Differentiation of fibrotic liver tissue using laser-induced breakdown spectroscopy

: Hepatic cirrhosis is a major cause of morbidity and mortality worldwide due to hepatitis C, alcoholism and fatty liver disease associated with obesity. Assessment of hepatic fibrosis relies in qualitative histological evaluation of biopsy samples. This method is time-consuming and depends on the histopathologists’ interpretation. In the last decades, noninvasive techniques were developed to detect and monitor hepatic fibrosis. Laser-induced breakdown spectroscopy (LIBS) is a good candidate for a real-time, independent and fast technique to diagnose hepatic fibrosis. In this work LIBS was employed to characterize rat liver tissues with different stages of fibrosis. Depth profiling measurements were carried out by using a nanosecond Nd:YAG laser operated at the fundamental wavelength and an echelle spectrometer coupled with an ICCD camera. Due to the soft nature of the samples, plasma conditions largely change between consecutives shots. Thus, a theoretically supported procedure to correct the spectral line intensities was implemented. This procedure allows the reduction of the intensities’ dispersion from 67% to 12%. After the correction, the LIBS signal shows an enhancement in calcium intensity by a factor of three as the fibrosis progressed. Calcium is known to increase


Introduction
Hepatic fibrosis is a reversible wound-repair response to chronic liver injury. The most important histological characteristic of hepatic fibrosis is the excessive accumulation of extracellular matrix (ECM) proteins (collagens, non-collagen glycoproteins, and proteoglycans). The thickening of ECM leads to progressive substitution of hepatic architecture by fibrous septa and chemical cross-linking of collagen and other ECM proteins; the end stage of this process leads to cirrhosis. The main causes of hepatic fibrosis are chronic hepatitis C virus (HCV) infection, fatty liver, alcohol and drugs abuse and metabolic or autoimmune diseases [1][2][3].
In spite of the fact that liver biopsy is useful as a reference for direct histological evaluation of hepatic diseases, it cannot be used as a serial monitoring technique. This is due to the technique's invasiveness, morbidity and mortality, and the heterogeneous distribution of fibrosis in liver. Other disadvantages of biopsies include: cost, time-consuming processes, lack of standardization of the staining used to highlight the fibrotic tissue and the intrinsic observer-dependent diagnostic variability [4]. These problems have resulted in the development of non-invasive techniques to detect and monitor fibrosis. Serum biomarkers provide accurate diagnosing for advanced cases of fibrosis and cirrhosis, but they do not provide a classification for liver fibrosis [5,6]. Elastography characterize the response and mechanical properties of tissue and accurately determine the higher grades of fibrosis [7]. This test is not practical for patients with accumulation of fluid in the peritoneal cavity, with small intercostal spaces or obesity [8]. However, the implementation of non-invasive diagnosis test for liver disease differs from one country to another. Lastly, liver biopsy remains the mainstream test to stage accurately liver fibrosis, generally through METAVIR system, which determinates the fibrosis level among five discrete levels [9,10]. The development of new techniques capable of detecting fibrosis at the earliest stages can be an important contribution to the early management of patients.
Laser-Induced Breakdown Spectroscopy (LIBS) is an optical technique that allows multielemental analysis of any kind of sample (solid, liquid and gas) [11][12][13]. In this technique, a pulsed laser is focused onto a sample ablating a small amount of material producing a plasma. The emitted radiation is collected and analyzed by a spectrometer to obtain qualitative and quantitative analytical information of sample species [14,15]. LIBS is an appealing technique because it allows local analysis of samples and the capability of multielement analysis that other analytical techniques are not capable to perform nor detect. LIBS has been used in several applications as the analysis of geological samples [16], environmental monitoring [17], process control [18], archeological artifacts [19,20] and most recently, in space exploration [21]. Furthermore, LIBS was used in characterization and differentiation of tissues [22], bacteria [23], normal and malignant tissues [24,25] and detection of biological agents [26].
In order to obtain the sample's quantitative composition by LIBS it is necessary to perform a calibration using samples with similar composition to the investigated ones. In the case of biological samples, is difficult to find matrix-matched standards. A theoretical tool, known as "calibration free LIBS" (CF-LIBS), has improved LIBS' capabilities [27]. To perform CF-LIBS, the plasma has to be optically thin and must be under local thermal equilibrium (LTE). Additionally, it is assumed that the sample is stoichiometrically ablated of and that it is necessary to acquire a wide range spectrum to detect all the constituents of the target. In simple matrices like binary alloys, CF-LIBS gives good quantitative results [28]; in rocks and siliceous soils results show a discrepancy for the major constituents up to 20-30% [29].
Other routines have been also exploited to carry out sample identifications. Kumar et al, for example, employed an internal standard to identify malign tumor cells in dogs' liver samples [30]. However, the use of internal standards could produce inconsistent results when water is present in the analyzed tissues [31]. Other works [22,32], implemented multivariate statistical techniques to distinguish between different samples. In particular, the application of Principal Component Analysis (PCA) has demonstrated its worth to find patterns in confusing data sets [33]. On the other hand, a theoretically supported correction to reduce shot-to-shot intensity fluctuations was applied to archeological ceramics samples [34]. The intensity correction allows to detect differences between the elemental composition of decorative layers and the bulk material.
The purpose of this work is to employ the LIBS technique to establish whether a correlation exists in the elemental concentration in fibrotic and healthy samples of liver rats. Hepatic fibrosis was induced in 16 rats by injection of carbon tetrachloride for LIBS characterization. Samples differentiation was performed by the statistical analysis through the PCA approach. Also, it was implemented the correction reported in the work of Lazic et al. [34] to reduce the relative standard deviation for several consecutive shots and perform a univariate analysis. Then, the LIBS signals obtained from liver samples with different grades of hepatic fibrosis were compared to histological analysis, which was taken as a reference.

