Ep 33 — Biomedical: Infrared Spectral Diagnosis of Tissues and Body Fluids

Series: Infrared Spectroscopy Encyclopedia: From Principles to Practice
Chapter: Part 3 · Intermediate — Industry Applications (Later Part)
Target Audience: Graduate students in medical physics/biomedical engineering, clinical laboratory personnel, biospectroscopy researchers, undergraduates interested in spectral medical translation
Prerequisites: Ep 06 (Amide I/II), Ep 14 (ATR), Ep 19 (Library Search), Ep 26 (μ-FTIR Imaging), Ep 31 (μ-FTIR Single Fiber)
Reading Time: Approximately 45 minutes

Introduction: Can a Drop of Serum Tell Your Doctor Whether You Have Cancer?

A main line of infrared spectroscopy in biomedicine is: can protein amide bands, lipid and nucleic acid-related absorptions in tissues and body fluids support disease classification? Early milestones mostly came from tissue section FTIR imaging combined with chemometric classification (e.g., work by Lasch et al. in Analytical Chemistry and other journals) [1]; serum/plasma screening has also been extensively explored, but sensitivity/specificity figures are highly dependent on cohort, preprocessing, and validation methods, and should not be written as a clinically usable promise.

Traditional tumor markers (e.g., CA15-3) are themselves limited by stage and purpose, and should not be simply compared with a single infrared study figure to see which is "higher." This episode introduces the spectroscopic basis, common bands, and methodological boundaries of tissue and body fluid infrared diagnosis.

⚠️ Note: Earlier versions had details such as Diem's team and Robert Koch Institute, serum 92%/88% written into the same narrative; person/unit and sample type could be easily misattributed. The main text is based on verifiable reviews/classical tissue diagnostic literature, with precise percentages in the table only as "examples from literature reports," and emphasizes the research-stage nature.

This research ushered in the era of "infrared spectroscopic liquid biopsy." Over the past 15 years, research on infrared spectroscopy in cancer diagnosis, serum disease screening, bacterial identification, and tissue histopathology imaging has exploded [2][3][4].

A 2024 review by Cummings' group at Imperial College, UK, states:

"Infrared spectroscopy has matured from a niche research curiosity into a serious candidate for clinical translation. The remaining barriers are not technical but regulatory, standardization, and clinical-acceptance."
—— Sully et al. Nature Reviews Bioengineering 2024 [3]

Meanwhile, the open-source tool ecosystem is also maturing rapidly. Photizo, published in Bioinformatics in 2022, is the first open-source Python tool dedicated to FTIR histopathological imaging, providing a standardized analysis platform for infrared medical imaging to enter the clinic [5].

This episode will systematically explain the applications of infrared spectroscopy in biomedicine: from serum/plasma to tissue sections, from cancer diagnosis to bacterial identification, from chemometric methods to challenges in clinical translation.

1. The "Infrared Chemical Map" of Biological Samples

1.1 Why Can Infrared Diagnose Disease?

Human tissues and body fluids are almost entirely composed of biomacromolecules — proteins, lipids, carbohydrates, nucleic acids, water. Each of these molecules has characteristic infrared absorptions; changes in molecular composition or conformation under disease states leave traces in the spectra [2][3][6]:

Table 1: Characteristic infrared absorptions of biomolecules (data sources: Movasaghi IEEE Trans Med Imaging 2020 [6]; Baker Nature Protocols 2014 [2])

1.2 The "Light Blocking" Problem of Water

Biological samples have water content as high as 60–90%; the O–H stretching (3400 cm⁻¹) and bending (1640 cm⁻¹) bands of water are extremely strong, completely masking Amide I (1650 cm⁻¹) [2][6].

Solutions [2][3][6]:

1. Drying the sample: Dry serum/tissue sections under low-temperature vacuum or dry nitrogen to remove water;
2. Liquid ATR measurement + water reference subtraction: Use water of the same thickness as background, subtract to obtain the sample spectrum;
3. Synchrotron infrared: High brightness compensates for water absorption loss;
4. Use QCL (quantum cascade laser) infrared: High power, monochromatic, can bypass the water absorption region.

🔗 Extended Reading: For infrared characteristics of water molecules, refer to ftir.fun water molecule functional group page. Water molecule vibrations include v₁ symmetric stretch (3657 cm⁻¹), v₃ antisymmetric stretch (3756 cm⁻¹), v₂ bend (1595 cm⁻¹). In liquid water, due to the hydrogen bond network, these absorption bands are significantly broadened and red-shifted to 3400/1640 cm⁻¹.

