v
Search
Advanced

Publications > Journals > Cancer Screening and Prevention> Article Full Text

  • Original Article
  • OPEN ACCESS

Diagnostic Performance of ATR-FTIR Spectroscopy in Discriminating Normal Breast Tissue and Breast Tumors

  • Samuel T. Adeleke1,* ,
  • Christopher Igbeneghu2 and
  • Sina Iyiola3
 Author information 

Abstract

Background and objectives

Histopathology is the gold standard in cancer diagnosis. However, attenuated total reflectance (ATR)-Fourier transform infrared (FTIR) spectroscopy has shown diagnostic potential in other settings. Therefore, this study aimed to evaluate the discriminatory performance of ATR-FTIR spectroscopy in differentiating histologically classified normal, benign, and malignant breast tissues.

Methods

This study was conducted on formalin-fixed, paraffin-embedded biopsy blocks received at Ladoke Akintola University of Technology Teaching Hospital between 2022 and 2023. The analyzed tissue set comprised 10 histologically normal breast tissue samples, 15 benign lesions, and 31 malignant specimens. Tissue sections of 15 µm were obtained during block trimming and transferred onto aluminum foil substrates. An additional 4 µm tissue section was stained with hematoxylin and eosin for tumor diagnosis and to identify suitable areas on the substrate. Spectrometer readings were taken within the range of 4,000–600 cm−1, 32 scans, and 16 cm−1 resolution, using the average of 10 preprocessed spectra per specimen. Biomarkers were calculated as peak-intensity ratios for A1632/A1543, A2922/A1632, A1632/A1080, A1080/A1543, A1237/A1080, and A1043/A1543, which represent protein, diagnostic marker, cytoplasm-nucleus ratio, carcinogenesis marker, phosphate, and glycogen, respectively. The receiver operating characteristic curve was used to determine sensitivity, specificity, and the area under the curve (AUC).

Results

The cytoplasm-nucleus ratio showed the highest reported AUCs in comparisons of normal, fibroadenoma, and fibrocystic tissues with malignant tissue (AUCs = 0.990, 0.935, and 0.976, respectively). Sensitivity for the cytoplasm-nucleus ratio was 100%, with specificity ranging from 50% to 100%. Glycogen showed an AUC of 1.000 with 100% sensitivity and specificity for differentiating fibroadenoma from fibrocystic changes.

Conclusions

ATR-FTIR spectroscopy showed promising discriminatory performance in differentiating normal, benign, and malignant breast tissues using specific spectral biomarkers. The cytoplasm-nucleus ratio marker showed potential as a spectral indicator for distinguishing the examined breast tissue categories.

Keywords

Terminal duct-lobular unit, ATR-FTIR spectroscopy, Fibroadenoma, Fibrocystic, Sensitivity, Specificity, Biomarkers, Area under the curve, Peak ratio, Point mapping, Performance, Invasive ductal carcinoma

Introduction

The breast holds significant importance in the reproductive aspects of women, while serving a more limited role in men. Its structure primarily consists of epithelium and stroma, with the terminal duct-lobular unit (TDLU) functioning as the essential component of the epithelial structure. The TDLU facilitates both secretory and collection functions during lactation.1 This unit is critical in reactive adaptations to physiological demands; however, most pathological lesions originate from epithelial cells in the TDLU area.2 Lesions are typically classified into benign and malignant categories, with benign conditions predominantly exhibiting inflammatory characteristics, epithelial hyperplasia, adenomas, fibrocystic changes, and fibroadenomas. Conventional diagnostic approaches, primarily reliant on histopathology of breast tissue, are limited by inconsistencies stemming from subjective interpretation and variability among pathologists.3,4 Consequently, there is a demonstrated need for the integration of objective methodologies to enhance diagnostic accuracy and reliability.

Fourier-transform infrared (FTIR) spectroscopy is a technique that involves exposing a sample to infrared light while measuring the energy that either passes through or reflects from the sample as a function of wavenumbers.5,6 Studies by Elshemey et al.4 and Luo et al.5 highlight FTIR spectroscopy’s sensitivity to various biomolecules, enabling it to generate biochemical signatures that differentiate proteins, lipids, and nucleic acids based on the distinct vibration energy requirements of their functional groups.4–6 Characteristic features like peak shapes, intensities, and ratios are utilized for both qualitative and quantitative assessments of the samples.5–7

Consequently, FTIR spectroscopy emerges as a powerful diagnostic tool, with clinical applications extending beyond screening and diagnosis to the ongoing monitoring of treatment responses. Its use has been documented in analyzing diverse clinical samples, including formalin-fixed paraffin-embedded tissues, as well as cases associated with disorders like diabetes and gastric cancer,8,9 neurodegenerative diseases such as Alzheimer’s,10 and conditions like atherosclerosis,11 and breast cancer.12

Among FTIR techniques, attenuated total reflectance (ATR) spectroscopy is particularly prevalent. It typically employs a diamond to gather data from the layer of the sample adjacent to the internal reflection element surface.11,12 Its higher signal-to-noise ratio and reduced scattering make ATR more efficient for biological material assessments, combined with minimal sample preparation since infrared light penetration depth remains unaffected by sample thickness.

While FTIR spectroscopy is increasingly utilized in Nigeria and Sub-Saharan Africa, its application to biological tissues, particularly in cancer research, remains limited.13–16 Prior studies on breast cancer using vibrational spectroscopy have primarily been conducted outside the region,17,18 and local investigations are scarce. To date, only X-ray emission techniques have been employed in Nigeria to differentiate between cancerous and non-cancerous breast tissues.19 Therefore, this study aimed to evaluate the discriminatory performance of ATR-FTIR spectroscopy in differentiating normal, benign, and malignant breast lesions in the examined Nigerian tissue samples, identifying specific spectral signatures for tumor discrimination.

