are processed by liver function-dependent metabolic pathways. Therefore, cytosolic, mitochondrial and microsomal processes can be investigated non-invasively to obtain information about site-specific physiological and pathological metabolism 5 . A novel approach, which is already used in clinical routine, is the 13 C-methacetin breath test ( 13 C-MBT) established by Stockmann et al. 11 . The principle underlying this real-time test is the ability to metabolize 13 C-labeled methacetin by the hepatocyte endoplasmic reticulum-located cytochrome P 450 1A2 (CYP1A2) into paracetamol and 13 C-labeled formaldehyde, which will be eliminated as 13 CO 2 5,[11][12][13] . After intravenous (i.v.) injection, 13 CO 2 will be exhaled, and a 13 CO 2 : 12 CO 2 ratio can be determined by a suitable device for breath analysis. Based on the values obtained by the 13 C-MBT, real-time information of patients' liver function can be directly obtained.
The hepatocyte-specific magnetic resonance imaging (MRI) contrast agent gadoxetic acid (Gd-EOB-DTPA) has been established and used for diagnostic purposes to detect and characterize focal liver lesions. Several studies have reported the use of Gd-EOB-DTPA-based MRI to evaluate liver function, usually expressed via the SI in T1-weighted MRI scans compared with 13 C-MBT readout. In a simple linear regression model, RE values were strong linear predictors of the logarithmic values of uncategorized 13 C-MBT readout values (r = 0.665, p < 0.001), while the SI values obtained from the HBP showed less predictive power (r = 0.554, p < 0.001). In contrast, SI values obtained without contrast enhancement showed no significant correlation (p = 0.120) ( Table 2). Scatterplots of SI pre , SI post and RE values plotted against the logarithmic values of 13 C-MBT are shown in Fig. 2. SI-based results compared with 13 C-MBT readout categories. Patients with normal liver function (Category 1) had a mean 13 C-MBT readout of 391.21 ± 53.28 µg/kg/h, whereas patients with intermediate liver function (category 2) had a mean 13 C-MBT readout value of 217.78 ± 49.39 µg/kg/h, and patients with severely impaired liver function had a mean 13 C-MBT readout value of 93.03 ± 33.59 µg/kg/h (Table 1). All pairwise comparisons between Categories 1, 2 and 3 showed significant differences (p ≤ 0.05) in the evaluated SI-based indices, except for SI pre values between Categories 1 and 3 and Categories 2 and 3 (Fig. 3A). The SI values obtained from the HBP (SI post ; p < 0.005; Fig. 3B) and the RE values (p < 0.05; Fig. 3C) differed significantly among the three 13 C-MBT based categories.