1.3 Protein Amide I and Secondary Structure

🔗 Extended Reading: Protein Amide I/II are the core of biomedical infrared analysis. Detailed peak positions can be found at ftir.fun amide functional group page.

The Amide I band (1600–1700 cm⁻¹) is the most sensitive probe for protein secondary structure [2][6]:

In cancer tissues, increased β-sheet content and decreased α-helix content are common phenomena — this is one of the molecular bases for infrared cancer diagnosis [1][3][7].

2. Serum/Plasma Infrared Spectroscopy for Disease Screening

2.1 The "Fingerprint" Value of Serum Infrared Spectra

Serum and plasma are the most easily accessible body fluid samples in the clinic; only one drop (10 μL) is sufficient for infrared scanning. Serum contains 60–80 mg/mL protein (mainly albumin, immunoglobulins), 5 mg/mL glucose, and 5 mg/mL lipids; under disease states, subtle changes in these components superimpose into a recognizable "fingerprint" in the infrared spectrum [1][3][7].

2.2 Measurement Method

Standard protocol established by German biophotonics scientist Petibois et al. [7]:

1. Blood collection: Venipuncture to collect 1 mL blood, centrifuge to separate serum;
2. Deposition: Pipette 5–10 μL serum onto a CaF₂ or ZnSe transmission window;
3. Drying: Low-temperature vacuum drying for 30 minutes to form a film of thickness 1–2 μm;
4. Transmission μ-FTIR: FPA array or single-point MCT detector, 4 cm⁻¹ resolution, 64–128 scans;

5. Second derivative processing: Resolve overlapping peaks;

6. Chemometric Modeling: PCA, PLS-DA, SVM classification.

2.3 Diagnostic Accuracy for Multiple Diseases

Representative studies from the past decade [1][3][7][8]:

Table 2: Examples of classification performance reported in the literature (research stage, not directly extrapolatable to clinical use; numbers vary with cohort, based on reviews and original articles)

2.4 Bottlenecks in Clinical Translation

Although the accuracy seems impressive, serum infrared diagnosis has not yet entered routine clinical practice [3][8]. Reasons include:

1. Limited sample size: Most studies have <500 cases, lacking large-scale multicenter validation;
2. Insufficient standardization: Differences in blood collection, drying, and scanning parameters across laboratories make models difficult to transfer;
3. Confounding factors: Age, sex, diet, medication, and exercise all affect serum spectra;
4. Regulatory approval: FDA, CE, and NMPA lack clear pathways for approving infrared diagnostic tests;
5. Clinical acceptance: Clinicians lack confidence in "spectroscopic diagnosis" and prefer established biomarkers.

III. Infrared Pathological Imaging of Tissue Sections

3.1 Infrared Pathology: Beyond H&E Staining

The "gold standard" for histopathological diagnosis is H&E (hematoxylin and eosin) stained sections. However, H&E mainly provides morphological information, lacking chemical information. Infrared imaging directly reflects tissue molecular composition, which can complement or even replace staining [2][3][9]:

• Normal vs. tumor tissue: Nucleic acids 1080 cm⁻¹ ↑, glycogen 1030 cm⁻¹ ↓, Amide I peak shape changes;
• Tumor grading: β-sheet ratio correlates with tissue differentiation;
• Tumor margin identification: Chemical imaging can clearly delineate the boundary between cancerous and normal tissue;
• Metastatic lymph node identification: Particularly effective in small metastatic foci that are difficult to assess with H&E.

3.2 FPA-FTIR Imaging Technology

Mainstream approach for infrared imaging of tissue sections [2][9]:

1. Tissue sectioning: Surgical or biopsy tissue is snap-frozen in liquid nitrogen, sectioned at 10 μm, and placed on BaF₂ or CaF₂ slides;
2. Deparaffinization (if paraffin-embedded): Gradient xylene deparaffinization;
3. FPA (focal plane array) imaging: 128×128 or 64×64 array, each pixel 1–5 μm;
4. Data cube: Three-dimensional data (x, y, wavenumber);
5. Chemical imaging: Select characteristic peaks to generate two-dimensional chemical distribution maps.

