Comparison of Mid-Infrared Handheld and Benchtop Spectrometers to Detect Staphylococcus epidermidis in Bone Grafts

Bone analyses using mid-infrared spectroscopy are gaining popularity, especially with handheld spectrometers that enable on-site testing as long as the data quality meets standards. In order to diagnose Staphylococcus epidermidis in human bone grafts, this study was carried out to compare the effectiveness of the Agilent 4300 Handheld Fourier-transform infrared with the Perkin Elmer Spectrum 100 attenuated-total-reflectance infrared spectroscopy benchtop instrument. The study analyzed 40 non-infected and 10 infected human bone samples with Staphylococcus epidermidis, collecting reflectance data between 650 cm−1 and 4000 cm−1, with a spectral resolution of 2 cm−1 (Agilent 4300 Handheld) and 0.5 cm−1 (Perkin Elmer Spectrum 100). The acquired spectral information was used for spectral and unsupervised classification, such as a principal component analysis. Both methods yielded significant results when using the recommended settings and data analysis strategies, detecting a loss in bone quality due to the infection. MIR spectroscopy provides a valuable diagnostic tool when there is a tissue shortage and time is of the essence. However, it is essential to conduct further research with larger sample sizes to verify its pros and cons thoroughly.


Introduction
Bone represents the second most commonly transplanted tissue behind blood [1]. In orthopedic surgery, the application of human bone allografts is prevalent in promoting spinal fusion and reconstructing bone defects resulting from trauma, tumors, or revision arthroplasty [2][3][4][5][6][7][8]. Removing loose implants and adjacent fibro cellular membranes is often necessary in the context of hip arthroplasty revision procedures. To address any resulting bone loss, particulate bone grafts are commonly employed as a compensatory measure. In order to induce bone remodeling and prevent early implant subsidence, the morselized allograft must be compacted [6]. Upon attaining initial stability via impaction procedures, the graft may be assimilated into the host skeleton through revascularization [9][10][11][12]. According to soil mechanics, firmly compacted aggregates should have well-graded particle confidently identify and select only the highest-quality bone grafts through handheld MIR spectroscopy. This tool provides crucial information about mineralization processes and can efficiently detect contamination [42,48,49]. In the last ten years, several studies have highlighted the usefulness of MIR spectroscopy for detecting different bone diseases in humans [34][35][36][37][38][39][40][41], including the evaluation of bone quality [43]. Bone comprises a variety of MIR bands, typically composed of phosphate (ν 3 PO 4 3− ), carbonate (ν 1 CO 3 2− ), collagen matrix, amide III, CH 2 of protein, and amide I [41,[50][51][52][53]. Staphylococcus epidermidis is the most prevalent pathogen in bone and implant-related infections. Therefore, this study aimed to compare the efficacy of the small handheld instrument with a benchtop mid-infrared spectrometer in detecting Staphylococcus epidermidis on bone grafts. The small Agilent 4300 Handheld Fourier-transform infrared (FTIR) scanner and the larger Perkin Elmer Spectrum 100 attenuated-total-reflectance infrared spectroscopy (ATR-IR) benchtop instrument were used, and a principal component analysis was conducted on fresh frozen bone samples. This study compares mid-infrared handheld and benchtop spectrometers to detect Staphylococcus epidermidis in bone grafts. The outcome of this study provides new insights into the potential of small MIR handheld spectrometers in identifying these pathogens in human bone grafts.

Sample Collection
The femoral heads used in our bone bank were sourced from individuals who had undergone hip replacement surgery due to advanced hip osteoarthritis or femoral neck fracture. Before their donation, all patients provided written informed consent, ensuring that their contribution was made willingly and with a full understanding of the process. The donated bone that did not fulfil the criteria for therapeutic use (e.g., due to incomplete screening and documentation) was kept at the bone bank and utilized in scientific studies. In general, severe osteoporosis and contaminated samples caused by various pathogens were eliminated and not collected, regardless of the research. There were no criteria based on age or gender. The bone was drained and chilled with 0.9% saline during osteotomy to prevent heat damage. Cartilage and cortical tissues from the femoral heads were removed using a bone saw. Bone chips 3-5 mm in diameter were extracted from the residual spongious tissue (Noviumagus Bone Mill; Spierings Meische Techniek BV, Nijmegen, The Netherlands) with a bone mill. A total of 40 human bone samples were examined. The local ethics council approved the retrospective study (EK 1291/2021) per the guiding principles outlined in the Declaration of Helsinki.

