2.1 Definition of the S_value

In clinical settings, accurate localization of exogenous elements within biological tissue is paramount. This necessitates the utilization of histological images obtained from adjacent sections as a referential framework. One primary aim of this elemental imaging study was to generate images depicting endogenous elements, such as phosphorus (P), magnesium, and sodium, in order to reconstruct a tissue architecture that closely mirrors its histological counterpart. Subsequent imaging focuses on the identification of exogenous elements, particularly various metals that may be present in either microparticle or nanoparticle form. The data collected from these imaging techniques can be substantial, posing a challenge for routine clinical application due to the sheer volume of specimens and images to be analyzed. Additionally, the contrast within the images complicates the quantification of exogenous elements, a challenge further exacerbated when examining metal particle distributions in lung tissue, which frequently manifest as isolated pixels. To address this issue, we introduced a quantitative metric, designated as $S_{\text{value}}$. This metric facilitates an easier interpretation by clinicians, enabling the rapid identification of abnormal elemental concentrations. When warranted, clinicians can then scrutinize the images in greater detail to elucidate the spatial distribution of these elements within the tissue.

The challenge in determining a $S_{\text{value}}$ arises from the intricate nature of the samples under investigation: tissues embedded in paraffin. Existing market offerings lack commercial calibration standards adapted to our study, i.e., paraffin-embedded lung tissue reference materials that possess a homogeneous and known concentration of chemical elements of interest, especially for Ti. Consequently, we adopted a strategy utilizing lung specimens from animals exposed to relevant conditions, thereby establishing a calibration curve specifically for the extraction of a metric $\left(S_{\text{value}}\right)$ for Ti. Our dual objectives encompassed evaluating analytical figures of merit, such as linearity, detection limits, and uncertainties, and validating the hypothesis that an elemental analysis of a lung section could be representative of the entire lung volume.

This calculation requires to address several challenges such as establishing a procedure to define the tissue area, defining an appropriate methodology to isolate the Ti line intensities and their corresponding uncertainties.

2.2 Spectral Characterization and Elemental Mapping of Endogenous and Exogenous Markers in Lung Tissue

In the initial phase of the study, emission lines for P and Ti were selected to avoid spectral interference, defined as the overlapping of emission lines from different elements. Phosphorus, an integral component of DNA and biomembranes, serves as an endogenous marker because its elemental image indirectly maps the overarching architecture of the biological tissue. Conversely, Ti serves as an exogenous marker, specifically denoting the presence of inhaled TiO₂ nanoparticles within the animal lung tissue. Through the employment of a carefully defined spectral window, we achieved the detection of intense emission lines corresponding to both elements. Figure 1a,b displays representative relative abundance maps for P and Ti obtained after LIBS imaging procedure.

To our knowledge, the megapixel elemental images presented in Figure 1a–c are among the few of their kind generated from biomedical specimens using LIBS. The 20 μm lateral resolution enables a comprehensive characterization of the tissue, and a closer examination of the selected specimen (Figure 1) reveals a porous lung tissue architecture delineated into two phases: a centrally located, phosphorus-rich region, and a peripherally located region with lower phosphorus concentration. Titanium displayed a largely homogeneous distribution within the lung tissue, as illustrated in Figure 1c. This homogeneity attests to the efficacy of the inhalation exposure procedures employed.^[20]

The spectral window chosen for the spectrometer accommodates multiple emission lines for the elements of interest, including two lines for P and numerous lines for Ti (as shown in Figure 1d), ensuring an accurate detection of the elements of interest. Moreover, this spectral window also encompasses primary or intense emission lines for other elements of interest in the context of inhalation exposure, such as aluminum (Al), copper (Cu), silicon (Si), chromium (Cr), iron (Fe), beryllium (Be), nickel (Ni), and zirconium (Zr). Figure 1d features four emission spectra with different Ti intensities, each shifted along the y-axis for clarity.

Due to its high dynamic range, LIBS emerges as an exceptionally suitable technique for detecting exogenous elements across a broad concentration spectrum—from ppm to tens of percentage points. It is also adapted for detecting all the endogenous elements which are ubiquitous in the tissues. This capability facilitates monitoring of specific elements across six orders of magnitude in concentration, all while maintaining rapid data acquisition and imposing minimal experimental constraints.

2.3 From Elemental Imaging to Quantitative Analysis

Upon the acquisition of LIBS elemental maps for abundance quantification, the background correction and surface contamination become critical because it becomes imperative to rigorously correct and control the signal to mitigate potential artifacts. While this issue of contamination is not unique to paraffin-embedded samples, we employed analytical chemistry-specific thresholds, namely, the limit of detection (LoD)^[21] and the limit of quantification (LoQ),^{[22, 23]} to effectively filter the entire data set.

