- Original research
- Open Access
A new compartmental method for the analysis of liver FDG kinetics in small animal models
- Sara Garbarino^{1, 2},
- Valentina Vivaldi^{1, 2},
- Fabrice Delbary^{1},
- Giacomo Caviglia^{1},
- Michele Piana^{1, 2}Email author,
- Cecilia Marini^{3},
- Selene Capitanio^{4},
- Iolanda Calamia^{4},
- Ambra Buschiazzo^{4} and
- Gianmario Sambuceti^{4, 5}
- Received: 10 February 2015
- Accepted: 15 April 2015
- Published: 11 June 2015
Abstract
Background
Compartmental analysis is a standard method to quantify metabolic processes using fluorodeoxyglucose-positron emission tomography (FDG-PET). For liver studies, this analysis is complex due to the hepatocyte capability to dephosphorylate and release glucose and FDG into the blood. Moreover, a tracer is supplied to the liver by both the hepatic artery and the portal vein, which is not visible in PET images. This study developed an innovative computational approach accounting for the reversible nature of FDG in the liver and directly computing the portal vein tracer concentration by means of gut radioactivity measurements.
Methods
Twenty-one mice were subdivided into three groups: the control group ‘CTR’ (n = 7) received no treatment, the short-term starvation group ‘STS’ (n = 7) was submitted to food deprivation with free access to water within 48 h before imaging, and the metformin group ‘MTF’ (n = 7) was treated with metformin (750 mg/Kg per day) for 1 month. All mice underwent a dynamic micro-PET study for 50 min after an ^{18}F-FDG injection. The compartmental analysis considered two FDG pools (phosphorylated and free) in both the gut and liver. A tracer was carried into the liver by the hepatic artery and the portal vein, and tracer delivery from the gut was considered as the sole input for portal vein tracer concentration. Accordingly, both the liver and gut were characterized by two compartments and two exchange coefficients. Each one of the two two-compartment models was mathematically described by a system of differential equations, and data optimization was performed by applying a Newton algorithm to the inverse problems associated to these differential systems.
Results
All rate constants were stable in each group. The tracer coefficient from the free to the metabolized compartment in the liver was increased by STS, while it was unaltered by MTF. By contrast, the tracer coefficient from the metabolized to the free compartment was reduced by MTF and increased by STS.
Conclusions
Data demonstrated that our method was able to analyze FDG kinetics under pharmacological or pathophysiological stimulation, quantifying the fraction of the tracer trapped in the liver or dephosphorylated and released into the bloodstream.
Keywords
- Compartmental analysis
- Liver physiology
- Dual input
- FDG-PET
- Metformin
Background
Hepatic glucose metabolism is of relevance in a number of clinical syndromes and, in particular, in diabetes [1-5]. Several authors previously evaluated the potential of positron emission tomography (PET) with ^{18}F-fluorodeoxyglucose (FDG) in this setting, both in humans and animals [2,3]. However, a standard method has not been developed so far because of the peculiarities of liver function and anatomy. From the physiological viewpoint, hepatocytes can dephosphorylate glucose-6-phosphate (G6P) to deliver glucose back into the bloodstream. This feature also applies to FDG-6-phosphate (FDG-6P) preventing the accumulation kinetics typical of most tissues and nicely suited to compartmental analysis. From the anatomical viewpoint, the dual blood supply from the hepatic artery (HA) and portal vein (PV) further hampers the definition of tracer delivery to hepatic tissue [6]. Actually, limiting the analysis to the arterial input function (IF) inevitably causes systematic underestimation of tracer distribution in the liver [7-11]. Accordingly, several attempts have been made to estimate the dual-input IFs from dynamic PET data [8,12,13]. However, no method successfully solved this problem so far, since the PV is hard to recognize in PET images and suffers from a relevant partial-volume effect related to vessel size and respiratory movements.
According to these limitations, several authors tried to indirectly predict the PV tracer concentration by the arterial IF. More precisely, one approach applied ad hoc algorithms to compute the PV IF as a heuristic modification of the arterial one [5,9]. A second approach directly and invasively sampled the PV IF [3-5] and then determined the dual IF as a weighed superposition of the two activities [10,14-18]. However, both methods are model-dependent and difficult to apply in a noninvasive setting, while their potential under non-physiological conditions has been scarcely explored.