Development of Biofilm on Bone Allografts
To examine Staphylococcus epidermidis ATCC 12228, Mueller-Hinton broth with confidence and precision was utilized. We carefully incubated the broth at 37 • C for 24 h. The inoculum was diluted to 106 CFU/mL, and 200 µL of the suspension was added to each well of a multi-well plate. Individual fresh frozen bone allografts and a substrate were inserted to create biofilms. The plates were then placed in an orbital shaker and incubated at 37 • C for 48 h to form the biofilms. Once completed, the supernatant was removed, and the bone samples were washed with new PBS to ensure the removal of any planktonic bacteria. After this procedure, the bone samples were dried in an aspirator (3.2 kPa) for 10 min at room temperature and measured. The drying time was sufficient, as prolonging the drying time to 24 h caused no differences in spectra quality.

Benchtop Perkin Elmer Spectrum 100 ATR-IR Spectrometer
The MIR ATR-IR spectra were collected using a Perkin Elmer Spectrum 100 ATR-IR spectrometer (Perkin Elmer, Waltham, MA, USA). There were eight scans per sample from three positions, with a resolution of 0.5 cm −1 and a wavenumber range of 4000 to 650 cm −1 . The measurement was conducted at a temperature of 22 • C under controlled humidity levels.

Agilent 4300 Handheld FTIR
MIR spectra were obtained using the Agilent 4300 Handheld FTIR (Agilent Technologies, Santa Clara, CA, USA) device. The spectral range covered was from 4000 to 650 cm −1 , with a spectral resolution of 2 cm −1 . Each sample was subjected to eight spectra recorded from three different positions. The measurement was conducted at a temperature of 22 • C under controlled humidity levels.
Spectral Parameters: The diagnostic parameters were studied using peak intensity (I) [54][55][56][57][58][59]. The determination of intensities was made effortless through an Excel spreadsheet that handles spectroscopic and chromatographic data [60]. A statistical analysis of the spectral parameters was conducted using GraphPad Prism software (version 9, San Diego, CA, USA) and compared through a two-sample t-test. A significant result is only considered if the p-value is less than 0.05.

Principal Component Analyses (PCA)
PCA models were created using Unscrambler X 10.5, which involved importing spectra into the program and applying various data pretreatments, such as a reduction factor 36, Savitzky-Golay smoothing with 15 smoothing points, and area normalization.