Materials and methods

Ethical consideration

Ethical approval for the study was granted by the Research Ethics Committee of Ladoke Akintola University of Technology (LAUTECH), Osun, Nigeria, with reference number: UTH/REC/2022/09/649. The study was carried out with a waiver of informed consent for the use of archived, anonymized samples in accordance with the principles of the Declaration of Helsinki. The ethics committee agreed that this study did not require informed consent due to its retrospective nature and the minimal risk involved.

Tissue retrieval and sorting

This retrospective study, conducted at LAUTECH, analyzed formalin-fixed paraffin-embedded biopsy samples from female patients collected between 2022 and 2023. The analyzed tissue set comprised 10 samples classified histologically as normal breast tissue, 15 benign lesions (fibroadenoma and fibrocystic changes), and 31 malignant specimens (invasive ductal carcinoma grade II), as shown in Figures 1 and 2. Using quota sampling, only primary tumors with complete information and that had not undergone neoadjuvant therapy during the study period were included.

Percentage proportion of normal (18%), fibroadenoma (14%), fibrocystic (13%), and malignant (invasive ductal carcinoma grade II) (55%) breast tissue.
Fig. 1  Percentage proportion of normal (18%), fibroadenoma (14%), fibrocystic (13%), and malignant (invasive ductal carcinoma grade II) (55%) breast tissue.
Hematoxylin and eosin-stained photomicrographs of (a) normal breast tissue section with lobules containing dark blue-stained acini nuclei (yellow arrow), separated by denser collagen fibers, compared with the stroma that separates the lobules ×400. (b) Breast lobules compressed into a slit-like shape (yellow arrow) due to excessive growth of the intervening stroma that separates the lobules (Fibroadenoma) ×400. (c) Cystic breast tissue with two large cysts (yellow arrow) filled with pink-staining amorphous material, lined by a single layer of epithelial cells ×400. (d) Invasive ductal carcinoma (Not Special Types).
Fig. 2  Hematoxylin and eosin-stained photomicrographs of (a) normal breast tissue section with lobules containing dark blue-stained acini nuclei (yellow arrow), separated by denser collagen fibers, compared with the stroma that separates the lobules ×400. (b) Breast lobules compressed into a slit-like shape (yellow arrow) due to excessive growth of the intervening stroma that separates the lobules (Fibroadenoma) ×400. (c) Cystic breast tissue with two large cysts (yellow arrow) filled with pink-staining amorphous material, lined by a single layer of epithelial cells ×400. (d) Invasive ductal carcinoma (Not Special Types).

This case was Nottingham grade II. Note the number of tubule formations, high pleomorphism, and mitotic figures in the picture ×400.

Tissue section preparation

Tissue sections measuring 15 µm were prepared from sorted tissue blocks using a Leica RM2125 microtome and transferred to aluminum foil substrates adapted for spectral acquisition according to Cui et al.20 An additional 4 µm section was stained with hematoxylin and eosin using a standard procedure.21 Tumor classification and identification of suitable areas were performed independently by two pathologists prior to FTIR analysis.

ATR-FTIR spectroscopic analysis on tissue samples

Following heat treatment at 55–60°C to dry the tissue sections, xylene (Surgipath Medical Industries, Inc.) was used for deparaffinization, followed by a descending ethanol series (Sigma-Aldrich) for rehydration, and then atmospheric drying. The Cary 630 Agilent spectrometer was calibrated against blank substrates before spectral acquisition. Point mapping was conducted on tissue sections placed on aluminum foil substrates, ensuring precise contact with the diamond ATR crystal at predefined normal and tumorous regions, previously annotated, following a methodology adapted from Baker et al.22 The tissue section fields were scanned across the mid-infrared range of 4,000 cm−1 to 600 cm−1, accumulating 32 scans at a resolution of 16 cm−1 and averaging ten spectra for each specimen. Spectra were preprocessed through baseline correction and smoothing techniques. Peaks and their relative intensities were identified, and spectral biomarkers corresponding to specific peak ratios were analyzed: A2922/A1632, A1632/A1543, A1632/A1080, A1080/A1543, A1237/A1080, and A1043/A1543, which were identified as biomarkers for diagnostic marker (DM),4 protein,5 cytoplasmic-nuclear ratio (CN),23 carcinogenesis marker,24 phosphate,25 and glycogen,26 respectively, as depicted in Figures 3 and 4.

Mean distribution of spectral biomarkers among normal, fibroadenoma, fibrocystic, and malignant breast tissues (invasive ductal carcinoma grade II).
Fig. 3  Mean distribution of spectral biomarkers among normal, fibroadenoma, fibrocystic, and malignant breast tissues (invasive ductal carcinoma grade II).
Spectra of (a) normal breast tissue showing peak positions and peak intensities, (b) breast fibroadenoma showing peak positions and peak intensities, (c) fibrocystic breast disease showing peak positions and peak intensities, (d) invasive breast cancer grade II showing peak positions and peak intensities.
Fig. 4  Spectra of (a) normal breast tissue showing peak positions and peak intensities, (b) breast fibroadenoma showing peak positions and peak intensities, (c) fibrocystic breast disease showing peak positions and peak intensities, (d) invasive breast cancer grade II showing peak positions and peak intensities.

Statistical analysis

Average values of these biomarkers were compared across normal, fibroadenoma, fibrocystic changes, and invasive ductal carcinoma grade II tissues, as displayed in Figure 3. Receiver operating characteristic (ROC) curves were employed to assess the sensitivity, specificity, and performance of the spectral biomarkers.4–6 The performance was quantified using area under the curve (AUC) metrics. An AUC of 0.5 indicates no discrimination, whereas values approaching 1.0 indicate greater discriminatory ability. Because AUC direction depends on the designated positive state, values below 0.5 were interpreted as separation in the opposite marker direction under the coding used, rather than as absence of separation. Data visualization and analysis were conducted using Microsoft Excel and SPSS 26, with results presented in both tables and graphs. Significance was set at a 95% confidence level (p < 0.05).