- 1.
The time-concentration curve in the arterial system.
- 2.
The time-concentration curve in the liver.
- 3.
The time-concentration curve in the whole gut as the source of the PV blood.
We used this information as input data of an iterative algorithm [19] that reduces a compartmental model in which gut tracer concentration is used to compute the PV IF, while liver activity is used to describe tracer kinetics in the hepatic system. In the present study, we aimed at validating this algorithm by testing its performance in mice treated with interventions able to blunt or enhance liver G6P dephosphorylation and glucose delivery.
Methods
Reagents
Metformin was provided by Sigma-Aldrich (St. Louis, MO, USA). FDG was produced in-house, according to standard methodology. Daily quality controls always documented adequate standards and, in particular, a radiochemical purity ≥98%.
Animal models
All animal experiments were reviewed and approved by the Licensing and Ethical Committee of the IRCCS San Martino IST, Genova, Italy, and by the Italian Ministero della Salute. A total of 21 six-week-old BALB/c female mice (Charles River Laboratories, Italy) were housed under specific pathogen-free conditions and subdivided into three different groups according to the treatment preceding the imaging study. The first group of controls (CTR) included seven untreated animals that did not receive any treatment and were kept under standard conditions for the whole study duration. The short-term starvation group (STS) included seven mice that were submitted to food deprivation with free access to water within 48 h before imaging. Finally, the last group (MTF) included seven mice treated with metformin for the month preceding the imaging test. In these animals, the drug was administered diluted in autoclaved drinking water at a concentration of 3 mg/mL as to account for a dose of 750 mg/Kg per day [11].
Animal preparation and image processing
Compartmental modeling of FDG kinetics
Our approach for the definition of the dual IF was based on processing a complete set of ROIs drawn on the aortic arc, gut, and liver. In particular, we aimed to describe the gut by means of a compartmental model with one arterial IF and one output directly delivering the tracer in the PV blood. This latter concentration was then used as the venous IF in a second compartmental model focused on the liver. Since both the gut and liver express G-6P-phosphatase, they were described by means of two functional compartments that account for the significant exchange between trapped and free tracer, i.e., between FDG-6P and FDG. This compartmental system is described in Figure 1.
Gut subsystem
In these equations, C _{ g }, C _{ t }, and C _{ a } represent the tracer concentrations in the free compartment g, in the phosphorylated compartment t, and in the arterial blood a, respectively; the superposed dot indicates differentiation with respect to time, and the coefficients k _{ ij } (measured in min^{−1}) denote the rate coefficients to the target compartment i from the source compartment j. Equations (1) and (2) can be formally solved to obtain the analytical expressions of C _{ g } and C _{ t }. We observe that such expressions depend on the tracer coefficients that at this stage of the solution process are still unknown.
Also, in the case of Equation (3), the analytical solution can be formally determined and depends on the kinetics parameter k _{ pg }.
Liver subsystem
According to standard results [17], also the liver can be described as a two-compartment model, one consisting of non-metabolized, free, tracer (compartment f), and one consisting of phosphorylated, metabolized tracer (compartment m), where, as in the case of the gut, dephosphorylation is explicitly allowed. Tracer inputs through the HA and PV (compartments a and p, respectively) are modeled in an independent manner, i.e., there is no mixing of blood from the two vessels before entrance into the liver. On the contrary, possible heterogeneities in perfusion source may occur as to consider the most general input model. Accordingly, exchange coefficients for the HA blood (k _{ fa }) and the PV blood (k _{ fp }) are considered as independent. On the other hand, FDG delivery to the PV blood is considered to occur only from the gut, tracer exchange throughout the vein is negligible, and thus, the whole tracer amount entered into the PV is available for liver uptake.
Equations (4) and (5) can be again formally solved to obtain the analytical expressions of C _{ f } and C _{ m }. We observe again that such expressions depend on the tracer coefficients that at this stage of the solution process are still unknown.