Results
The main objective of this study was to compare the effectiveness of MIR handheld and benchtop spectrometers in detecting Staphylococcus epidermidis in bone grafts. Of 40 noninfected human bone samples, 10 were intentionally infected with Staphylococcus epidermidis ATCC 12228. Twenty-two females and eighteen males provided non-infected human bones (n = 40), while the infectious human bones were collected from seven females and three males (n = 70/30%). The sample characteristics are summarized in Table 1. Figure 1 shows the advantages and disadvantages of conventional infection diagnosis, MIR handheld, and benchtop spectrometry. Compared to MIR, the traditional diagnosis of infection can be time-consuming, resource-intensive, and labor-intensive.
0.0004626 ± 7.614 × 10 to 0.0007611 ± 5.644 × 10 . Notably, the positions and shape of the v3PO4 3− band from 1200 to 900 cm −1 , the amide I band from 1730 to 1585 cm −1 , and the bending and stretching modes of C-H groups from ~3000 to 2800 cm −1 coincided nicely. However, there were some differences in the Agilent 4300 Handheld, particularly in the shape of the feature between 650 and 1800 cm −1 (i.e., v3CO3 2− and amide III), and the amide II feature at around 1650 cm −1 was not displayed in the Agilent 4300 Handheld as it was in the Perkin Elmer Spectrum 100 instrument.  Figure 3 clearly compares the average spectra of non-infected and infected human bone with Staphylococcus epidermidis. This was achieved through measurements taken using the Agilent 4300 Handheld and Perkin Elmer Spectrum 100 instruments. The average spectra were derived from eight spectra by using the Perkin Elmer Spectrum 100 and four by using the Agilent 4300 Handheld. Notably, infected and non-infected human bones exhibit distinct spectral features, including the v3PO4 3− band, the amide I band, and C-H groups from ~3000 to 2800 cm −1 . However, significant differences between infected and non-infected human bones are present in the fingerprint region from 1800 to 650 cm −1 , the v3PO4 3− band, and the amide I band. Specifically, the Agilent 4300 Handheld instrument demonstrates a substantial loss of the amide I band in infected human bones.  Figure 3 clearly compares the average spectra of non-infected and infected human bone with Staphylococcus epidermidis. This was achieved through measurements taken using the Agilent 4300 Handheld and Perkin Elmer Spectrum 100 instruments. The average spectra were derived from eight spectra by using the Perkin Elmer Spectrum 100 and four by using the Agilent 4300 Handheld. Notably, infected and non-infected human bones exhibit distinct spectral features, including the v3PO 4 3− band, the amide I band, and C-H groups from~3000 to 2800 cm −1 . However, significant differences between infected and non-infected human bones are present in the fingerprint region from 1800 to 650 cm −1 , the v 3 PO 4 3− band, and the amide I band. Specifically, the Agilent 4300 Handheld instrument demonstrates a substantial loss of the amide I band in infected human bones. Table 2 concisely overviews the essential parameters scrutinized in MIR spectra. These parameters encompass the intensities (I) of the most influential bands, culminating in single-band and multiple-band ratios [54,55,[57][58][59]. They serve as markers that furnish precise diagnostic data on infections that impact human bones, as depicted in Figure 4.
The results from a two-sample t-test, displayed in Figure 4, indicate a significant difference in the levels of phosphate, amide I, and CH-Aliphatic content (CHACont) between infected and non-infected samples, using the Agilent 4300 Handheld instrument. The Perkin Elmer Spectrum 100 instrument revealed a significant difference in mineral/matrix (MMR), mineral quality and crystallinity, and mineral carbonate content (MinCarb) between infected and non-infected cases. Additionally, bone quality and strength can be assessed through the mineral-to-matrix ratio (MMR), which measures mineralization levels [41,42,69,70]. Calculating the ratio of mineral-specific MIR band intensities (phosphate and carbonate bands) to the intensity of the amide I band or the ratio of phosphate band intensity to the total intensity of proline and hydroxyproline MIR bands is crucial in determining changes in bone strength. It is important to note that bacterial infection can result in a higher loss of relative mineral content in bones. This makes weaker bones more susceptible to fractures than non-infected human bones [71,72]. The ratio of carbonate to phosphate intensities in MMR indicates certain properties.  Table 2 concisely overviews the essential parameters scrutinized in MIR spectra. These parameters encompass the intensities (I) of the most influential bands, culminating in single-band and multiple-band ratios [54,55,[57][58][59]. They serve as markers that furnish precise diagnostic data on infections that impact human bones, as depicted in Figure 4.   Additionally, the amide I band can help analyze changes in the collagen network caused by infection, as it is a typical protein conformation indicator due to its role in crosslinking and bonding. This band serves as an indicator of the protein structure [42]. Figure 4 clearly demonstrates that bacterial infection in human bones decreases structural organization and relative collagen. To measure the organic and inorganic components within the bone, the CH-aliphatic content (CHACont) is measured, which is typically attributed to the presence of proteins and lipids [73]. The study results reveal that bacterial infection causes a significantly higher loss of CHACont in bones. Co-culturing Staphylococcus epidermidis with human bone samples results in a significant deterioration of both bone quality and protein conformation. PCA was conducted to thoroughly describe the full range of spectral variations, as it is impossible to identify exact correlations with this type of processing.

Diagnostic Performance PCA
With the aim of facilitating a prompt diagnosis of the causative agent of infection, molecular techniques are commonly used in routine diagnostics. Early detection of infections is critical for guiding treatment decisions and improving patient outcomes [74]. The study successfully utilized the Agilent 4300 Handheld and Perkin Elmer Spectrum 100 instruments in the MIR with PCA analysis to expertly distinguish between bone graft samples infected with Staphylococcus epidermidis and those not infected. The potential diagnostic utility of PCA analysis of spectroscopic data has been demonstrated in previous studies [42,[75][76][77][78][79][80][81][82][83]. PCA was utilized on the averaged spectra of Agilent 4300 Handheld and Perkin Elmer Spectrum 100 instruments to conduct the analysis. Our study involved examining 40 non-infected and 10 infected bone samples, following the methodology outlined in previous studies [42,83,84]. Figure 5 presents the results of the spectral analyses conducted using PCA. It compares non-infected and infected human bone samples through a score plot of the first and second principal components. Table 3 provides a detailed analysis of five wavenumber ranges used in PCA. . MIR-derived Agilent 4300 Handheld and Perkin Elmer Spectrum 100 instrument spectral markers. * Significant; ** high significant; **** highly significant difference between means; ns, not significant.