Results

Distribution of breast tissue types

In the study, the sampled breast tissue types were categorized into 10 normal, 8 fibroadenoma, 7 fibrocystic, and 31 malignant samples, as shown in Figure 1 below. The photomicrographs also showed the characteristics of different breast tissue types (Fig. 2).

Mean distribution of spectral biomarkers

The mean distribution of biomarkers was assessed across normal breast, fibroadenoma, fibrocystic, and malignant breast tissues, with the results presented in Figure 3. This provides a graphical highlight of how biomarkers differ among different breast tissue categories (normal and abnormal).

Quantitative determination of spectral biomarkers attributed to cytoplasm: nucleus ratios was particularly elevated in normal breast tissue (2.29), followed by fibrocystic (2.11), fibroadenoma (1.80), and lowest in malignant tissue (1.41). Similarly, glycogen levels were highest in malignant breast tissue (0.82), followed by fibroadenoma (0.58), fibrocystic (0.484), and normal (0.448). The carcinogenesis marker showed an increase in cancer tissue (0.81), relative to fibroadenoma (0.60), fibrocystic (0.52), and normal (0.48). Phosphate followed a similar pattern, with higher levels in malignant breast tissue compared to fibroadenoma, fibrocystic, and normal tissues. Protein and diagnostic markers in malignant, fibroadenoma, fibrocystic, and normal breast tissues were 1.13, 1.07, 1.10, and 1.10, and 0.53, 0.40, 0.39, and 0.45, respectively. The biomarkers showed distinct distribution patterns across the tissue groups; malignant tissue showed higher glycogen, carcinogenesis-marker, phosphate, protein, and diagnostic-marker values, whereas the cytoplasm:nucleus ratio was lower.

Diagnostic performance in pairwise breast tissue comparisons

To ascertain the discriminatory performance of biomarkers in discriminating between normal and fibroadenoma, normal and fibrocystic, normal and malignant breast tissue, fibroadenoma and malignant, fibrocystic and malignant, and fibroadenoma and fibrocystic tissues, ROC was performed to highlight the discriminatory power of these biomarkers within a 95% confidence interval, as presented in Tables 1–6. Furthermore, the sensitivity and specificity of these biomarkers were evaluated to assess their classification performance at the selected cut-off values (Table 7).

Table 1

Model performance of biomarkers between normal breast tissue and fibroadenoma

AreaSE95% CI
p-value
Lower boundUpper bound
Protein0.8100.1420.531.0000.138
DM0.8100.1790.4661.0000.138
CN1.0000.0011.0001.0000.002
CM0.0480.0690.0010.1830.03
Phosphate0.0950.1070.0010.3050.053
Glycogen0.0010.000130.0010.0010.017

For normal breast tissue versus fibroadenoma, the cytoplasm:nucleus ratio yielded an AUC of 1.000 (p = 0.002). Protein and the diagnostic marker each yielded an AUC of 0.810 (p = 0.138). Carcinogenesis marker, phosphate, and glycogen yielded AUCs below 0.5 under the coding used, reflecting separation in the opposite marker direction rather than absence of discrimination. AUC, area under the curve; CI, confidence interval; CM, carcinogenesis marker; CN, cytoplasmic:nucleus ratio; DM, diagnostic marker; SE, standard error.

Table 2

Model performance of biomarkers between normal breast and fibrocystic breast tissues

AreaSE95% CI
p-value
Lower boundUpper bound
Protein0.5830.2370.1191.0000.724
DM0.9170.1150.6911.0000.077
CN0.8330.1730.4931.0000.157
CM0.0830.1150.0010.3090.077
Phosphate0.3330.2720.0010.8670.480
Glycogen0.0010.000130.0010.0010.034

For normal versus fibrocystic breast tissue, the diagnostic marker and cytoplasm:nucleus ratio yielded AUCs of 0.917 (p = 0.077) and 0.833 (p = 0.157), respectively. Protein yielded an AUC of 0.583 (p = 0.724), whereas carcinogenesis marker, phosphate, and glycogen yielded AUCs below 0.5 under the coding used, reflecting separation in the opposite marker direction. AUC, area under the curve; CI, confidence interval; CM, carcinogenesis marker; CN, cytoplasmic:nucleus ratio; DM, diagnostic marker; SE, standard error.

Table 3

Model performance of biomarkers between normal breast and malignant (invasive ductal carcinoma grade II) breast tissues

AreaSE95% CI
p-value
Lower boundUpper bound
Protein0.3050.0750.1580.4580.256
DM0.2080.0870.0380.3770.089
CN0.9900.0120.9651.0000.004
CM0.0010.001NRNR0.004
Phosphate0.0050.0080.0010.020.004
Glycogen0.0050.0080.0010.020.004

For normal versus malignant breast tissue, the cytoplasm:nucleus ratio yielded an AUC of 0.990 (p = 0.004), as shown in Figure S3. Protein and the diagnostic marker yielded AUCs of 0.305 (p = 0.256) and 0.208 (p = 0.089), respectively. Carcinogenesis marker, phosphate, and glycogen yielded AUCs below 0.5 (p = 0.004 for each), reflecting separation in the opposite marker direction under the coding used. AUC, area under the curve; CI, confidence interval; CM, carcinogenesis marker; CN, cytoplasmic:nucleus ratio; DM, diagnostic marker; NR, not reported; SE, standard error.

Table 4

Model performance of biomarkers between fibroadenoma and malignant (invasive ductal carcinoma grade II) breast tissues

AreaSE95% CI
p-value
Lower boundUpper bound
Protein0.1480.0570.0360.2590.002
DM0.0250.0170.010.0590.001
CN0.9350.0280.8800.9910.001
CM0.0370.0210.0010.0770.001
Phosphate0.00810.0130.0010.0430.001
Glycogen0.0410.0220.0010.0840.001

For fibroadenoma versus malignant breast tissue, the cytoplasm:nucleus ratio yielded an AUC of 0.935 (p = 0.001), as shown in Figure S4. The remaining biomarkers yielded AUCs below 0.5 (p ≤ 0.002), reflecting separation in the opposite marker direction under the coding used. AUC, area under the curve; CI, confidence interval; CM, carcinogenesis marker; CN, cytoplasmic:nucleus ratio; DM, diagnostic marker; SE, standard error.