Samples
The animal model to induce fibrosis in rats was described elsewhere [9]. Briefly, 16 male Wistar rats were separated in 4 groups (named T1, T2, T3 and T4) of 4 rats each. Accordingly, 16 rats received an initial dose of 250 µl/g of carbon tetrachloride (CCl 4 ) diluted 1:3 in olive oil, twice a week by intraperitoneal administration. Later doses were adjusted according to the weight of the rodents. In addition, 4 control rats were injected only olive oil which acted as a placebo. The specimens in groups T1 to T4 received CCl 4 for 4, 6, 10 and 18 weeks, respectively. After the established time for each group, the animals were slaughtered and the livers were removed and fixed in 3.8% phosphate-buffered formaldehyde.

Histological analysis
Tissue samples of the left lateral lobe's livers were histologically evaluated according to the METAVIR scoring system in order to diagnose the disease stage [9]. Ten images with a 4x magnification were taken from each sample and three histopathologists independently evaluated all the images to diagnose the degree of fibrosis. Hence, 160 images corresponding to 16 samples were analyzed following this procedure. It is worth to mention that this analysis is larger in extent and with more repetitions than a conventional histopathological examination performed on a biopsy sample.
The METAVIR system generates a score from 0 to 4. The system describes fibrosis as follows: F0, no fibrosis; F1, portal fibrosis without septa; F2, portal fibrosis with a few septa; F3, numerous septa without cirrhosis and F4, cirrhosis [10]. Figure 1 shows the frequency histogram of the METAVIR scores among the groups T1, T2, T3 and T4. As it can be observed, the dispersion of the disease for different groups is not uniform, probably due to individual response to CCl 4 and liver fibrosis expression in aggregates. For the four groups T1, T2, T3 and T4 there is predominance of F1, F2, F3 and F4 METAVIR score, respectively, for at least 50% of the evaluations. However, there is some statistical dispersion of the data to the neighboring scores and the larger data dispersion was observed for group T1. The diagnosis for each sample was the statistical mode of the obtained results. Finally, it was diagnosed that 3 livers belong to F1, 5 to F2, 5 to F3 and 3 to F4. The other 4 control rats were assumed to be F0.