Diagnostic Performance PCA
With the aim of facilitating a prompt diagnosis of the causative agent of infection, molecular techniques are commonly used in routine diagnostics. Early detection of infections is critical for guiding treatment decisions and improving patient outcomes [74]. The study successfully utilized the Agilent 4300 Handheld and Perkin Elmer Spectrum 100 instruments in the MIR with PCA analysis to expertly distinguish between bone graft samples infected with Staphylococcus epidermidis and those not infected. The potential diagnostic utility of PCA analysis of spectroscopic data has been demonstrated in previous studies [42,[75][76][77][78][79][80][81][82][83]. PCA was utilized on the averaged spectra of Agilent 4300 Handheld and Perkin Elmer Spectrum 100 instruments to conduct the analysis. Our study involved examining 40 non-infected and 10 infected bone samples, following the methodology outlined in previous studies [42,83,84]. Figure 5 presents the results of the spectral analyses conducted using PCA. It compares non-infected and infected human bone samples through a score plot of the first and second principal components. Table 3 provides a detailed analysis of five wavenumber ranges used in PCA.
Handheld in the spectral regions of amide III, amide I, and bending and stretching modes of C-H groups, as depicted in Figure 5 II, IV, and V Both methods are trustworthy and robust and can be enhanced to discriminate Staphylococcus epidermidis, Staphylococcus aureus, and other bacteria. PCA analysis is automated and objective, making it a beneficial tool for the routine laboratory screening of bone samples for bacterial infections. This method can significantly improve diagnostic performance for laboratories not specifically detecting bacterial bone infections.   According to the PCA models, the most informative data are located in regions II, IV, and V, as displayed in Figure 5. The score plots demonstrate the correlation between PC1 and PC2 for non-infected and infected bone samples within the II, IV, and V ranges, along with the corresponding loadings of PC1. The red symbols signify non-infected samples, while the blue symbols represent infected samples. PC1 accounts for 95%, 100%, and 91% for Perkin Elmer Spectrum 100 instrument and 97%, 98%, and 91% for Agilent 4300 Handheld in the spectral regions of amide III, amide I, and bending and stretching modes of C-H groups, as depicted in Figure 5 II, IV, and V Both methods are trustworthy and robust and can be enhanced to discriminate Staphylococcus epidermidis, Staphylococcus aureus, and other bacteria. PCA analysis is automated and objective, making it a beneficial tool for the routine laboratory screening of bone samples for bacterial infections. This method can significantly improve diagnostic performance for laboratories not specifically detecting bacterial bone infections.