Table 5

Model performance of biomarkers between fibrocystic and malignant (invasive ductal carcinoma grade II) breast tissues

AreaSE95% CI
p-value
Lower boundUpper bound
Protein0.2980.0880.1250.470.177
DM0.0080.010.0010.0280.001
CN0.9760.0190.941.0000.02
CM0.0080.010.0010.0270.001
Phosphate0.0010.0010.0010.0020.001
Glycogen0.0040.0060.0010.0160.001

For fibrocystic versus malignant breast tissue, the cytoplasm:nucleus ratio yielded an AUC of 0.976 (p = 0.02), as shown in Figure S4. Protein yielded an AUC of 0.298 (p = 0.177), while the diagnostic marker, carcinogenesis marker, phosphate, and glycogen yielded AUCs below 0.5 (p = 0.001 for each), reflecting separation in the opposite marker direction under the coding used. AUC, area under the curve; CI, confidence interval; CM, carcinogenesis marker; CN, cytoplasmic:nucleus ratio; DM, diagnostic marker; SE, standard error.

Table 6

Model performance of biomarkers between fibrocystic and fibroadenoma breast tissues

AreaSE95% CI
p-value
Lower boundUpper bound
Protein0.2140.150.0010.5080.131
DM0.6430.1740.3010.9840.45
CN0.0710.0820.0010.2320.023
CM0.8570.1320.5981.0000.059
Phosphate0.8210.1370.5521.0000.089
Glycogen1.0000.0011.0001.0000.008

For fibrocystic versus fibroadenoma breast tissue, the diagnostic marker, carcinogenesis marker, phosphate, and glycogen yielded AUCs of 0.643 (p = 0.45), 0.857 (p = 0.059), 0.821 (p = 0.089), and 1.000 (p = 0.008), respectively. The cytoplasm:nucleus ratio yielded an AUC of 0.071 (p = 0.023), reflecting separation in the opposite marker direction under the coding used. AUC, area under the curve; CI, confidence interval; CM, carcinogenesis marker; CN, cytoplasmic:nucleus ratio; DM, diagnostic marker; SE, standard error.

Table 7

Peak ratios, important biomarker assignment, sensitivity, specificity, and cut-off points of comparison among normal, fibroadenoma, fibrocystic, and malignant breast tissues

Histology classificationPeak ratiosBiomarker assignmentSensitivity%Specificity%Cut-off points
Normal
  MalignantA1632/A1080CN100692.10
Fibroadenoma
  MalignantA1632/A1080CN100861.58
Fibrocystic
  MalignantA1632/A1080CN100972.10
Normal
  FibroadenomaA1632/A1543Protein100711.09
Normal
  FibroadenomaA2922/A1632DM100430.39
Normal
  FibroadenomaA1632/A1080CN1001002.10
Normal
  FibrocysticA2922/A1632DM100750.39
Normal
  FibrocysticA1632/A1080CN100502.08
Fibroadenoma
  FibrocysticA2922/A1632DM71750.39
Fibroadenoma
  FibrocysticA1080/A1543CM861000.54
Fibroadenoma
  FibrocysticA1237/A1080Phosphate821000.77
Fibroadenoma
  FibrocysticA1043/A1543Glycogen1001000.51

CM, carcinogenesis marker; CN, cytoplasmic:nucleus ratio; DM, diagnostic marker.

Table 7 summarizes the sensitivity and specificity of the selected pairwise comparisons. For the CN marker, sensitivity was 100% in all listed comparisons, while specificity ranged from 50% to 100%. In comparisons with malignant tissue, CN specificity was 69% for normal versus malignant, 86% for fibroadenoma versus malignant, and 97% for fibrocystic versus malignant tissue. The protein marker yielded 100% sensitivity and 71% specificity for normal versus fibroadenoma. Glycogen yielded 100% sensitivity and specificity for fibroadenoma versus fibrocystic tissue. Other markers showed comparison-specific sensitivity and specificity values, as detailed in Table 7.

The selected CN cut-offs were 2.10 for normal versus malignant, 1.58 for fibroadenoma versus malignant, 2.10 for fibrocystic versus malignant, 2.10 for normal versus fibroadenoma, and 2.08 for normal versus fibrocystic tissue. A diagnostic-marker cut-off of 0.39 was used for normal versus fibroadenoma, normal versus fibrocystic, and fibroadenoma versus fibrocystic comparisons (Table 7).

Discussion

The current study analyzed breast tissue classified histopathologically as normal, benign (fibroadenoma and fibrocystic), and malignant using ATR-FTIR spectroscopy to identify spectral peak intensities as potential biomarkers for proteins, diagnostic markers, carcinogenesis markers, cytoplasmic-nuclear ratios, phosphate, and glycogen. ROC curve analysis evaluated these biomarkers based on their ability to distinguish the examined breast tissue types. In particular, biomarkers demonstrating high discriminatory power, as reflected in AUC values, were prioritized. The AUC provides a threshold-independent summary of discriminatory performance.27–30 Thus, ROC curve analysis was systematically employed across the selected biomarkers to establish their respective sensitivities, specificities, AUCs, and cut-off points when contrasting breast tissue types.