Data optimization
In order to numerically solve Equations (6) and (7) and therefore to determine the tracer coefficients, we applied, separately and in cascade, a regularized multi-dimensional Newton algorithm [19], where a nice trade-off between the numerical stability of the problem solution and an appropriate fitting of the measured data were obtained by means of an optimized selection of the regularization parameter. To this aim, we first observed using simulations that the regularized Newton algorithm is rather robust with respect to the choice of the regularization parameter. In fact, in the case of a synthetic dataset, there exists a unique value of the regularization parameter that minimizes the distance between the reconstructed and ground-truth tracer coefficient vector. For all simulations performed, this value had an order of magnitude of around 10^{4}, and tuning such value in the range of 10^{3} to 10^{5} changed the reconstructed coefficients of less than 0.5%. In the case of experimental data, for each mouse, we applied a discrepancy approach: we chose as optimal value of the regularization parameter the value for which the discrepancy between the experimental data and the data predicted by the regularized solution coincided with the uncertainty on the measurement [20]. This uncertainty was computed by assuming that the noise on the activity is Poisson and the values of the regularization parameter we obtained were around 10^{4}, as in the simulated cases. We are aware that the Poisson assumption on the uncertainty does not account for correlation and other effects induced by the reconstruction algorithm; however, the robustness of regularization guarantees for the reliability of this approach.
From a computational viewpoint, the codes implementing the reduction of these compartmental models are extremely fast (less than 1 min for each analysis). Further, in order to assess the numerical robustness of the approach, for each animal model we have perturbed 50 times the PV TAC (obtained from the reduction of the gut subsystem) with 50 different Poisson random components, and accordingly, we have computed the tracer coefficients for the liver. For each coefficient, the standard deviation corresponding to the mean value over the several runs of the algorithm represents a reliable measure of the robustness of the approach with respect to uncertainties in the computation of the PV TAC. We found out that all standard deviations were below 4% of the mean values.
Results
Experimental protocol
Means and standard deviations for weight, glycemia, and SUV over the three sets
Weight ( g ) | Glycemia (mg/dl) | SUV gut | SUV liver | |
---|---|---|---|---|
CTR | 24.4 ± 1.8 | 131.1 ± 12.1 | 1.07 ± 0.36 | 1.07 ± 0.33 |
MTF | 23.9 ± 1.3 | 89.1 ± 18.6 | 2.49 ± 0.97 | 1.03 ± 0.31 |
STS | 23.7 ± 1.6 | 54.0 ± 24.3 | 1.51 ± 0.39 | 1.17 ± 0.27 |
Estimation of the PV TAC
Effects on liver FDG kinetics
Means and standard deviations for the gut and liver kinetics parameters over the three sets
k _{ fa } | k _{ mf } | k _{ sf } | k _{ fm } | k _{ fp } | k _{ ga } | k _{ tg } | k _{ gt } | |
---|---|---|---|---|---|---|---|---|
CTR | 2.180 ± 0.568 | 0.003 ± 0.003 | 2.51 ± 0.51 | 0.154 ± 0.059 | 2.291 ± 0.530 | 1.109 ± 0.331 | 0.036 ± 0.029 | 0.047 ± 0.029 |
MTF | 2.112 ± 0.420 | 0.004 ± 0.003 | 2.478 ± 0.389 | 0.041 ± 0.026** | 2.068 ± 0.757 | 0.902 ± 0.305 | 0.118 ± 0.073 | 0.029 ± 0.012 |
STS | 2.384 ± 0.689 | 0.095 ± 0.051* | 3.801 ± 1.650 | 0.563 ± 0.221** | 2.135 ± 0.646 | 0.963 ± 0.373 | 0.060 ± 0.008 | 0.181 ± 0.10 |
We further observe that a Patlak analysis of the same data did not allow us to differentiate the three groups, since we found statistically comparable Patlak slopes (1.09 ± 0.11 for CTR, 1.08 ± 0.12 for MTF, and 1.11 ± 0.24 for STS). Table 2 also reports an increase of k _{ sf } (3.801 ± 1.650 min^{−1} vs 2.51 ± 0.51 min^{−1}, p = 0.07 vs CTR) and an increase of k _{ fa } in the STS models (2.384 ± 0.689 min^{−1} vs 2.180 ± 0.568 min^{−1}, p = 0.07 vs CTR), which corresponds to a significant release of FDG from the mice liver to the bloodstream and a less significant input increase from the blood to the liver, induced by starvation. The effect of MTF treatment on these coefficients was negligible.
Finally, no statistically significant change in the values of k _{ fp } for the three groups was observed.