Laser-induced breakdown spectroscopy
For LIBS measurements, liver samples were desiccated in a vacuum chamber at 10 −3 Torr. This was performed to enhance LIBS signal and reduce pulse-to-pulse fluctuations. The experimental arrangement is shown in Fig. 2. The laser employed was a Nd:YAG laser (Quantel, Brilliant EaZy) emitting 4-6 ns pulses at 1064 nm and the frequency was set to 1 Hz. Laser beam was expanded with a telescope up to 25-mm diameter and was focused onto the sample by using a 15 cm plano-convex lens. Laser energy was kept fixed at 50 mJ corresponding to a fluence of about 170 J cm −2 . All experiments were performed in air at atmospheric pressure. Samples were mounted on an x-y-z translation stage for an accurate positioning of the target. Analysis was performed at 5 different sample locations acquiring 80 individual shots in each of them. The spatially integrated plasma emission was collected by a fused silica collimator located at 4 cm from the laser focusing point at about 45° with respect to the target surface. The collimator was connected to a fiber optic bundle that channeled the light to an Echelle spectrograph (LTB, Aryelle 200) with a resolving power of Δλ/λ = 7500 in the spectral range of 220 -830 nm. Dispersed light was detected with an intensified charge-coupled device (ICCD) camera (ANDOR, iStar 334-18F-03). To optimize the signalto-noise ratio (SNR), spectra with different time delays and gate widths were recorded; the optimum obtained values were 580 ns and 20 µs, correspondingly. Finally, transition intensities were calculated from the area under the curve of the observed lines.  Figure 3 shows a typical spectrum of a healthy rat liver tissue. The elements detected were: carbon, phosphorus, iron, magnesium, CN band, calcium, sodium, potassium, nitrogen, hydrogen and oxygen. This spectrum results from the averaging of 20 single-shot spectra. However, for copper, iron, phosphorous and potassium, the measured SNR of the single shots experiments were near the limit of detections. The contribution of oxygen, nitrogen and hydrogen was not considered in this work as they are elements present in the surrounding air. The list of the main transitions observed is given in Table 1.  Even though all the experimental conditions were kept fixed, due to the soft nature of the samples and the variation of plasma conditions during drilling, a large pulse-to-pulse fluctuation took place (see Fig. 4). Our results show that the relative standard deviation (RSD) recorded at different spots under the same shot count can reach the 50% mark. Besides, it was observed that the first 3-5 shots present higher intensity values. However, after these initial pulses, the intensity profile remained almost constant. A similar trend has been reported elsewhere in analysis of thick samples, such as metals and ceramics [35]. The first high intensities values observed in Fig. 4 can be attributed to surface roughness and sample curvature. Subsequent signal fluctuations for different spots and between shots can be mainly attributed to variation in plasma conditions. The differentiation of fibrosis degree may be accomplished through statistical multivariate analysis. Thus, the Mathematica multivariate statistic package was employed to reduce the dimensionality of the original data set and differentiate between samples. For this purpose, five spectral regions were selected including the most important observed transitions, namely, carbon, calcium, magnesium, sodium and potassium. Afterwards, all spectra from samples with the same METAVIR score were averaged (i.e. equivalent shots for different spots). Then PCA was applied to the 240 averaged spectra (60 shots x 5 METAVIR scores). The first 20 shots were not included in the calculation due to high fluctuation observed in transition intensities.

Results and discussion
PCA calculates principal components (PC), which are linear combinations of variables that carry most of the information of the original data. PC are typically displayed in a score plot where samples with similar characteristics form clusters [36]. Figure 5 shows the plot of PCA scores for the first two principal components covering the 61.9% and 26.0% of the total variance of the system. Also, the 95% confidence ellipses are shown in the Figure. As it can be appreciated, cirrhotic samples (F4) can be differentiated from samples with different fibrosis diagnostics. However, it was observed a partial overlap of the confidence ellipses for the other stages (F0-F3). Furthermore, univariate analysis using an internal standard such as carbon and intensity averaging between consecutive shots, does not allow the identification between METAVIR stages. This can be mainly attributed to intensity fluctuations due to plasma temperature variations. In Ref [34], a procedure to reduce the RSD of homogeneous samples due to variable plasma conditions was proposed. This method could be applied to increase LIBS precision without performing the calculation of plasma temperature. The homogeneity of rat liver samples has been previously investigated by Cockell et al. [37] using different analytical techniques. This analysis demonstrated that there is no statistically differences in the elemental composition among different regions of liver. Thus, it is possible to implement the procedure developed in [34] in this work.
The corrected line intensity (I c ) can be written in terms of the experimentally measured value I e as: ij ij e c P ij where the spectral integral S is the area under the curve of the whole intensity-wavelength graph, the "i" index is referred to the shot number (up to 80) and the "j" index correspond to the sampling spot number (from 1 to 5). P is a coefficient that minimizes the averaged relative standard deviation (RSD av ) of the corrected line intensities: where N is the considered number of shots per sampling point and RSD i is the relative standard deviation of the corrected intensities (I c ij ) between equivalent shots in the five different sampling spots: where Ī c i is the average intensity for the five equivalent shots. As an example of the method, we illustrate the correction with the Mg II 279.55 nm transition. For each sample, five depth profiles were completed and using Eqs. (1), (2) the coefficient P was determined when the index i is varied from 40 to 80. Using the value obtained for P, we corrected the LIBS depth profiles intensities. Afterwards, we performed the average of the corrected line intensities for the same shot at different spots (Ī c i ). Figure 6 shows a comparison between uncorrected and corrected averaged intensities. As can be seen, the dispersion in intensity for different shot numbers was significant reduced, from 49% to 16%. Furthermore, the RSD for the same shot number in different spots was also reduced, which is translated in smaller error bars. Table 2 shows a summary of the calculated P coefficient for the main observed transitions and its corresponding RSD before and after the correction. To relate the obtained LIBS signal values with the corresponding METAVIR score, we averaged the corrected line intensities for all the samples with the same diagnostic. For instance, since we have only three livers with F1 diagnostic, we performed the average of the corrected LIBS signals for these three traces. Finally, the obtained trace was averaged over the 80 shots to obtain a single value score with its standard deviation. Figure 7 shows normalized results values for magnesium (a), carbon (b) and calcium (c). As can be seen from Fig. 7(a) and 7(b), the obtained intensities remain almost unchanged for the different fibrosis stages. On the other hand, calcium presents an increasing trend with the fibrosis stage (see Fig. 7(c)), reaching a maximum value close to three. Here, a high correlation was observed between LIBS intensity for calcium and the METAVIR scores, which is reflected in a Pearson's correlation coefficient of 0.98.
The increment in calcium content in liver could be related with the age of the animals. Nevertheless, Shimamura et al. [38] measured the concentrations of 22 elements in rat's livers samples up to 113 weeks age by inductively coupled plasma analysis. Here, it was obtained that concentration of Mg, P, K, Fe and Cu was almost constant across the ages. On the contrary, calcium content showed variations related with age, but with non-monotonic behavior.
Although calcium is not a major constituent of the buffered formaldehyde employed, we investigated a possible contamination of samples during the fixation process. Therefore, we analyzed the LIBS signal of an additional healthy rat liver. For this experiment, part of the left lateral lobe's liver was desiccated in vacuum as described above. The other part of the sample was fixed with the employed buffered formaldehyde for several weeks and afterwards it was also desiccated. As it should be expected, the LIBS signal intensity of both samples did not present any meaningful variation. A similar result was obtained in previous work when comparing the elemental composition of liver samples fixed in acetone with others dried in air [30].
Consequently, calcium enhancement can be linked to the fibrosis progression. Calcium ions (Ca 2+ ) regulate several cellular processes, such as hormonal responses and cell death and its concentration rises in response to physiological stimuli [39]. Schanne et al. [40] studied the role of intracellular Ca 2+ in liver injury caused by carbon tetrachloride (CCl 4 ). Here, it was concluded that intracellular accumulation of calcium is the final common pathway by which toxic cell death occurs. On the other hand, crosslinking of collagen and other ECM proteins as proteoglycans, is a crucial process in the stabilization of fibrous septa that contributes to the progress and avoids the reversibility of the disease [41]. Extracellular Ca 2+ binds to ECM proteins by electrostatic force [42] and augments their stability. Also, calcium is needed for crosslinking activity of enzymes as transglutaminase 2 [43,44]. Therefore, an increase in Ca 2+ could aid the fibrosis progression.

Conclusions
METAVIR is the gold standard for assessing liver fibrosis stage in clinical management of patients. Besides the invasiveness of this technique, this method is arduous, time-consuming and it requires histopathologists interpretation.
In this work, LIBS was proposed to characterize hepatic fibrosis induced in rats. Also, liver samples were histologically evaluated according to the METAVIR scoring system to serve as a reference. Results shows that PCA was able to distinguish cirrhotic samples, but a superposition of other stages was observed. This is probably due to variable plasma conditions causing a large shot-to-shot intensity variation. The semi-empirical employed method allowed to reduce the RSD by a factor of five. Besides it was possible to successfully link calcium intensities with METAVIR score values. The increment observed in calcium concentration can be correlated with intracellular or extracellular calcium accumulation observed in cirrhosis. Although the obtained results are preliminary, LIBS could serve as an independent fast technique to accurately diagnose an early stage of hepatic fibrosis.

Funding
This work was supported by the Dirección General de Asuntos del Personal Académico, Universidad Nacional Autónoma de México (DGAPA-UNAM IN105510) and the Consejo Nacional de Ciencia y Tecnología, Mexico (CONACyT 221506-F).

Disclosures
The authors declare that there are no conflicts of interest related to this article.