Discussion
Both the Agilent 4300 Handheld instrument and the Perkin Elmer Spectrum 100 are trustworthy and robust and can be enhanced to discriminate Staphylococcus epidermidis, Staphylococcus aureus, and other bacteria. However, different wavelengths are responsible for the comparable outcome of the two instruments: For non-infected bones, the Agilent 4300 Handheld and the Perkin Elmer Spectrum 100 instrument equally detect phosphate, amide I (region IV), and stretching modes of CH groups (region V) with different intensities, especially for phosphate ( Figure 2). The Agilent 4300 Handheld, however, has a lower resolution for ν 3 CO 3 2− than the Perkin Elmer Spectrum 100 instrument. Thus, bone quality and strength cannot be assessed equally through the mineral-to-matrix ratio (MMR), which measures mineralization levels [41,42,69,70]. Additionally, mineral quality, crystallinity, and mineral carbonate content (MinCarb) cannot be assessed equally. Concerning bones infected with Staphylococcus epidermidis, the Agilent 4300 Handheld instrument especially differentiates the levels of phosphate, amide I (region IV), and CH-Aliphatic content (CHACont, region V), whereas the Perkin Elmer Spectrum 100 instrument revealed a significant difference in amide I (region IV), mineral/matrix (MMR), mineral quality and crystallinity, and mineral carbonate content (MinCarb) between infected and non-infected bones. In summary, according to the PCA models, the most informative data are located in regions II, IV, and V ( Figure 5). This shows that the Agilent 4300 Handheld instrument does not perform well between 650 and 1800 cm −1 , whereas the result of amide I (region IV) especially shows comparable data concerning the spectral analysis and PCA. The correlation between PC1 and PC2 was examined for non-infected and infected bone samples in the II, IV, and V ranges. The loadings of PC1 were also analyzed, with PC1 accounting for 97%, 98%, and 91% for the Agilent 4300 Handheld and 95%, 100%, and 91% for the Perkin Elmer Spectrum 100 instrument in the spectral regions of II (amide III), IV (amide I), and V (bending and stretching modes of C-H groups) (as shown in Figure 5).
Further investigations utilizing the carbonate/phosphate ratio reveal that human bone samples co-cultured with Staphylococcus epidermidis exhibit a significant reduction in bone quality and protein conformation. Infected bones, in particular, demonstrate a more pronounced decrease in relative mineral content compared to non-infected bones. Additionally, changes in the collagen network can be identified through the amide I band. PCA allows us to detect Staphylococcus epidermidis in multiple spectral regions, primarily from amide III, amide I, and C-H groups' bending and stretching modes, thereby validating its presence. These results using MIR spectroscopy are in line with previously published Raman investigations [41,42]. Such a direct detection of bone infection via MIR or Raman spectroscopy is challenging due to the complexity of bone tissue, low pathogen concentrations, and spectral overlaps. Nevertheless, with the integration of complementary techniques and careful optimization of sample preparation and data analysis, there is potential for MIR spectroscopy to contribute to the laser-independent detection and characterization of bone infections.
MIR spectroscopy is a speedy, robust, and automatable technique with significant advantages. MIR spectroscopy also requires a tiny untreated bone sample, making it suitable for situations with limited bone availability. Finally, this technique may benefit patients requiring urgent treatment because the analysis results are obtained without delay. In contrast, intraoperative tissue cultures, which are the current gold standard in diagnosing periprosthetic joint infections, are resource-intensive, and results can be expected only 5 to 11 days after tissue samples have been intraoperatively obtained [85]. Similarly, a histopathological workup of tissue samples requires significant resources and time, while sensitivity and specificity are even lower [86,87]. Another issue of tissue cultures is that inadequate processing and transport to the laboratory may result in either false-positive results due to contamination or culture-negative infections [88]. Overall, the sensitivity and specificity of intraoperative tissue cultures to detect periprosthetic joint infection have been reported to range from 0.51 to 0.90 and from 0.67 to 1.00, respectively, and diagnostic accuracy may be increased by additional sonication fluid cultures from explanted prosthetic components [89]. Similar values have been reported for spinal implant infections [90]. However, antibiotic therapy before culture sampling further decreases the sensitivity of intraoperative tissue cultures and results in culture negativity in a significant portion of periprosthetic joint infections [91]. MIR spectroscopy not only appears to help overcome the suboptimal results and well-known limitations of current infection diagnosis but may also contribute to further reducing the risk of transmitting bacterial contamination by bone allografts.
Our research has several limitations, including the small sample size and the prespecified incubation time of 48 h for bone fragments to promote the bacterial growth of Staphylococcus epidemridis. Therefore, the pathogen's limit of detection (LOD) could not be determined. Further studies to achieve calibration standards with known pathogen concentrations assisting in LOD determination are warranted. Our findings highlight the potential of spectroscopic analyses to link molecular changes with pathological conditions. Based on these findings, we anticipate that handheld and benchtop MIR spectroscopy have a different potential and will be used to detect bacterial infections in human bone samples either in biobanks or under surgical conditions, thus adding rapid information to other microbial diagnostic procedures. Further research with larger sample sizes and various incubation times are required to control for potential confounding variables and to validate this technique as a new diagnostic tool for the clinical handling of bones, combining MIR data and machine-learning analysis.

Conclusions
These findings highlight the comparable potential of the Agilent 4300 Handheld and Perkin Elmer Spectrum 100 for detecting infections in bone grafts. The data are consistent with previous research that suggests MIR spectrometry's effectiveness in identifying bacterial infections and extend the work now even for handheld instruments.
The Agilent 4300 Handheld and Perkin Elmer Spectrum 100 Benchtop spectrometers have unique strengths and limitations. The handheld spectrometer is beneficial for on-site and immediate clinical use, while the benchtop spectrometer is more suitable for research and complex analytical tasks in a laboratory setting. Thus, the choice between the two depends on specific requirements, portability needs and the budget as pros of the Agilent 4300 Handheld instrument, and the resources needed as con for the Perkin Elmer Spectrum 100 instrument. Using a handheld instrument has broad implications for the medical community, as such instruments allow MIR spectrometry to effectively and efficiently detect infections in bone grafts, potentially reducing healthcare costs and improving patient outcomes. Researchers and analysts should carefully consider these factors when selecting the most appropriate spectrometer for bacterial detection in bone grafts.
In summary, Agilent 4300 Handheld and Perkin Elmer Spectrum 100 instruments are both promising tools to diagnose human bone infections in situations with limited tissue availability and high urgency for treatment. However, further assessment is necessary to further validate the technique's benefits and drawbacks with a larger sample size and for more microbes and fungi.

Institutional Review Board Statement:
The study was conducted in accordance with the Declaration of Helsinki and approved by the Ethics Committee of the Medical University of Innsbruck (EK 1291/2021).

Informed Consent Statement: Not applicable.
Data Availability Statement: The data presented in this study are available upon request from the corresponding author.