Discussion
One of the most common types of liver cancer is HCC, with a global cancer mortality rate of 9.1% 14 . As HCC is mostly based on preexisting liver cirrhosis, the early detection and assessment of liver cirrhosis is of high clinical relevance. Patients benefit from rapid and accurate examinations of the liver condition. Thus, the 13 C-MBT has become more common, as it can provide liver failure predictions, liver transplant control, and liver disease severity estimations 13,[15][16][17] . The test is based on the analysis of volatile components exhaled during the 13 C-methacetin metabolism by CYP1A2 in the endoplasmic reticulum 12 . Different studies have demonstrated the predictive power of enzymatic hepatocyte functionality for liver resection 13,15,[17][18][19][20][21] . Even in cases of non-cirrhotic, early-stage and primary biliary cirrhosis, the 13 C-MBT can reliably indicate decreased liver function 10 . Nevertheless, this tool has some diagnostic restrictions, as it is unable to determine and distinguish between areas of reduced hepatocyte function and healthy areas. Therefore, 13 C-MBT values describe the function of the whole liver. However, diagnostic imaging techniques might provide these details, which are crucial for both liver resection and liver transplantation. To reduce the number of different examinations, researchers aim to establish liver tests capable of not only reliably reflecting liver function but also revealing hepatic lesions in a single examination.
To this end, Gd-EOB-DTPA-enhanced MRI has been established as a promising approach for assessing liver function in the past few years. Gd-EOB-DTPA consists of a gadolinium ion covalently bound to a lipophilic ethoxybenzyl group and is absorbed by intact hepatocytes via triggered ATP-dependent organic anion-transporting polypeptide (OATP1 B1/B3) channels. Subsequently, Gd-EOB-DTPA will be excreted by multidrug-resistance protein 2 (MRP2) into the biliary system [22][23][24] . The strength of Gd-EOB-DTPA is its hepatocyte-specific character, which improves the contrast-to-noise ratio, as it will not be taken up by cells other than hepatocytes (e.g., cells in  The lesion observed in the liver with impaired function (D,E) was caused by former radiofrequency ablations treatments. The delta-over-baseline (DOB) was assessed inline automatically and describes the increase in the R PDB -corrected 13 CO 2 : 12 CO 2 ratio to the basal value (blue line). The evaluated 13 C-MBT value was calculated as the product of the DOB max , R PDB , CO 2 production and molar mass of 13 C-methacetin per body weight 13 . The DOB max (green line) was defined after an increase in DOB was no longer observable. At the time point 0, the 13 C-methacetin was applied via bolus injection. metastatic lesions) [25][26][27] . At 20 min after contrast agent application, a significant increase in the SI of the liver parenchyma is recorded, and the extent of this phenomenon is dependent on the integrity of the liver parenchyma. The highest SI is exhibited by healthy liver tissue, while cirrhotic liver parenchyma shows only a slight increase in SI 28,29 .
A loss of functioning hepatocytes, bridging of portal spaces or nodular regeneration of the liver parenchyma is associated with liver cirrhosis and hindered hepatocyte contrast agent uptake, causing decreased SI-based values 30 .
The estimation of SI-based indices provides insights into regional hepatocyte-specific function, efficiency, functionality and condition 31 . Nevertheless, the analysis is restricted by the relative character of the obtained SI values 32 . However, by the mathematical computation of SI post with SI pre , a liver function index with increased reliability can be achieved. By calculating the RE, we correct the enhanced SI values to gain a SI ratio independent from artificial signal enhancement with increased explanatory power 29,[33][34][35][36][37][38] .
To the best of our knowledge, this is the first study comparing SI-based MRI values reflecting hepatocyte OATP1 B1/B3 and MRP2 pathway activity with the 13 C-MBT.
We could demonstrate that the SI-based MRI values assessed in the HBP reflect liver function in a suitable manner, as estimated by 13 C-MBT readout values. Similar to a previous study 39  . In a correlation analysis of SI-based indices to 13 C-MBT values, we were able to show that SI-based RE values significantly support the 13 C-MBT findings (RE, r = 0.665, p < 0.001) and therefore reflect liver functionality. Similar to Utsunomiya et al. 31 , in this study, we tested the correlation of the mean SI obtained after contrast agent application (SI post ) to the results of a liver function test. In their case, the tested SI values showed a higher prediction of indocyanine green dye (ICG) retention at 15 min (r = −0.67, p < 0.01) than we obtained when testing against 13 C-MBT values (r = 0.554, p < 0.001). However, this difference seems plausible because the ICG clearance test is, similar to the GD-EOB-DTPA pathway, dependent from OATP transporter activity. Therefore, we expected a slightly less pronounced correlation between the 13 C-MBT and SI-based values, as the 13 C-MBT relies on an enzymatic metabolism, whereas contrast-enhanced MRI relies on OATP channel-triggered contrast agent uptake. In general, the ICG clearance test has certain limitations, as constant hemodynamic conditions (stable liver perfusion rate and hepatic blood flow) are required for liver function analysis [40][41][42] and, in cases of cholestasis and hyperbilirubinemia, carrier competition of bilirubin and ICG at the OATP1 transporter might occur 43,44 . It is also known that various drugs (e.g., rifampicin) exert inhibitory effects on the OATP pathway and might influence hepatocyte ICG uptake 45,46 . Similar to the findings of Tamada et al. 36 , we could show that the RE of the liver parenchyma serves as a reliable tool for liver function classification, as the RE significantly differs among different stages of liver function (p ≤ 0.02). Additionally, it has been shown that the hepatic enhancement during Gd-EOB-DTPA-enhanced MRI is strongly affected by the degree of liver cirrhosis, as expressed by the ICG test, the Child-Pugh score or the MELD score 34,47 . These studies have shown that Gd-EOB-DTPA-enhanced MRI has potential as a reliable tool for liver function estimation in addition to its already established implementation for hepatic lesion detection.
Our study has several limitations. First, ROI placement may cause some variations due to the possible nonhomogeneous distribution of parenchymal changes. However, using the average of six repeated ROI measurements across an area of the liver parenchyma should provide reliable values. Second, this study was retrospective in nature, with only a limited patient population. Third, the lack of histopathology is another potential limitation.
In conclusion, SI-based indices, such as the RE and contrast-enhanced SI values, can be used to determine liver function as assessed by 13 C-MBT.

Materials and Methods
Patients. Local institutional review board approval of the University Hospital Regensburg was obtained for 13 C-MBT and Gd-EOB-DTPA-enhanced MRI at 3 T. Only data from written informed consent patients were included for this analysis, also the study was performed in accordance with the relevant guidelines and regulations. The retrospective analysis includes 110 patients (83 men and 27 women; median age, 61 years) who underwent both a 13 C-MBT and Gd-EOB-DTPA-enhanced MRI at 3 T. The patients underwent Gd-EOB-DTPA-enhanced T1-weighted volume-interpolated breath-hold examination (VIBE) MRI sequences with fat suppression. The included patients did not have known reactivity to liver-specific MRI contrast media, 13 C-methacetin intolerance or renal-specific contraindications to either MRI and Gd-EOB-DTPA administration.
For the statistical analysis, the patients were grouped according to their 13 C-MBT readout into 3 categories: patients with normal liver function (Category 1): 13   The patients received a Gd-EOB-DTPA dose (0.025 mmol/kg body weight) adapted to their body weight administered via bolus injection at a flow rate of 1 mL/s, followed by 20 mL of 0.9% sodium chloride.

Image analysis.
Operator-defined region-of-interest (ROI) measurements were used to obtain the mean SI values from the T1-weighted VIBE images (before and after Gd-EOB-DTPA injection). ROIs were manually placed at identical locations in every sequence, avoiding liver lesions, major branches of the portal and hepatic veins, and imaging artifacts. In total, 6 ROIs (3 each in the right and left lobes) were defined in the VIBE images (Fig. 4). Each ROI was a circle that was made as large as possible (liver parenchyma: 1.1 cm 2 -4.6 cm 2 ) and manually adjusted between sequences if necessary.
The relative enhancement (RE) of the liver was calculated according to following formula: SI post SI pre SI pre (1) Statistical analysis. The different 13 C-MBT readout categories were compared as non-parametric independent samples by the Mann-Whitney-U test. The predictive power of SI-based indices was determined by simple linear regression models, and the optimal curve fit was assessed visually. In all tests, the statistical significance level was set to 0.05 (two-sided). All analyses were performed using SPSS software (version 24; IBM, Chicago, IL, USA).

Data availability.
All data that support the findings of this study are provided in the manuscript. Raw data used in this work are available on reasonable request.