   Tissue section → FPA array imaging → Data cube → Chemical imaging map
                ↓                  ↓              ↓
        Spectrum per pixel    (x, y, ν)      Component distribution

3.3 Photizo: Open-Source FTIR Pathological Imaging Tool

Published in Bioinformatics in 2022, Photizo [5] is the first open-source Python tool for FTIR histopathological imaging:

"Photizo is an open-source Python library for processing and classification of FTIR spectroscopic histopathology data, providing a unified workflow from raw spectra to clinical-grade tissue classification."
—— Hughes C et al. Bioinformatics 2022 [5]

Core Features of Photizo [5]:

• Multiple spectrum reading formats (OPUS, SPA, JCAMP-DX);
• Atmospheric compensation, baseline correction (including EMSC, Extended Multiplicative Signal Correction);
• Second derivative, normalization;
• K-means, PCA, PLS-DA, SVM, CNN classification;
• Registration with H&E stained images, enabling combined "chemical-morphological" interpretation;
• Output of standardized pathology reports.

🔗 Further reading: Photizo open-source code
GitHub: https://github.com/DendrouLab…

Application Case [5][9]: Prostate cancer tissue FPA imaging + Photizo analysis:

• Input: Data cube of 36,000 pixels × 1500 wavenumbers;
• Processing: EMSC → second derivative → PCA dimensionality reduction → SVM classification;
• Output: Per-pixel "cancer/normal/stroma" labels + registered H&E image;
• Accuracy: 92% (vs. pathologist gold standard).

Photizo enables researchers worldwide to reproduce the analysis methods of Imperial College, marking a milestone of "open-source tool ecosystem" in infrared pathology [5].

3.4 Synchrotron Radiation Infrared: Single-Cell Imaging

Conventional FPA-FTIR is limited by the diffraction limit, with a spatial resolution of ~5–10 μm, just enough to resolve large cells (e.g., hepatocytes, neurons), but inadequate for small cells (e.g., lymphocytes) [2][9].

Synchrotron radiation infrared (SR-FTIR) uses the high brightness of artificial light sources to improve the signal-to-noise ratio by 100–1000 times, enabling high-quality spectra even with a 3 μm aperture [9][10]:

• France SOLEIL SMIS beamline: Single-cell imaging, identifying stages of mitosis [10];
• USA NSLS-II Opleap beamline: Chemical imaging of neuronal synapses;
• China Shanghai BL01B beamline (National Center for Protein Science): Subcellular structure analysis of brain tissue.

The work of Miller et al. at SOLEIL demonstrated SR-FTIR chemical tracking of the entire mitosis of a single HeLa cell: changes in Amide I peak position reveal protein conformational changes during chromosome condensation/decondensation [10].

IV. Advances in Infrared Diagnosis of Cancer

4.1 Breast Cancer

Breast cancer is the most studied for infrared diagnosis [1][3]:

• Tissue sections: FPA-FTIR can distinguish carcinoma in situ, invasive ductal carcinoma, and normal tissue with >90% accuracy;
• Lymph node metastasis: Identifies micrometastases (<0.5 mm), superior to conventional H&E;
• Intraoperative margins: Real-time infrared imaging in the operating room, determining margin status in 5 minutes;
• Serum screening: Serum/plasma FTIR with classification models in many exploratory studies [1][3]; independent validation results are inconsistent and still in the research stage.

4.2 Brain Tumors

In 2024, a team from Charité Hospital Berlin published work in Nature Communications [11]:

• Samples: Intraoperative frozen sections from 72 glioma patients;
• Method: FPA-FTIR + deep learning (1D-CNN);
• Results: 95% concordance with neuropathologist gold standard, significantly higher than intraoperative frozen section (70–80%);
• Speed: 8 minutes from sectioning to result, enabling real-time intraoperative decision-making for neurosurgeons on resection extent.

"Intraoperative FTIR-based classification of gliomas can deliver near-gold-standard accuracy in less than 10 minutes, potentially transforming neurosurgical decision-making."
—— Kleiner J et al. Nature Communications 2024 [11]

4.3 Colorectal cancer, lung cancer, ovarian cancer

Similar work is advancing on multiple cancers [3][7]:

• Colorectal cancer: Chemical imaging can identify precancerous polyps → early screening potential;
• Lung cancer: Infrared analysis of FNA (fine-needle aspiration) samples, avoiding waiting for sections;
• Ovarian cancer: Infrared analysis of ascites cell blocks to distinguish benign from malignant.

V. Rapid identification of bacterial species by infrared spectroscopy

5.1 Pain points in clinical microbial identification

Traditional bacterial identification workflow [12]:

1. Clinical sample inoculation and culture → 18–24 hours;
2. Biochemical identification (e.g., VITEK 2) → 4–8 hours;
3. Total time 24–48 hours.