ROC curve analysis provides a visual and quantitative method for evaluating overall test performance by plotting sensitivity against specificity. The AUC summarizes discrimination across thresholds. An AUC of 0.5 indicates no discrimination. Values approaching 1.0 indicate greater discriminatory ability, whereas values below 0.5 indicate separation in the opposite direction to the designated positive state under the coding used.29,30

Sensitivity measures the likelihood of accurately identifying diseased patients, while specificity assesses the capability of a test to exclude healthy individuals correctly.6 The necessity to reduce the proportion of false positives is crucial in ensuring proper screening for positive subjects.27,28 In the examined tissue groups, ATR-FTIR showed comparison-specific discriminatory performance across the assessed tissue categories. In Table 3 and Table 7, the CN ratio, shown by the peak ratio A1632/A1080, was significantly elevated in normal compared to malignant tissues (p = 0.004, AUC = 0.990, sensitivity = 100%, specificity = 69%). This ratio also demonstrated remarkable discriminatory ability (p = 0.002; AUC = 1.00, sensitivity = 100%, specificity = 100%) between normal and fibroadenoma, as shown in Table 1 and Table 7. Moreover, the protein peak ratio A1632/A1543 in Tables 1 and 7 showed an AUC of 0.810 with 100% sensitivity and 71% specificity (p = 0.138). The diagnostic marker (A2922/A1632) yielded AUC = 0.810, specificity = 43% (p = 0.138) with 100% sensitivity in Tables 1 and 7.

For normal versus fibrocystic tissue, the diagnostic marker yielded an AUC of 0.917 with 100% sensitivity and 75% specificity (p = 0.077), while the cytoplasmic-nuclear ratio yielded an AUC of 0.833 with 100% sensitivity and 50% specificity (p = 0.157) (Tables 2 and 7). The cytoplasmic-nuclear ratio yielded an AUC of 0.935 with 100% sensitivity and 86% specificity for fibroadenoma versus malignant tissue (p = 0.001) and an AUC of 0.976 with 100% sensitivity and 97% specificity for fibrocystic versus malignant tissue (p = 0.02) (Tables 4, 5, and 7). For fibrocystic versus fibroadenoma tissue, glycogen yielded an AUC of 1.000 with 100% sensitivity and specificity; carcinogenesis marker and phosphate yielded AUCs of 0.857 and 0.821, respectively, and the diagnostic marker yielded an AUC of 0.643 (Tables 6 and 7).

The cytoplasmic-nuclear ratio showed potential as a discriminatory marker in the examined tissue groups. The cytoplasmic-nuclear ratio has traditionally been used in histopathological assessment.1,2 Related analytical methods have evaluated nucleus-to-cytoplasm metrics.31–33 Pathology references also describe morphologic differences among benign and malignant breast lesions.34 Across the comparisons listed in Table 7, CN specificity varied from 50% to 100%, indicating that performance depended on the tissue comparison.

While the peak ratio A2922/A1650 has exhibited diagnostic capability in previous studies, approaching AUC values of 0.908 and 100% sensitivity in distinguishing diseased from healthy tissues,4,6 these findings were not uniformly observed in the present study. Similar 100% sensitivity with 71% specificity and an AUC value of 0.810 were noted primarily between normal and fibroadenoma tissues. This suggests that further exploration of these biomarkers could yield significant insights into breast cancer dynamics—a notion echoed in complementary studies.6 The protein peaks, generally assumed to elevate in cancerous tissues compared to normal counterparts, did not exhibit notable statistical significance in this study. While statistical significance was generally elusive, the protein marker yielded an AUC below 0.5 under the coding used (AUC = 0.305, p = 0.256), and this result was not statistically significant. This pattern may reflect biological or pre-analytical variability, including differences in protein secondary structure,35,36 unlike the more stable α-sheet derived proteins used in prior studies.37 This contrasts with reports demonstrating the utility of α:β ratios and Amide II/Amide III ratios in serum for distinguishing breast cancer patients from healthy subjects.37 Variations in peak ratios from differing protein structures can impact diagnostic sensitivity, and despite high sensitivity and diagnostic performance between normal and fibroadenoma tissues, statistical significance was lacking.33 Thus, the implementation of this protein biomarker warrants careful consideration due to its lack of statistical significance, potentially stemming from confounding variables like formalin fixation.38 Nonetheless, the observed trends suggest potential diagnostic utility.

Furthermore, the elevated glycogen-related signal in cancerous tissues is consistent with previous evidence of increased glycogen accumulation in metastatic breast cancer cells.39 In the present study, the glycogen-related marker distinguished fibroadenoma from fibrocystic changes, with an AUC of 1.00 and 100% sensitivity and specificity. Elsewhere, salivary carbohydrate profiles, specifically the peak at 1,041 cm−1, have been reported to exhibit moderate diagnostic utility in breast cancer detection.6 Ferreira et al.6 reported an AUC of 0.765–0.770, with 80% sensitivity and 70% specificity for differentiating healthy controls from breast cancer patients. The same peak also demonstrated 70% sensitivity and 70% specificity in distinguishing patients with benign breast conditions from healthy individuals using saliva samples.

Additionally, the phosphate peak was notably higher in malignant tissues compared to fibrocystic and benign forms, aligning with prior findings that link phosphate levels to nucleic acid activity in malignant transformations.40,41 Phosphate showed comparison-dependent performance; AUC values below 0.5 in malignant comparisons reflected the coding direction, whereas the fibroadenoma-versus-fibrocystic comparison yielded an AUC of 0.821. Its overall discriminatory performance therefore varied by tissue comparison. It showed discrimination between fibrocystic and fibroadenoma tissues with sensitivity and specificity rates of 82% and 100%, respectively, though this contrasts with other studies reporting less than 90% for both metrics.4,6 The carcinogenesis marker appeared to aid primarily in differentiating fibrocystic tissues from fibroadenoma,23,24 which may warrant further evaluation for benign variant characterization and diagnosis.

While this study shows promising discriminatory findings, several limitations warrant cautious interpretation of the findings and restrict their broad applicability. These include a modest sample size and the potential for confounding biases, particularly stemming from uncontrolled pre-analytical and analytical variables during tissue block preparation. Future investigations should address these limitations by employing larger, prospectively controlled cohorts to validate these results. The ROC cut-offs and performance estimates were also derived from the same relatively small cohort without independent validation, which may overestimate diagnostic performance.