Discussion
This study introduces a computation-based, model-dependent method for the use of FDG to study glucose kinetics in the liver. The main advantage and novelty of this approach are that it is able to determine the dual input typical of liver physiology just by means of computation, thus avoiding: 1) the use of invasive catheter-based measurement procedures and 2) the major drawback of methods relying on the sole arterial IF, which are not able to account for the intrinsic variability of perfusion and glucose consumption in splanchnic organs. Other attempts to compute the PV IF have been performed ([8] and references within), but those cases rely on a heuristic modification of the arterial IF which is obtained once for all and applied to all animal models. On the contrary, our method is intrinsically model-dependent, since it utilizes compartmental analysis of the gut to directly estimate the PV IF, which, in its turn, is combined with the arterial TAC (obtained via ROIs on the aortic arc) to perform the compartmental analysis of the liver. From a numerical viewpoint, this task was accomplished applying the regularized Newton algorithm to the inverse problems associated to the multi-dimensional Cauchy problems describing the two-compartment gut and liver systems, respectively. In this setting, the computation of this algorithm, which is rather similar to other optimization algorithms applied in compartmental analysis [22-25], is rather effective, since the matrix differentiation step required at some stage can be realized analytically, thus, avoiding time consuming numerical differentiation. The advantages of this technique are the ability to correctly identify the response of rate constants to experimental interventions able to modify hepatic glucose metabolism and a notable robustness with respect to data optimization. The robustness of the method has been tested by computing the tracer coefficients for the liver in the case of different perturbed values of the PV TAC for each animal model. The corresponding standard deviations were always below 4% of the mean values, showing the notable numerical robustness of the numerical process.
Pathophysiological considerations
The liver plays a crucial role in the maintenance of glycemic levels despite a large variability in glucose consumption in the different body tissues. To this purpose, liver metabolism is regulated as to counterbalance the hyperglycemic response to food intake and to adapt serum glucose concentration to whole body needs under fasting periods. This buffer function is of pivotal importance since - once entered in the cytosol of almost all mammalian cells - glucose is sequestered by its phosphorylation to G6P. The irreversible nature of this reaction also applies to FDG whose accumulation as FDG-6P accounts for the potential of this tracer in the study of brain and tumor metabolism [26].
Coherently with its role in maintaining metabolic homeostasis, this simplified model does not apply to hepatocytes whose G6P content - either newly synthetized or coming from glycogen breakdown - can be hydrolyzed to P_{i} and glucose by G6Pase. This reaction replenishes cytosol content of glucose to be released into the bloodstream. In agreement with this homeostatic role, liver G6Pase catalytic function is downregulated by insulin response to feeding [12] while it is markedly enhanced by the high glucagon levels occurring during fasting periods [13].
Our results favorably compared with this well-validated biochemical model. In the present study, effect of feeding was not tested, as this condition abolishes the steady-state of serum glucose and insulin needed for an accurate compartmental analysis of liver metabolism. Nevertheless, STS increased rate constant k _{ fm } (as an index of FDG-6P dephosphorylation to FDG) about threefold with respect to 6 h of fasting, indicating a reliability of our measurements.
This finding was confirmed by the data obtained under the opposite condition. To this purpose, we studied animals exposed to prolonged treatment with high doses of MTF, the most widely used drug in the treatment of type 2 diabetes. MTF exerts its anti-hyperglycemic effect mainly decreasing hepatic gluconeogenesis via AMP-dependent protein kinase (AMPK) activation through the upstream kinase liver kinase B1 (LKB1) [7]. More importantly, this molecular response also downregulates or even silences the expression of all gluconeogenic enzymes and mostly of G6Pase [27]. Again, our approach to liver FDG kinetics nicely agreed with this mechanism of action. In fact, treatment with high MTF doses profoundly reduced k _{ fm } values, indicating a reduced rate of G6P hydrolysis as expected on the basis of a reduced G6Pase abundance. In this line, our data extend previous studies documenting an increase in hepatic glucose uptake under MTF treatment in subjects with type 2 diabetes [3-5]. This finding has been usually attributed to a direct increase in glucokinase function induced by the improved metabolic control. Our data indicate that, besides the enhancement in retention mechanism, MTF also decreases glucose and FDG output by inhibiting G6Pase activity. Even more interestingly, the response of k _{ fm } to both STS and MTF was paralleled by a relative invariance of k _{ fp } and k _{ sf } in the three mice groups analyzed, demonstrating a similar blood input (PV) and output (SV) in the liver vein system.