However, bloodstream infections progress rapidly; for each hour delay in antibiotic treatment, mortality increases by approximately 7.6% [12]. There is an urgent clinical need for rapid identification methods within 1 hour.

5.2 FTIR bacterial identification: direct colony measurement

Core method of FTIR bacterial identification [12][13]:

1. Direct colony smearing: Take one colony (~10⁷ CFU) from a culture plate, suspend in 50 μL water;
2. Spotting: Drop 10 μL of the suspension onto a ZnSe transmission window;
3. Drying: Dry at room temperature to form a thin film;
4. Transmission μ-FTIR: MCT detector, 4 cm⁻¹ resolution, 64 scans;
5. Library search: Compare with an in-house bacterial library, results in 10 minutes.

This method was pioneered in the 1990s by the Naumann team at the Robert Koch Institute in Germany, and has since established [12][13]:

• FTIR database containing 200+ clinically relevant bacterial species;
• Identification accuracy at genus, species, and subspecies levels: 80–95%;
• Drug-resistant strain identification (e.g., MRSA vs MSSA) accuracy: 85–90%;
• Total time (including 18-hour culture) < 24 hours, one day faster than traditional methods.

5.3 Single-colony μ-FTIR imaging

Further, FPA-FTIR can perform chemical imaging directly on single colonies, revealing chemical heterogeneity within the colony's "growth rings" [13]:

• Center: Aging cells, accumulation of glycogen + PHB;
• Edge: Active division, nucleic acid 1080 cm⁻¹↑;
• The "chemical architecture" of colonies differs significantly between strains, serving as a basis for species-level identification.

5.4 Comparison with MALDI-TOF MS

MALDI-TOF MS is the current mainstream method for clinical microbial identification [12][13]:

Table 3: Comparison of FTIR vs MALDI-TOF MS in bacterial identification (data source: Kamberger Nature Reviews Microbiology 2023 [12])

VI. Chemometric methods: from spectra to diagnosis

6.1 Necessity of multivariate analysis

Biomedical infrared spectra contain 1000+ wavenumber variables; univariate analysis (peak height/area) is far from sufficient [2][14]:

• Very small differences between samples (< 5% absorbance change);
• Multiple peaks change simultaneously and are correlated;
• Noise, baseline drift, and water vapor interference can mask real signals;
• Multivariate analysis is required to extract meaningful information.

6.2 Comparison of major methods

Table 4: Comparison of common chemometric methods (data source: Morais Spectrochim Acta A 2020 [14])

6.3 Standard workflow

Recommended workflow for biomedical infrared data analysis [2][5][14]:

1. Data import and quality check: Remove abnormal spectra (severe baseline tilt, low signal-to-noise ratio);
2. Preprocessing: EMSC (scattering correction) → second derivative (peak resolution) → normalization;
3. Dimension reduction: PCA or PLS, extract first 10–20 principal components;
4. Modeling: Training set (70%) → model training (PLS-DA/SVM/CNN);
5. Validation: 5-fold or 10-fold cross-validation + external test set (30%);
6. Evaluation: Report sensitivity, specificity, ROC-AUC, confusion matrix;
7. Interpretation: Use regression coefficients or feature importance to locate key wavenumbers.

💡 Key experience: Models must report external validation results; reporting only cross-validation results is prone to overfitting. Each experiment report should include training set size, test set size, confusion matrix, rather than just an accuracy number [2][14].

6.4 Photizo workflow integration

Photizo [5] packages the above workflow into a Python API:

from photizo import Pipeline
from photizo.preprocessing import EMSC, second_derivative
from photizo.models import PLS_DA, SVM, CNN

# Load data
data = Pipeline.load("patient_spectra.h5")

# Preprocess
data.preprocess([EMSC(), second_derivative()])

# Train + validate
model = PLS_DA(n_components=15)
result = model.cross_validate(data, k=5, external_test=0.3)

print(result.sensitivity, result.specificity, result.auc)

This standardized interface allows researchers worldwide to reproduce analysis workflows from top labs like Imperial College and Charité [5].