Conclusions

ATR-FTIR spectroscopy showed promising discriminatory performance among the examined normal, benign, and malignant breast tissue groups using selected spectral peak ratios. The cytoplasmic-nuclear ratio yielded high reported AUCs in selected comparisons with malignant tissue, while glycogen yielded an AUC of 1.000 with 100% sensitivity and specificity for fibroadenoma versus fibrocystic change. These findings should be interpreted as exploratory and require validation in larger independent cohorts.

Supporting information

Supplementary material for this article is available at https://doi.org/10.14218/CSP.2025.00004.

Fig. S1

ROC curve of biomarkers showing moderate to excellent diagnostic performance discriminating normal and fibroadenoma at p-value (0.05). (a) ROC curve with AUC of protein (0.810)*** at p = 0.138 between normal and fibroadenoma breasts. (b) ROC curve with AUC of diagnostic marker (0.810)*** at p = 0.138 between normal and fibroadenoma breasts. (c) ROC curve with AUC of cytoplasm:nucleus ratio (1.00)*** at p = 0.002 between normal and fibroadenoma breasts. **moderate diagnostic performance, ***excellent diagnostic performance. AUC, area under the curve; ROC, receiver operating characteristic.

(TIF)

Click here for additional data file.

Fig. S2

ROC curve of biomarkers showing moderate to excellent diagnostic performance discriminating normal and fibrocystic at p-value (0.05). (a) ROC curve with AUC of diagnostic marker (0.917)*** at p = 0.077 between normal and fibrocystic breasts. (b) ROC curve with AUC of cytoplasm:nucleus ratio (0.833)*** at p = 0.157 between normal and fibrocystic breasts. AUC, area under the curve; ROC, receiver operating characteristic. **moderate diagnostic performance, ***excellent diagnostic performance.

(TIF)

Click here for additional data file.

Fig. S3

ROC curve of biomarkers showing moderate to excellent diagnostic performance discriminating normal and malignant breast and benign and malignant breast at p-value (0.05). (a) ROC curve with AUC of cytoplasmic-nucleus ratio (0.990)*** at p = 0.004 between normal and malignant breast. (b) ROC curve with AUC of cytoplasmic-nucleus ratio (0.943)*** at p = 0.001 between benign and malignant breast. **moderate diagnostic performance, ***excellent diagnostic performance. AUC, area under the curve; ROC, receiver operating characteristic.

(TIF)

Click here for additional data file.

Fig. S4

ROC curve of biomarkers showing moderate to excellent diagnostic performance discriminating fibroadenoma and malignant breast and fibrocystic and malignant breast at p-value (0.05). (a) ROC curve with AUC of cytoplasm:nucleus ratio (0.935)*** at p = 0.001 between fibroadenoma and malignant breast. (b) ROC curve with AUC of cytoplasm:nucleus ratio (0.976)*** at p = 0.02 between fibrocystic and malignant breast. **moderate diagnostic performance, ***excellent diagnostic performance. AUC, area under the curve; ROC, receiver operating characteristic.

(TIF)

Click here for additional data file.

Fig. S5

ROC curve of biomarkers showing moderate to excellent diagnostic performance discriminating fibroadenoma and fibrocystic at p-value (0.05). (a) ROC curve with AUC of diagnostic marker (0.643)* at p = 0.45 between fibroadenoma and fibrocystic breasts. (b) ROC curve with AUC of phosphate (0.821)*** at p = 0.09 between fibroadenoma and fibrocystic breasts. (c) ROC curve with AUC of carcinogenesis marker (0.857)*** at p = 0.06 between fibroadenoma and fibrocystic breasts. (d) ROC curve with AUC of glycogen (1.00)*** at p = 0.008 between fibroadenoma and fibrocystic breasts. *fair diagnostic performance, **moderate diagnostic performance, ***excellent diagnostic performance. AUC, area under the curve; ROC, receiver operating characteristic.

(TIF)

Click here for additional data file.

Fig. S6

ROC curve of biomarkers showing moderate to excellent diagnostic performance discriminating normal and benign breast at p-value (0.05). (a) ROC curve with AUC of protein (0.727)** at p = 0.243 between normal and benign breast. (b) ROC curve with AUC of diagnostic marker (0.848)*** at p = 0.073 between normal and benign breast. (c) ROC curve with AUC of cytoplasmic-nucleus ratio at p (0.939)*** = 0.024 between normal and benign breast. **moderate diagnostic performance, ***excellent diagnostic performance. AUC, area under the curve; ROC, receiver operating characteristic.

(TIF)

Click here for additional data file.

Declarations

Acknowledgement

The authors are grateful to Dr. Blessings Effiong (Biochemistry Department, University of Uyo, Akwa Ibom, Nigeria) for advice on analyzing a tissue section with ATR-FTIR, Dr. Waheed Oluogun, and Mr. Richard Oduru (LAUTECH Teaching Hospital, Osogbo) for technical support and assistance during sample collection.

Ethical statement

Ethical approval for the study was granted by the Research Ethics Committee of Ladoke Akintola University of Technology (LAUTECH), Osun, Nigeria, with reference number: UTH/REC/2022/09/649. The study was carried out with a waiver of informed consent for the use of archived, anonymized samples in accordance with the principles of the Declaration of Helsinki. The ethics committee agreed that this study did not require informed consent due to its retrospective nature and the minimal risk involved.

Data sharing statement

The spectral and receiver operating characteristic (ROC) plots supporting the findings of this study are included in the article and supplementary materials.

Funding

None.

Conflict of interest

The authors have no conflict of interest related to this publication.

Authors’ contributions

Study concept and design (STA, CI), acquisition of data (STA, SI), analysis and interpretation of data (STA, CI, SI), drafting of the manuscript (STA), critical revision of the manuscript for important intellectual content (STA, CI, SI), administrative, technical, or material support (CI, SI), and study supervision (STA, CI). All authors have made significant contributions to this study and have approved the final manuscript.