Interestingly, in STS models, the value of the rate constant from arterial blood to the liver free compartment (k _{ fa }) was increased, probably due to the low glucose concentration in the blood during caloric restriction and the consequent increased demand by the liver.
Conclusions
In conclusion, our data demonstrated the effectiveness of an innovative computational method to analyze the liver FDG kinetics under pharmacological or pathophysiological stimuli. Although we used a rather complex compartmental model characterized by nine unknowns, the regularized Newton algorithm was able to estimate all unknowns with a satisfactory degree of robustness. This permitted to quantify the fraction of tracer trapped in the liver or dephosphorylated and released into the bloodstream and thus conveyed to the different tissues.
Consent
Informed consent was obtained from all individual participants included in the study.
Declarations
Acknowledgements
SG, VV, and MP kindly acknowledge the support of a grant of the Italian INdAM-GNCS.
Authors’ Affiliations
References
- Basu A, Basu R, Shah P, Vella A, Johnson CM, Nair KS, et al. Effects of type 2 diabetes on the ability of insulin and glucose to regulate splanchnic and muscle glucose metabolism: evidence for a defect in hepatic glucokinase activity. Diabetes. 2000;49:272–83.PubMedView ArticleGoogle Scholar
- Choi Y, Hawkins RA, Huang SC, Brunken RC, Hoh CK, Messa C, et al. Evaluation of the effect of glucose ingestion and kinetic model configurations of FDG in the normal liver. J Nucl Med. 1994;35:818–23.PubMedGoogle Scholar
- Iozzo P, Hallsten K, Oikonen V, Virtanen KA, Kemppainen J, Solin O, et al. Insulin-mediated hepatic glucose uptake is impaired in type 2 diabetes: evidence for a relationship with glycemic control. J Clin Endocrinol Metab. 2003;88:2055–60.PubMedView ArticleGoogle Scholar
- Iozzo P, Geisler F, Oikonen V, Mäki M, Takala T, Solin O, et al. Insulin stimulates liver glucose uptake in humans: an 18 F-FDG PET study. J Nucl Med. 2003;44:682–9.PubMedGoogle Scholar
- Iozzo P, Hallsten K, Oikonen V, Virtanen KA, Parkkola R, Kemppainen J, et al. Effects of metformin and rosiglitazone monotherapy on insulin-mediated hepatic glucose uptake and their relation to visceral fat in type 2 diabetes. Diabetes Care. 2003;26:2069–74.PubMedView ArticleGoogle Scholar
- Munk OL, Bass L, Roelsgaard K, Bender D, Hansen S, Keiding S. Liver kinetics of glucose analogs measured in pigs by PET: importance of dual-input blood sampling. J Nucl Med. 2001;42(5):795–801.PubMedGoogle Scholar
- Jalving M, Gietema JA, Lefrandt JD, De Jong S, Reyners AK, Gans RO, et al. Metformin: taking away the candy for cancer? Eur J Cancer. 2010;46(13):2369–80.PubMedView ArticleGoogle Scholar
- Keiding S. Bringing physiology into PET of the liver. J Nucl Med. 2012;53(3):425–33.PubMed CentralPubMedView ArticleGoogle Scholar
- Ferl GZ, Zhang X, Wu HM, Kreissi MC, Huang SC. Estimation of the FDG input function in mice by use of dynamic small animal PET and minimal blood sample data. J Nucl Med. 2007;48(12):2037–45.PubMed CentralPubMedView ArticleGoogle Scholar
- Kudomi N, Järvisalo MJ, Kiss J, Borra R, Viljianen A, Savunen T, et al. Non-invasive estimation of hepatic glucose uptake from [18 F] FDG PET images using tissue-derived input functions. Eur J Nucl Med Mol Imaging. 2009;36(12):2014–26.PubMedView ArticleGoogle Scholar
- Massollo M, Marini C, Brignone M, Emionite L, Salani B, Riondato M, et al. Metformin temporal and localized effects on gut glucose metabolism assessed using 18 F-FDG-PET in mice. J Nucl Med. 2013;54(2):259–66.PubMedView ArticleGoogle Scholar