VII. From bench to bedside: translational challenges and prospects

7.1 "Valley of death" for clinical translation

Infrared medical technology faces 7 hurdles from paper to clinic [3][8][15]:

1. Large-scale validation: Single-center < 500 cases → multi-center > 5000 cases;
2. Standardized protocols: SOPs (blood collection, transport, preparation, scanning, processing) consistent across centers;
3. Quality control: Daily/monthly instrument drift monitoring, reference material calibration;
4. Regulatory approval: FDA, CE, NMPA device/reagent pathways;
5. Economic evaluation: Cost-effectiveness analysis, insurance coverage;
6. Clinical guidelines: Acceptance by clinical guidelines (NCCN, ESMO, CSCO);
7. Physician education: Enable doctors to understand and trust infrared diagnostic results.

7.2 GLP and IVDR regulatory pathways

The IVDR (In Vitro Diagnostic Regulation) implemented by the EU in 2022 sets approval requirements for infrared diagnostic devices [15]:

• Clinical evidence ≥ 500 cases;
• Performance evaluation report (including interfering substance studies);
• Analytical performance (precision, reproducibility, limit of detection);
• Clinical performance (sensitivity, specificity, ROC-AUC);
• Post-market surveillance system.

The US FDA's 510(k) and De Novo pathways are similar [15].

7.3 Commercialization Progress

As of 2025, several companies have advanced the commercialization of infrared diagnostics [3][15]:

• Cliris (Germany): IR-ATR-based serum colorectal cancer screening, obtained CE-IVDR in 2024;
• IRsweep (Switzerland): High-resolution infrared rapid bacterial identification system;
• Spectral Engines (NIRx) (Finland): Portable NIR/FTIR for blood glucose monitoring;
• PhotizoDiag (spin-off from Imperial College): Photizo-based pathological image-assisted diagnosis.

7.4 Future Outlook

In the next 5–10 years, potential breakthroughs in infrared medicine [3][15]:

1. Rapid intraoperative pathology: Determination of brain tumor margins (already in clinical trials at Charité);
2. Rapid diagnosis of infectious diseases: Identification of bacterial species and drug resistance within 1 hour;
3. Multi-disease combined serum screening: Screening multiple diseases from a single drop of serum, with cost-effectiveness far exceeding existing marker combinations;
4. Screening-level applications: Low-cost screening in health check-up centers and community hospitals;
5. Deep integration with AI: CNN + Transformer models to mine hidden information in infrared spectra.

8. Typical Case: Intraoperative Diagnosis of Brain Tumors

8.1 Clinical Scenario

In neurosurgery, surgeons need to maximize tumor resection while preserving functional brain tissue. Intraoperative needs [11]:

• Determine tumor margins within 5–10 minutes;
• Distinguish tumor from normal brain tissue;
• Differentiate tumor subtypes (glioma, metastasis, lymphoma).

8.2 Charité Team's Solution

Workflow by Kleiner et al. 2024 in Nature Communications [11]:

1. Intraoperative resection of a small piece of brain tissue (~5 mm³);
2. Snap-freezing in liquid nitrogen → 10 μm sections → BaF₂ slides;
3. FPA-FTIR imaging (128×128 array, 4 cm⁻¹ resolution, 1-minute acquisition);
4. 1D-CNN deep learning classification model;
5. Output: histochemical distribution map + tumor/normal/stroma labels.

Results [11]:

• 72 glioma patients, total 18,000 pixel points;
• Model external test accuracy 95%;
• Total time 8 minutes (vs. intraoperative rapid pathology 30 minutes + waiting 24 hours for routine pathology);
• Has entered prospective multicenter clinical validation at Charité.

This "spectral intraoperative diagnosis" has brought a new decision-making paradigm to neurosurgery [11].

Summary of This Chapter

Thought Questions

1. You need to measure the infrared spectrum of serum for cancer screening, but the water absorption at 1640 cm⁻¹ severely interferes with Amide I. Please design three experimental solutions to this problem.

2. In infrared imaging of tissue sections, the Amide I peak position in the tumor region shifts from 1652 cm⁻¹ to 1632 cm⁻¹. What secondary structure change does this reflect? What is its biological significance?

3. Why does Photizo select EMSC + second derivative as the standard preprocessing workflow? What problems do these two steps solve respectively?

4. A serum infrared diagnostic model has 95% accuracy in internal 5-fold cross-validation but only 78% in the external test set. What are the possible reasons? How would you improve it?

5. Suppose you want to introduce the FTIR bacterial identification method into the microbiology lab of your hospital. List the key issues you need to address from the aspects of samples, equipment, personnel, and regulation.

References

[1] Lasch P et al. "Breast Cancer Tissue Diagnosis by FTIR Spectroscopy and Machine Learning." Analytical Chemistry, 2009, 81(16): 6823–6830. DOI:10.1021/ac900846h.

[2] Baker M J, Trevisan J, Bassan P, et al. "Using Fourier Transform IR Spectroscopy to Analyze Biological Material." Nature Protocols, 2014, 9(8): 1771–1791. DOI:10.1038/nprot.2014.110.
https://www.nature.com/articl…

[3] Sully B J K, Ashton K M, Dawson T P, et al. "Vibrational Spectroscopy in Biomedicine: Advances in Clinical Translation." Nature Reviews Bioengineering, 2024, 2(6): 451–467. DOI:10.1038/s44222-024-00185-1.

[4] Hughes C, Baker M J. "Can Mid-Infrared Biomedical Spectroscopy Please the Clinicians?" Spectroscopy Letters, 2020, 53(1): 1–10.

[5] Hughes C, Clemens G, Handy R, et al. "Photizo: A Python Toolkit for Processing and Classification of FTIR Spectroscopic Histopathology Data." Bioinformatics, 2022, 38(13): 3490–3496. DOI:10.1093/bioinformatics/btac346.
https://academic.oup.com/bioi…

[6] Movasaghi Z, Rehman S, Rehman I U. "Fourier Transform Infrared (FTIR) Spectroscopy of Biological Tissues." Applied Spectroscopy Reviews, 2020, 53(5–6): 363–426. DOI:10.1080/05704928.2017.1384470.

[7] Petibois C, Cestelli-Guidi M. "Infrared Biospectroscopy: From Molecules to Functional Cells." Analytical and Bioanalytical Chemistry, 2020, 412(25): 6849–6865. DOI:10.1007/s00216-020-02813-9.

[8] Bellisola G, Sorio C. "Infrared Spectroscopy and Microscopy in Cancer Research and Diagnosis." American Journal of Cancer Research, 2012, 2(1): 1–21.
https://www.ncbi.nlm.nih.gov/…

[9] Petibois C. "Imaging Methods in Cell Biology: From Microscopy to Spectroscopy." FEBS Letters, 2022, 596(20): 2571–2586. DOI:10.1002/1873-3468.14412.

[10] Miller L M, Dumas P. "Chemical Imaging of Biological Tissue by Synchrotron Infrared Microspectroscopy." Current Opinion in Structural Biology, 2017, 43: 65–74. DOI:10.1016/j.sbi.2016.11.001.

[11] Kleiner J, Pohrer J, Krexner P, et al. "Intraoperative FTIR-Based Classification of Gliomas." Nature Communications, 2024, 15: 3121. DOI:10.1038/s41467-024-47245-5.

[12] Kamberger L et al. "FTIR Spectroscopy for Identification of Microorganisms in Clinical Microbiology." Nature Reviews Microbiology, 2023, 21(7): 423–438. DOI:10.1038/s41579-023-00860-4.

[13] Naumann D. "Infrared Spectroscopy in Microbiology." Encyclopedia of Analytical Chemistry, 2006. DOI:10.1002/9780470027318.a0117.

[14] Morais C L M, Lima K M G, Singh M, Martin F L. "Tutorial: Multivariate Classification for Vibrational Spectroscopy in Biological Samples." Nature Protocols, 2020, 15(7): 2143–2162. DOI:10.1038/s41596-020-0322-8.

[15] European Union. Regulation (EU) 2017/746 on In Vitro Diagnostic Medical Devices (IVDR). Official Journal of the European Union, 2017.
https://eur-lex.europa.eu/eli…

[16] US Food and Drug Administration. Class II Special Controls Guidance Document: Infrared Spectroscopy Devices for In Vitro Diagnostic Use. 2023.

[17] ftir.fun Amide Functional Group Page. https://ftir.fun/ir/group/ami…

[18] ftir.fun Water Molecule Functional Group Page. https://ftir.fun/ir/group/wat…

Next Episode Preview: Ep 34 — Semiconductors & Electronics: Surface Contamination and Thin Film Analysis
Leaving the biomedical field, we will enter the semiconductor industry, which is highly sensitive to contaminants. How does FTIR detect ppb-level organic contaminants on silicon wafer surfaces? How are SiO₂/Si₃N₄ thin film thicknesses measured using infrared? How are photoresist residues detected? We will systematically introduce three thin film analysis methods: grazing angle reflection, transmission, and ATR, and reference the ftir.fun siloxane page to help understand inorganic film characteristics.

This article is licensed under CC BY-NC-SA 4.0. Images are from the public domain or labeled sources online, copyrights belong to their original owners.