References

  1. Kumar A, Abbas AK, Aster JC. Robbins Basic Pathology. Tenth Edition. Pennsylvania: Elsevier Philadelphia; 2018:736–747
  2. Kumar V, Abbas AK, Aster JC, Deyrup AT, editors. Kumar V, Abbas AK, Aster JC, Deyrup AT, editors. Robbins & Kumar Basic Pathology. 11th ed. Philadelphia: Elsevier; 2022
  3. Zhu J, Liu M, Li X. Progress on deep learning in digital pathology of breast cancer: a narrative review. Gland Surg 2022;11(4):751–766 View Article PubMed/NCBI
  4. Elshemey WM, Ismail AM, Elbialy NS. Molecular-Level Characterization of normal, benign, and malignant breast tissues using FTIR spectroscopy. J Med Biol Eng 2016;36(3):369–378 View Article
  5. Luo Y, Liu H, Wu C, Paraskevaidi M, Deng Y, Shi W, et al. Diagnostic segregation of human breast tumours using Fourier-transform infrared spectroscopy coupled with multivariate analysis: Classifying cancer subtypes. Spectrochim Acta A Mol Biomol Spectrosc 2021;255:119694 View Article PubMed/NCBI
  6. Ferreira ICC, Aguiar EMG, Silva ATF, Santos LLD, Cardoso-Sousa L, Araújo TG, et al. Attenuated Total Reflection-Fourier Transform Infrared (ATR-FTIR) Spectroscopy Analysis of Saliva for Breast Cancer Diagnosis. J Oncol 2020;2020:4343590 View Article PubMed/NCBI
  7. Alhajjeh R, Al-Ali H. Application of FTIR spectroscopy method for the quantification of ascorbic acid in bulk materials and pharmaceutical formulation. Iraqi J Pharm Sci 2023;32(3):186–194 View Article
  8. Farooq S, Zezell DM. Diabetes Monitoring through Urine Analysis Using ATR-FTIR Spectroscopy and Machine Learning. Chemosensors 2023;11(11):565 View Article
  9. Pang N, Yang W, Yang G, Yang C, Tong K, Yu R, et al. The utilization of blood serum ATR-FTIR spectroscopy for the identification of gastric cancer. Discov Oncol 2024;15(1):350 View Article PubMed/NCBI
  10. Paraskevaidi M, Karim S, Santos M, Lima K, Crean S. The use of ATR-FTIR spectroscopy for the diagnosis of Alzheimer’s disease using oral buccal cells. Appl Spectrosc Rev 2024;59(8):1021–1035 View Article
  11. Kęsik JJ, Paja W, Jakubczyk P, Khalavka M, Terlecki P, Iłżecki M, et al. Determination of spectroscopy marker of atherosclerotic carotid stenosis using FTIR-ATR combined with machine learning and chemometrics analyses. Nanomedicine 2024;62:102788 View Article PubMed/NCBI
  12. Sitnikova VE, Kotkova MA, Nosenko TN, Kotkova TN, Martynova DM, Uspenskaya MV. Breast cancer detection by ATR-FTIR spectroscopy of blood serum and multivariate data-analysis. Talanta 2020;214:120857 View Article PubMed/NCBI
  13. Shehu UF, Awwalu S, Adeshina GO. FTIR analysis and antimicrobial evaluation of aqueous ethanol leaf extract and fractions of pavonia snegalensis (Cav.) listener (Malvaceae). J Curr Biomed Rep 2021;1(3)
  14. Amponsah IK, Boakye A, Orman E, Armah FA, Borquaye LS, Adjei S, et al. Assessment of some quality parameters and chemometric-assisted FTIR spectral analysis of commercial powdered ginger products on the Ghanaian market. Heliyon 2022;8(3):e09150 View Article PubMed/NCBI
  15. Njokunwogbu AN, Amarachukwu O, Ene FC, Iyioku MU, Enyi CN. Separation And Infrared (Ft-Ir) Characterization Of Aromatic Hydrocarbons From High Boiling Fractions (≫200 Oc) Of Nigeria Crude Oil (Using Trichloroethylene As The Eluant). Int J Res Appl Sci Eng Technol 2022;8(8):1–9
  16. Ityo SD, Anhwange BA, Okoye PAC, Feka PD. Sensory, GC-MS and FTIR analysis of aqueous extract of hibiscus sabdarriffa and Vernonia amygdalina herbal tea with blends of ginger and lemon zest. J Chem Soc Nigeria 2023;48(4):662–676 View Article
  17. Lazaro-Pachecho D, Shaaban A, Baldwin G, Titiloye NA, Rehman S, Rehman IU. Deciphering the structural and chemical composition of breast cancer using FTIR spectroscopy. Appl Spectrosc Rev 2020;57(3):234–248 View Article
  18. Lazaro-Pacheco D, Shaaban AM, Titiloye NA, Rehman S, Rehman IU. Elucidating the chemical and structural composition of breast cancer using Raman micro-spectroscopy. EXCLI J 2021;20:1118–1132 View Article PubMed/NCBI
  19. Olaiya DO, Alatise OI, Oketayo OO, Abiye OE, Obianjunwa EI, Balogun FA. Trace Element Analysis of Cancerous and Non-cancerous Breast Tissues of African Women in Southwest Nigeria Using Particle-Induced X-ray Emission Technique. Breast Cancer (Auckl) 2019;13:1178223419840694 View Article PubMed/NCBI
  20. Cui L, Butler HJ, Martin-Hirsch PL, Martin FL. Aluminium foil as a potential substrate for ATR-FTIR, transflection FTIR, or Raman spectrochemical analysis of biological specimen. Anal Methods 2016;8:481–487 View Article
  21. Carson FL, Cappellano CH. Histotechnology: A Self Instructional Text. 4th ed. Chicago, Illinois: American Society College of Pathology press; 2015:264–282
  22. Baker MJ, Trevisan J, Bassan P, Bhargava R, Butler HJ, Dorling KM, et al. Using Fourier transform IR spectroscopy to analyze biological materials. Nat Protoc 2014;9(8):1771–1791 View Article PubMed/NCBI
  23. Sahu RK, Argov S, Salman A, Huleihel M, Grossman N, Hammody Z, et al. Characteristic absorbance of nucleic acids in the Mid-IR region as possible common biomarkers for diagnosis of malignancy. Technol Cancer Res Treat 2004;3(6):629–638 View Article PubMed/NCBI
  24. Yu G, Xu JL, Niu Y, Zhang CZ, Zhang CP. Studies on breast tumor tissues with ATR-FTIR spectroscopy. In: Chance B, Chen MZ, Chiou AET, Luo QM, editors. Optics in Health Care and Biomedical Optics: Diagnostics and Treatment II, Vol. 5630. SPIE; 2005:796–801 View Article
  25. Kumar S, Verma T, Mukherjee R, Ariese F, Somasundaram K, Umapathy S. Raman and infra-red microspectroscopy: towards quantitative evaluation for clinical research by ratiometric analysis. Chem Soc Rev 2016;45(7):1879–1900 View Article PubMed/NCBI
  26. Yano K, Sakamoto Y, Hirosawa N, Tanooka S, Katayama H, Kumaido K, et al. Applications of fourier transform infrared spectroscopy, fourier transform infrared microscopy and near infrared spectroscopy to cancer research. J Spectrosc 2003;17(2-3):315–321 View Article
  27. Bradley AP. The use of the area under the ROC curve in the evaluation of machine learning algorithms. Pattern Recognit 1997;30(7):1145–1159 View Article
  28. Vrbin CM. Recommendations for reporting measures of diagnostic accuracy. Cytopathology 2023;34(3):185–190 View Article PubMed/NCBI
  29. Shreffler J, Huecker MR. Diagnostic Testing Accuracy: Sensitivity, Specificity, Predictive Values and Likelihood Ratios. Treasure Island (FL): StatPearls Publishing; 2025 PubMed/NCBI
  30. Çorbacıoğlu ŞK, Aksel G. Receiver operating characteristic curve analysis in diagnostic accuracy studies: A guide to interpreting the area under the curve value. Turk J Emerg Med 2023;23(4):195–198 View Article PubMed/NCBI
  31. Moore MJ, Sebastian JA, Kolios MC. Determination of cell nucleus-to-cytoplasmic ratio using imaging flow cytometry and a combined ultrasound and photoacoustic technique: a comparison study. J Biomed Opt 2019;24(10):1–10 View Article PubMed/NCBI
  32. Sebastian JA, Moore MJ, Berndl ESL, Kolios MC. An image-based flow cytometric approach to the assessment of the nucleus-to-cytoplasm ratio. PLoS One 2021;16(6):e0253439 View Article PubMed/NCBI
  33. Andrus PG. Cancer monitoring by FTIR spectroscopy. Technol Cancer Res Treat 2006;5(2):157–167 PubMed/NCBI
  34. Carton J. Oxford handbook of pathology. 3rd ed. Oxford: Oxford University Press; 2024:349–364
  35. Zelig U, Barlev E, Bar O, Gross I, Flomen F, Mordechai S, et al. Early detection of breast cancer using total biochemical analysis of peripheral blood components: a preliminary study. BMC Cancer 2015;15:408 View Article PubMed/NCBI
  36. Wang H, Logan DT, Danielsson J, Oliveberg M. Exposing the distinctive modular behavior of β-strands and α-helices in folded proteins. Proc Natl Acad Sci U S A 2020;117(46):28775–28783 View Article PubMed/NCBI
  37. Ghimire H, Garlapati C, Janssen EAM, Krishnamurti U, Qin G, Aneja R, et al. Protein Conformational Changes in Breast Cancer Sera Using Infrared Spectroscopic Analysis. Cancers (Basel) 2020;12(7):1708 View Article PubMed/NCBI
  38. Bauer DR, Chafin DR. Assessing Tissue Fixation Time and Quality with Label-free Mid Infrared Spectroscopy and Machine Learning. Biopreserv Biobank 2023;21(2):208–216 View Article PubMed/NCBI
  39. Hicks E, Layosa MA, Andolino C, Truffer C, Song Y, Heden TD, et al. Gluconeogenesis and glycogenolysis required in metastatic breast cancer cells. Front Oncol 2024;14:1476459 View Article PubMed/NCBI
  40. Amjad M, Ullah H, Andleeb F, Batool Z, Nazir A, Gilanie G. Fourier transform infrared spectroscopy for investigation of human carcinoma and leukaemia. Lasers Eng 2021;51(1-5):217–233
  41. Depciuch J, Barnaś E, Skręt-Magierło J, Skręt A, Kaznowska E, Łach K, et al. Spectroscopic evaluation of carcinogenesis in endometrial cancer. Sci Rep 2021;11(1):9079 View Article PubMed/NCBI
Copyright © 2025 Authors. This is an Open Access article distributed under the terms of the Creative Commons Attribution-Noncommercial 4.0 License (CC BY-NC 4.0), permitting all non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.

About this Article

Cite this article
Adeleke ST, Igbeneghu C, Iyiola S. Diagnostic Performance of ATR-FTIR Spectroscopy in Discriminating Normal Breast Tissue and Breast Tumors. Cancer Screen Prev. 2025;4(2):79-88. doi: 10.14218/CSP.2025.00004.
Copy        Export to RIS        Export to EndNote
Article History
Received Revised Accepted Published
January 30, 2025 March 27, 2025 April 27, 2025 May 26, 2025
DOI http://dx.doi.org/10.14218/CSP.2025.00004
  • Cancer Screening and Prevention
  • pISSN 2993-6314
  • eISSN 2835-3315
  • © 2026 Xia & He Publishing Inc.
Back to Top

Diagnostic Performance of ATR-FTIR Spectroscopy in Discriminating Normal Breast Tissue and Breast Tumors

Samuel T. Adeleke, Christopher Igbeneghu, Sina Iyiola
  • Reset Zoom
  • Download TIFF