These thresholds are conventionally derived from signal of the noise. Specifically, the LoD functions as a tissue-specific threshold, while the LoQ operates as a signal threshold. For the purposes of noise extraction, we focused on spectra corresponding to tissue within a 2 nm spectral window, ranging from 320 to 322 nm—a range out but close to the Ti emission line of interest. The noise was determined using a substantial dataset of 32 000 spectra, as illustrated in Figure 2d.

As a first analysis step, we delineated the area of interest by constructing a mask derived from the P elemental map, which corresponds to the tissue surface, denoted as $A_{\text{tissue}}$ (Figure 2a,b). The LoD serves as the tissue threshold, facilitating the differentiation between paraffin-embedded regions and genuine tissue. Specifically, if the P signal surpasses this LoD threshold, the region is classified as tissue. Conversely, signals falling below the LoD are likely attributable to noise fluctuations and are consequently excluded from consideration. To quantify the uncertainty associated with this threshold, we assessed the deviation in $A_{\text{tissue}}$ when altering the standard deviation (σ) thresholds from 3σ to both 2σ and 4σ. This sensitivity analysis yielded a surface deviation of ≈6%.

As a second step, a filtration process is applied specifically to the Ti intensities within the tissue mask. The signal threshold for this stage is established at the LoQ, and intensities falling below this cutoff are systematically excluded. As a third step, the sum of all intensities corresponding to all Ti-positive pixels is then normalized by the tissue area ($A_{\text{tissue}}$). The resultant metric, termed $S_{\text{value}}$, serves as a quantitative indicator of the total Ti content per unit of tissue area and allows for intersample comparability.

In alignment with analytical chemistry standards, a 10σ threshold was employed for signal discrimination. It should be noted, however, that variations in the number of σ applied for signal thresholding exert a minimal influence on $S_{\text{value}}$, given that nontissue spectra have already been eliminated through the initial tissue-specific thresholding. Detailed sensitivity analyses regarding the deviation of $S_{\text{value}}$ for different σ signal thresholds linked to the Ti accumulation revealed an associated uncertainty margin of 0.89% by comparing the reference $S_{\text{value}}$ to those for ± 3σ (Figure 2e).

2.4 Calibration Methodology

A cohort of six rat specimens from each exposure group—categorized as low, medium, and high—was subjected to ICP-MS for the quantification of total Ti mass in the exposed lung tissues. In each animal, the cranial right lung lobe underwent digestion and subsequent ICP-MS analysis for Ti quantification. Concurrently, the caudal right lung lobe was preserved through formalin fixation and paraffin embedding for LIBS analysis.

The computational approach delineated earlier for calculating $S_{\text{value}}(\mathrm{Ti})$ was effectively implemented on the set of 12 selected lung samples, facilitated by a fully automated MATLAB code. A comparative analysis was performed between the Ti concentrations obtained through ICP-MS and the $S_{\text{value}}(\mathrm{Ti})$ ascertained via LIBS imaging. This comparison was executed on a subset of three samples from each exposure group, as illustrated in Figure 3.

Despite uniform exposure to TiO₂ nanoparticles across all exposure groups, heterogeneities were observed in Ti concentrations within individual samples (e.g., variations in $S_{\text{value}}(\mathrm{Ti})$ within the low-exposure cohort, refer to Figure 3a). The calibration curve, depicting $S_{\text{value}}(\mathrm{Ti})$ against ICP-MS (Ti) concentrations (see Figure 3b), demonstrated remarkable linearity, with a coefficient of determination (R²) exceeding 0.999. Importantly, the calibration curve intersects the origin within the bounds of model uncertainties, substantiating the adequacy of the $S_{\mathrm{value}}$ extraction methodology.

Here, ${\sigma }_{\text{fit}}$ and m denote the standard error of the fit and the slope of the calibration curve, respectively. This innovative methodology achieves sub-ppm sensitivity, with a LoD registering at 0.05 μg g⁻¹ and a LoQ at 0.16 μg g⁻¹.

The calibration curve's exceptional linearity yields two pivotal insights: i) The $S_{\text{value}}(\mathrm{Ti})$ obtained through LIBS imaging strongly correlates with the total Ti content measured via ICP-MS. This suggests that $S_{\text{value}}\hspace{0.17em}$ serves as a robust quantitative indicator for total Ti content per unit mass of tissue; and ii) $S_{\text{value}}(\mathrm{Ti})$ is proportional to the levels of exposure to inhaled TiO₂ nanoparticles in rats, affirming the robustness of LIBS imaging for evaluating respiratory exposure to xenobiotic substances.

While ICP-MS provides volumetric measurements for entire lung lobes, LIBS remains a surface-sensitive technique with a depth of analysis approximating 10 μm in paraffin-embedded samples. The observed positive correlation between surface-level Ti signals obtained via LIBS and global Ti mass quantified by ICP-MS substantiates that LIBS surface measurements may serve as viable proxies for volumetric lung content. This is contingent upon a homogeneous distribution of the particles of interest—specifically, TiO₂ nanoparticles—induced by the exposure. In the current study, the similarity in Ti content between cranial and caudal lung lobes further corroborates that the exposure protocol ensured a uniform distribution of inhaled TiO₂ nanoparticles throughout the lung volume.

2.5 Influence of Lateral Resolution on Analytical Efficacy

The high-definition elemental image illustrated in Figure 1 demands ≈3 h for comprehensive analysis, given that the used LIBS instrumentation, operating at 100 Hz, can record up to 360 000 spectra per hour. While this time investment aligns well with academic research requirements, it presents logistical challenges for routine clinical applications. Notably, the acquisition time is inherently dependent on two factors: the lateral resolution and the surface area of the sample under investigation. For human specimens, which can span up to 6 cm², the time commitment becomes prohibitive, requiring around 40 min per cm² at the lowest feasible lateral resolution of 20 μm.

To reconcile the need for rapid analysis with analytical rigor, we scrutinized the impact of diminishing lateral resolution—defined as the step size or inter-laser-shot distance—on the $S_{\text{value}}$ metric. Given that exogenous elements in lung tissues often manifest as discrete particles (isolated pixels), excessive degradation of the resolution could compromise the elemental image accuracy. To model this, we employed the original high-definition image from Figure 1 to artificially generate submaps with reduced resolution via random pixel selection. For example, within a 4 × 4 pixel square (initially captured at a 20 μm step size), a single pixel was randomly chosen to reconstruct an image at an 80 μm resolution. This process was iteratively applied to generate 100 degraded images for each examined step size, ranging from 20 to 200 μm, and the corresponding $S_{\text{value}}(\mathrm{Ti})$ was computed for each.

As illustrated in Figure 4, the mean $S_{\text{value}}(\mathrm{Ti})$ for degraded images closely approximates the reference $S_{\text{value}}(\mathrm{Ti})$ derived from high-resolution imaging. However, statistical uncertainties, represented by standard deviations, increase with resolution deterioration, peaking at ≈10% for a tenfold reduction in resolution (i.e., 200 μm). This deterioration in resolution, however, leads to a significant reduction in analysis time by a factor of 100. Importantly, a moderate resolution shift from 20 to 50 μm decreased the acquisition time from 3 h to 25 min while maintaining a standard deviation below 5%. Moreover, the elemental distribution of Ti remained consistent at this 50 μm resolution, mirroring the homogeneity observed in the original high-definition image. Consequently, a 50 μm resolution emerges as an optimal compromise between analytical speed and confidence in quantitative assessments. Note that recent technologies have improved the laser repetition rate up to the kHz, and kHz LIBS imaging has been demonstrated.^[25]

2.6 Quantification of Analytical Uncertainties

Here, ${\sigma }_{\Delta x}$, ${\sigma }_{\text{surf}}$, ${\sigma }_{\sum I}$, and ${\sigma }_{\text{CC}}$ correspond to the uncertainties associated with lateral resolution (from sample shown Figure 1), tissue thresholding, signal thresholding for intensity summation, and the confidence curve at the median value, respectively. These individual error components are quantified as 1.8%, 6%, 0.89%, and 2.2%, cumulatively yielding a total error of 6.7%.

2.7 Application to Human FFPE Specimens

Having established a Ti calibration curve via LIBS quantitative analysis on animal FFPE lung tissues exposed to varying concentrations of TiO₂ P25 nanoparticles, we proceeded to adapt this novel methodological framework for quantifying elemental Ti content in autopsied human FFPE lung samples.

Figure 5a,b provide histological and phosphorus elemental images, respectively, to elucidate the tissue architecture of the three human lung specimens. Subsequently, the LIBS-derived elemental images for Ti are displayed in Figure 5c.

After employing the calibration curve illustrated in Figure 3b, the calculated yielded $S_{\text{value}}(\mathrm{Ti})$ corresponding Ti concentrations for these patients measured at 72.3, 15.7, and 1.03 μg g⁻¹. Notably, these observed concentrations align well with literature, wherein the typical range for human lung Ti content spans between 0.1 and 200 μg g⁻¹.^[26-28] Drawing upon the empirical data obtained from controlled inhalation exposure studies in animals (Figure 3b), the observed Ti concentrations in the three human lung specimens can be interpretatively aligned with distinct levels of exposure. Specifically, the elevated Ti content in Patient 1's lungs suggests a high level of exposure, while the moderate and minimal concentrations in Patients 2 and 3, respectively, could be indicative of medium and low exposure levels. The multielemental capability of LIBS imaging was also leveraged to generate elemental images for both endogenous elements, such as sodium and magnesium, and exogenous elements, including silicon and aluminum (Figure S1, Supporting Information). The LIBS-based analytical framework presented herein holds significant promise as a diagnostic tool for medical pathologists and clinicians. Specifically, it offers the capability to directly quantify concentrations of specific chemical elements—particularly metals—in lung tissue biopsies. This could serve as a valuable adjunct in assessing patient exposure levels to environmental or occupational hazards.