- Minassian C, Daniele N, Bordet JC, Zitoun C, Mithieux G. Liver glucose-6 phosphatase activity is inhibited by refeeding in rats. J Nutr. 1995;125(11):2727–32.PubMedGoogle Scholar
- Mithieux G, Vidal H, Zitoun C, Bruni N, Daniele N, Minassian C. Glucose-6-phosphatase mRNA and activity are increased to the same extent in kidney and liver of diabetic rats. Diabetes. 1996;45(7):891–6.PubMedView ArticleGoogle Scholar
- Munk OL, Keiding S, Bass L. Impulse-response function of splanchnic circulation with model-independent constraints: theory and experimental validation. Am J Physiol Gastrointest Liver Physiol. 2003;285(4):G671–80.PubMedView ArticleGoogle Scholar
- Munk OL, Bass L, Feng H, Keiding S. Determination of regional flow by use of intravascular PET tracers: microvascular theory and experimental validation for pig livers. J Nucl Med. 2003;44(11):1862–70.PubMedGoogle Scholar
- Winterdahl M, Keiding S, Sørensen M, Mortensen FV, Alstrup AK, Munk OL. Tracer input for kinetic modelling of liver physiology determined without sampling portal venous blood in pigs. Eur J Nucl Med Mol Imaging. 2011;38:263–70.PubMed CentralPubMedView ArticleGoogle Scholar
- Rani SD, Nemanich ST, Fettig N, Shoghi KI. Kinetic analysis of FDG in rat liver: effect of dietary intervention on arterial and portal vein input. Nucl Med Biol. 2013;40(4):537–46.PubMedView ArticleGoogle Scholar
- Winterdahl M, Munk OL, Sørensen M, Mortensen FV, Keiding S. Hepatic blood perfusion measured by 3-minute dynamic 18 F-FDG PET in pigs. J Nucl Med. 2011;52(7):1119–24.PubMed CentralPubMedView ArticleGoogle Scholar
- Nocedal J, Wright S. Numerical optimization. New York, NY: Springer; 1999. p. 1–10.View ArticleGoogle Scholar
- Tikhonov AN, Goncharsky AV, Stepanov VV, Yagola AG. Numerical methods for the solution of ill-posed problems. Dordrecht: Kluwer; 1995.View ArticleGoogle Scholar
- Marini C, Salani B, Massollo M, Amaro A, Esposito AI, Orengo AM, et al. Direct inhibition of hexokinase activity by metformin at least partially impairs glucose metabolism and tumor growth in experimental breast cancer. Cell Cycle. 2013;12(22):3490–9.PubMed CentralPubMedView ArticleGoogle Scholar
- Slimani L, Kudomi N, Oikonen V, Jarvisalo M, Kiss J, Naum A, et al. Quantification of liver perfusion with [(15)O]H(2)O-PET and its relationship with glucose metabolism and substrate levels. J Hepatol. 2008;48(6):974–82.PubMedView ArticleGoogle Scholar
- Garbarino S, Caviglia G, Sambuceti G, Benvenuto F, Piana M. A novel description of FDG excretion in the renal system: application to metformin-treated models. Phys Med Biol. 2014;59(10):2469–84.PubMedView ArticleGoogle Scholar
- Garbarino S, Caviglia G, Brignone M, Massollo M, Sambuceti G, Piana M. Estimate of FDG excretion by means of compartmental analysis and ant colony optimization. Comput Math Methods Med. 2013;2013:10.View ArticleGoogle Scholar
- Limber MA, Limber MN, Cellar A, Barney JS, Borwein JM. Direct reconstruction of functional parameters for dynamic SPECT. IEEE Trans Nucl Sci. 1995;42(4):1249–56.View ArticleGoogle Scholar
- Wang G, Qi J. Direct estimation of kinetic parametric images for dynamic PET. Theranostics. 2013;3(10):802–15.PubMed CentralPubMedView ArticleGoogle Scholar
- Sokoloff L, Reivich M, Kennedy C, Des Rosiers MH, Patlak CS, Pettigrew KD, et al. The [^{14}C]deoxyglucose method for the measurement of local cerebral glucose utilization: theory, procedure and normal values in the conscious and anesthetized albino rat. J Neurochem. 1977;28(5):897–916.PubMedView ArticleGoogle Scholar
Copyright
This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited.