Advances in the clinical application of Raman spectroscopy in breast cancer

Abstract

Recently, Raman spectroscopy has made phased progress in clinical research of breast cancer. Raman spectroscopy is a nondestructive optical analysis technology that can quickly provide biochemical molecular specific information of samples. The analysis of Raman spectral characteristics at different stages of clinical activity in breast cancer patients is helpful to formulate individualized diagnosis and treatment plans, improve surgical results, adjust treatment strategies, and promote prognosis. Due to the development of optical technology and the diversity of clinical research needs, derivative Raman techniques based on Raman effect have emerged, such as surface-enhanced Raman spectroscopy, resonance Raman spectroscopy, spatial shifted Raman spectroscopy and coherent Raman scattering spectroscopy. In this paper, the research progress of Raman spectroscopy and its derivative techniques in the diagnosis and treatment of breast cancer in recent years is reviewed. The diagnostic value of different Raman techniques in different levels of breast cancer cells, body fluids and breast tissue and microcalcification molecules is evaluated. In addition, the application status and prospects of Raman spectroscopy and its derivative techniques in breast cancer surgery, chemotherapy and radiotherapy are analyzed and discussed.

Keywords:

• Breast cancer
• Raman spectroscopy
• Raman biopsy
• Raman imaging
• intra-operative assessment
• lymph nodes

Introduction

Breast cancer has become the malignant disease with the highest incidence of cancer in women, accounting for about 6.9% of cancer-related deaths every year. In addition, the incidence of breast cancer is also increasing year by year. Projections from the International Agency for Research on Cancer (IARC) indicate that by 2040, the number of new breast cancer cases is expected to exceed 3 million annually, accompanied by over 1 million deaths [1, 2]. As the protracted course of the disease and the irreversible damage to health resulting from breast cancer, it is imperative to implement timely and comprehensive management in breast cancer treatment. Specifically, the precision of early diagnosis and the rational formulation of treatment strategies hold paramount significance in controlling disease progression and enhancing the prognosis for patients.

Currently, histopathological detection is the primary clinical diagnosis method of breast cancer, which is recognized as the gold standard for breast cancer detection [3, 4]. However, histological detection has been found to lead to a high rate of missed diagnosis in breast cancer detection. The main reason is that this method can only detect the marker points rather than all parts of the resection margin, and some resection margin tissues are inevitably lost during the freezing process of pathological samples [5–8]. Therefore, there is a need for a noninvasive, convenient and rapid detection method for the diagnosis of breast cancer. Raman spectroscopy (RS), distinguished by its features of no sample preparation, nondestructive sampling, straightforward and rapid analysis, and high resolution, has gained widespread utilization in the breast cancer diagnosis [9–12]. Furthermore, the RS has the potential to detect pathological changes prior to imaging detection and the ability to detect biochemical molecular changes in samples [13–16]. However, to the best of our knowledge, the existing reviews predominantly concentrate on the application of RS in breast cancer diagnosis or the evaluation of surgical margins, with a noticeable gap in reviews addressing its broader application in breast cancer diagnosis and treatment [17–21]. This article fills that gap by providing a comprehensive review of the clinical application of RS in recent years, encompassing breast cancer diagnosis, surgical treatment, drug therapy, and radiotherapy. It aims to summarize research progress, identify challenges to be addressed, and outline future development trends in this field.

Principles and methods of Raman spectroscopy for biological detection

When a monochromatic laser irradiates a sample, it induces both elastic scattering and inelastic scattering within the molecules present. Elastic scattering, characterized by no energy exchange between the incident photon and the molecule, leads to the generation of photons moving in different directions without altering their frequency—a phenomenon known as Rayleigh scattering. In contrast, inelastic scattering, also referred to as Raman scattering, involves energy exchange between incident photons and molecules during collision. Some energy is either absorbed by photons or sample molecules, resulting in scattered photons with changes in both direction and frequency—an occurrence termed the Raman effect. Upon irradiation by incident light, the sample molecule’s energy level undergoes excitation, reaching a higher virtual energy level. As the energy attenuates, the molecule’s final energy level in Rayleigh scattering returns to its ground energy level. In Raman scattering, however, the final energy level of the molecule can be either higher or lower than its original energy level. When the final level is higher, it is termed Stokes Raman scattering; conversely, when it is lower, it is referred to as anti-Stokes Raman scattering. The energy level transition of Raman effect is illustrated in Figure 1.

Figure 1. The energy-level diagram of Rayleigh scattering, Stokes Raman scattering and anti-Stokes Raman scattering.

Raman spectroscopy, as a powerful spectroscopic technique, has the capability to discern different chemical components by detecting the vibration modes of chemical bonds within samples. After a laser beam is incident on the sample, there are three possible outcomes: some of the light becomes transmitted light, some is absorbed by the sample, and some becomes scattered light. Most of the scattered light has the same wavelength as the incident light, but a small portion experiences a wavelength shift due to the vibration and rotation of molecules in the sample. This Raman-shifted light carries information about the molecular composition and functional groups of the sample (e.g., C–H, C–N, O–H). Typically, this information is collected by a Raman spectrometer and processed by its computer system to generate a plot of Raman scattering intensity as a function of wavelength, known as a Raman spectrum. The wavenumber is the reciprocal of the wavelength. Since the wavelength is usually measured in centimeters, the unit of the wavenumber is cm-1. The scattered light, carrying molecular information, is eventually manifested as distinct peaks in the Raman spectrum, allowing for the identification of the molecular concentration and properties within the sample through spectral analysis. The RS are categorized into three wavenumber regions: high wavenumber region, silent region, and the fingerprint region. The high wavenumber region, ranging between 2800 cm⁻¹ and 3000 cm⁻¹, encompasses the stretching vibration of C–H, O–H, N–H, and other chemical bonds found in lipids, proteins, and water. The silent region, usually situated between the high wavenumber region and fingerprint region (typically 1800–2800 cm⁻¹), features low Raman scattering intensity, primarily comprising molecular internal vibration and lattice vibration. The fingerprint region, with a wavenumber frequency band typically between 600 cm⁻¹ and 1800 cm⁻¹, is positioned between the silent and high wavenumber regions. This region contains abundant molecular vibration information, including characteristic peaks of proteins, lipids, carbohydrates, and nucleic acids in biological samples. The Raman spectrum of healthy breast tissue and breast cancer tissue are shown in Figure 2. By leveraging the molecular characteristic fingerprints provided by RS, the qualitative and quantitative analysis of the sample’s molecular composition becomes possible. Table 1 shows the biochemical molecular distribution of RS peaks in detail.

Figure 2. Information content of laser Raman spectra of healthy and cancerous tissue. The average spectral with 1-sigma error bars for the histopathology 100% (tumor) and 0% (healthy) quintiles of the study population. Reprinted with permission from [22]. Copyright [2021] Springer nature.

Table 1. Assignment of spectral bands in the Raman spectrum.

Peak position (cm^−1)	Assignment
796	DNA|RNA ring breathing modes and O–P–O backbone
828	Proline, hydroxyproline, tyrosine, O–P–O); mono- and polysaccharides
1048	Symmetric stretch vibration of ν3PO4 ^3− in hydroxyapatite and glycogen
1300	Lipids, collagen, protein amide III
1437	Lipids, fatty acids, triglycerides, collagen, and phospholipids
1654	Amide I if collagen assignment, and/or C = C of lipids in normal tissue
2853	Symmetric stretch of lipids
2896	Asymmetric stretch of protein, lipids, glycogen
2937	C–H vibrations in lipids, proteins, glycogen, DNA|RNA

Reprinted with permission from [22]. Copyright [2021] Springer nature.

Biological metabolism undergoes significant changes throughout the transition from a physiological to a pathological state within the human body [23, 24]. Raman spectroscopy and other spectral detection derivatization techniques, has demonstrated its efficacy in detecting optical properties in samples and monitoring disease progression by reflecting molecular changes in biological metabolism [25–29]. Then, RS emerges as a suitable and effective tool throughout the entire clinical intervention process for breast cancer. Its distinctive features, such as the absence of a requirement for sample preparation, nondestructive impact on sample structure, ease of detection, and high resolution, make it instrumental in tasks ranging from distinguishing benign and malignant tissues to monitoring disease progression and guiding treatment plans. Nevertheless, the faint signal of Raman spectroscopy remains a significant impediment, constraining its clinical applications. To address diverse clinical requirements, various enhanced Raman spectroscopy variants have emerged [30, 31]. Table 2 outlines the technical characteristics and application scope of several widely utilized Raman spectroscopy derivative techniques in clinical detection.

Table 2. Summary of the technical characteristics and applications of Raman spectral derived techniques in breast cancer detection.

Raman derived technology	Basic principles	Technical features	Application examples in breast cancer
Surface enhanced Raman spectroscopy (SERS)	SERS can enhance the intensity of Raman signal through the interaction between molecules and rough metal surface (nanoparticles)	Nondestructive technique, minimal sample preparation, minimal interference.	Detection through analyzing liquid biopsy, tissue samples, exosome miRNAs and in vivo imaging [14, 32–36].
Spatially offset Raman spectroscopy (SORS)	SORS involves collecting Raman spectra at different spatial offsets from the point of laser excitation, and then mathematically analyze the extract information from different depths within the sample.	Noninvasive and in-depth analysis of complex samples.	Detection of breast microcalcifications buried in tissue at depths up to 20–40mm [37–40].
surface-enhanced spatially offset resonance Raman spectroscopy (SESORS)	SESORS combines the technical advantages of SORS with high probing depth and SERS with signal enhancement to achieve deeper and precise probing than conventional Raman spectroscopy.	Deep analysis capability, enhanced signal sensitivity, potential for Raman imaging	SESORRS imaging of tumor profile [41, 42], detection of tumor in vitro using hand-held SORS spectrometer [43, 44].
Transmission Raman spectroscopy (TRS)	The excitation and collection optics of the TRS are located on either side of the sample. By collecting the signal on the opposite side of the laser illumination, the Raman spectrum obtained from the transmitted light will be representative of the entire tissue within its detection depth range.	Nondestructive, Bulk analysis ability. fluorescence signal suppression ability.	The microcalcification components buried in 16 mm chicken breast tissue and 27 mm pig tissue were identified [40, 45].
Kerr-gated Raman spectroscopy	Kerr-gated Raman spectroscopy can separate biochemical information from different tissue depths by using excitation with a picosecond pulsed laser and collecting Raman scattered light at different time delays.	Deep analysis, fluorescence signal suppression ability	Detection of microcalcification foci in early breast cancer [38].
Near-infrared Fourier transform Raman spectroscopy (NIR FT-RS)	Nir spectroscopy and FT-Raman spectroscopy are combined to simultaneously obtain the relevant chemical composition information of the sample (from NIR) and the molecular vibration details (from Raman).	Fluorescence signal suppression ability, high spectral resolution,	Analysis of breast cancer tissue samples from in vivo and in vitro experiments [46–48].
Shell-isolated nanoparticle-enhanced Raman spectroscopy (SHINERS)	SHINERS is a new generation technology based on SERS, the enhanced Raman signal of the substance to be measured can be obtained by spreading the nanoparticles on the surface of the sample to be measured. Core-shell nanoparticles are prepared by coating the surface of Au or Ag nanoparticles with an extremely thin and dense inert shell layer (such as SiO2, Al2O3, etc.).	The signal interference caused by impurity adsorption of nanoparticles is greatly avoided, extended service life of nanoparticles, high detection sensitivity	Investigation of Raman characteristics of breast cancer in different pathologic types [37, 49, 50].

Raman spectroscopy in tissue biopsy of breast cancer

As the remarkable capability of RS in accurate biological information recognition, substantial efforts have been directed toward its development for the early diagnosis of breast cancer over the past decades [26, 51]. The onset of breast cancer is associated with significant alterations in the levels of biochemical molecules within breast tissue. Gene mutations and changes in gene expression impact the biochemical metabolism of proteins, lipids, and carbohydrates, leading to metabolic mechanism disorders that facilitate the gradual development of breast tissue into cancer [52–55]. RS plays a crucial role in identifying these biochemical changes in cancerous breast tissue, enabling early diagnosis and tumor grading of breast cancer [56–58].

Alfano et al. first applied RS to detect the spectral characteristics of benign and malignant breast tissues [59]. They conducted RS on 14 breast tissues, comprising three normal tissues, four benign lesions, and seven breast cancer tissues. Their findings revealed differences in the spectral characteristics of proteins and other components among these tissues. Similarly, Chowdary et al. explored Raman spectra in normal tissues, benign breast lesions, and malignant breast tissues [60]. The spectral results indicated that lipids in normal tissues predominantly exhibited distribution at 1078, 1267, 1301, 1440, 1654, and 1746 cm⁻¹, while proteins in benign and malignant lesions showed distribution at stronger amide I, red-shifted Delta CH2, broad and strong amide III, 1002, 1033, 1530, and 1556 cm⁻¹. By observing the characteristic peak distribution, they identified a key Raman spectral difference between benign and malignant lesions, specifically the presence of excess lipids in the spectra of malignant tissues (1082, 1301, 1440 cm⁻¹). The difference spectra of normal and pathological tissues are shown in Figure 3.

Figure 3. Difference spectra of normal, malignant, and benign breast tissue spectra: (a) malignant – normal; (b) benign – normal; (c) malignant – benign. Reprinted with permission from [60]. Copyright [2006] Wiley Periodicals, Inc.

Raman spectroscopy in progress of breast cancer

After the ability of RS to accurately distinguish benign and malignant tissues has been proved, a new research direction has emerged—exploring whether RS can be employed to monitor the changes in biochemical mechanisms as benign breast tissue progresses toward malignant transformation. Consequently, recent studies have delved into the molecular mechanisms elucidated by RS in the context of the transformation of benign breast tissue into cancer. For instance, Han et al. utilized Raman spectroscopy to detect the biochemical characteristics of normal breast tissue, atypical ductal hyperplasia (ADH), ductal carcinoma in situ (DCIS), and invasive ductal carcinoma (IDC) lesions. Employing leave-one-out cross-validation (LOOCV) and radial basis function (RBF), they constructed a support vector machine (SVM) diagnostic model [61]. Their findings indicated significant RS differences among normal tissues, ADH, DCIS, and IDC tissues. Notably, they observed that the protein level of ADH tissue was higher than that of normal tissue, while the lipid level was lower. Spectral differences among ADH, DCIS, and IDC were characterized by a red shift in the CH2 broad peak (1301 cm⁻¹), a carotenoid stretching vibration peak (1526 cm⁻¹), a relatively strong amide I band (1656 cm⁻¹), and a nucleic acid peak (882 cm⁻¹). The classifier employed was the SVM model, achieving an overall accuracy of 74.39%. Sensitivities for normal tissue, ADH, DCIS, and IDC were 62.5%, 50%, 90%, and 66.7%, respectively, with specificities of 100%, 100%, 66.7%, and 89.06%, respectively.

Currently, the primary traditional technique used in clinical practice to differentiate between different tissue types is tissue pathology diagnosis. This technique relies on observing histological differences such as cytologic monotony and uniformity, nuclear characteristics, and lesion severity to distinguish between these tissue types. According to a study by Elmore et al. [62], there is a discrepancy between the interpretation of pathological diagnosis by clinical pathologists and the reference diagnosis by experts, with an overall consistency of 75.3%. The consistency is highest for invasive carcinoma at 96%, while it is lower for ADH and DCIS, at 48% and 84% respectively. Therefore, the subjectivity in pathological interpretation among observers is considered one of the main challenges in tissue pathology diagnosis. On the other hand, pathological diagnosis can only identify breast tissue that has already undergone malignant transformation, lacking the ability to predict disease progression in breast tissue that has not yet developed cancer. Raman spectroscopy, however, has the capability to differentiate normal tissue, ADH, DCIS, and IDC at the molecular level. Thus, it can provide early indications of disease type and potential progression in breast tissue lesions. This information contributes to enhancing the reliability of current breast cancer diagnosis and treatment strategies. In summary, we believe that the application of Raman spectroscopy in breast cancer diagnosis can facilitate early clinical intervention for breast cancer, thereby providing patients with better treatment and prognosis.

Li et al. conducted a study on the RS differences among healthy breast tissues, solid papilloma (SPC), mucinous carcinoma (MC), ductal carcinoma in situ (DCIS), and invasive ductal carcinoma (IDC) in vitro [63]. They utilized principal components analysis (PCA) combined with linear discriminant analysis (LDA) to identify spectral differences. The results, obtained through a cross-validation method, demonstrated a 100% overall accuracy in distinguishing breast tissue types. This suggests that the integration of multivariate analysis with RS has the potential to enhance the diagnostic accuracy of breast cancer. Figure 4 shows the spectral intensities corresponding to the different biochemical components in HC, IDC, DCIS, SPC, and MC samples. In another study, Surmacki et al. employed confocal Raman imaging, RS, and infrared spectroscopy to detect the vibration characteristics of normal breast tissue and malignant breast tissue, including invasive ductal carcinoma and invasive lobular carcinoma [64]. They utilized K-means clustering, basis analysis, PCA, and Partial least squares Discriminant Analysis (PLS-DA) to identify Raman features in 164 samples from 82 patients. In RS, the regions corresponding to carotenoids, fatty acids, and proteins were identified, allowing for the differentiation of biochemical components between normal and cancerous tissues based on the intensity, frequency, and distribution of the average Raman spectrum. The discovery of this spectroscopic information provides valuable insights into the biochemical metabolic mechanisms underlying the malignant transformation of breast tissue. Such insights can guide the timing of treatment choices and contribute to the development of therapeutic interventions.

Figure 4. Bar graphs showing variations in the intensity of spectral peak corresponding to various biochemical components in HC, IDC, DCIS, SPC, and MC samples. The mean ± standard deviation is indicated. (a) Collagen (868 cm⁻¹), (b) lipid (1267 cm⁻¹), (c) lipid (1302 cm⁻¹), (d) lipid (1440 cm⁻¹), (e) protein (1608 cm⁻¹), (f) lipid (2890 cm⁻¹). Statistical significance was determined by one-way ANOVA followed by a Tukey’s HSD post-hoc test. Asterisks indicate levels of significance, * p < .05, ** p < .01. Reprinted with permission from [63]. Copyright [2021] MDPI.

With its robust capability to visualize biochemical metabolic information, RS exhibits significant potential for monitoring the progression of breast cancer. Sabtu et al. conducted a study using RS to investigate the metabolic changes associated with epithelial-mesenchymal transition (EMT) in breast tissue—an essential process in the progression and metastasis of breast cancer [65]. RS was applied to non-lesional, EMT, and non-EMT breast cancer samples from 23 patients. Results revealed substantial differences in spectral intensity within the 600–1800 cm⁻¹ band between the non-lesion group and the EMT group, as well as between the EMT group and the non-EMT group. Multivariate analysis, including PCA, independent component analysis (ICA), and the non-negative least squares method (NNLS), indicated significant differences in lipids, proteins, and nucleic acids between EMT and non-EMT tumors. In a separate study, Daniela et al. utilized Raman microscopy to analyze tissue microarrays of breast cancer biopsy specimens (n = 499) and normal breast specimens (n = 79). PCA and LDA were employed for feature extraction and verification [66]. The results demonstrated that peak intensities of carotenoids, β-carotene, and cholesterol were higher in normal breasts, while ceramide-related peaks were predominantly visible in breast cancer spectra.

The application of automated algorithms is inevitably accompanied by some limitations, such as oversimplification of data characteristics and incorrect assumptions about the relationship between variables, which may lead to misunderstanding of Raman spectroscopy data [10, 67]. Fortunately, due to the robust advancements in artificial intelligence (AI) technology in recent years, data processing approaches that integrate automated algorithm models with deep learning have demonstrated promising application potential in Raman spectroscopy detection of breast cancer [22, 68, 69]. Deep learning, as a machine learning method, automatically extracts learning features for advanced data analysis by simulating the neural network of the human brain [70–75]. Ma et al. explored the effectiveness of a one-dimensional convolutional neural network combined with RS for breast cancer diagnosis. RS was conducted on breast cancer samples from 20 patients, and classification was performed using a one-dimensional convolutional neural network, Fisher discriminant analysis (FDA), and SVM classifier. The one-dimensional convolutional neural network exhibited the best classification performance, achieving an overall recognition accuracy of 92% [76]. In a recent study, Shang et al. investigated the accuracy of breast cancer detection and recognition using two-dimensional convolutional neural networks combined with RS. The average recognition accuracy of this approach was 96.01%, surpassing the performance of one-dimensional convolutional neural networks. Their results demonstrated that the incorporation of deep learning algorithms is beneficial for improving the ability of RS to identify and automatically diagnose cancer [77]. Furthermore, the same research group evaluated the feasibility of employing deep learning algorithms in combination with fluorescence imaging and RS for multimodal breast cancer diagnosis. They applied a pseudo-colour enhancement algorithm and convolutional neural network to fluorescent image processing. Utilizing a partial least squares algorithm, they achieved a prediction accuracy of 100%. This study underscores the versatility of deep learning algorithms in various diagnostic techniques while ensuring high classification accuracy [78].

The advancement of portable Raman spectrometers has enabled rapid and noninvasive tumor diagnosis during the initial stages of assessment. Li et al. were among the pioneers to use a miniature laser Raman spectrometer for the analysis and detection of isolated breast tissues [79]. Their results indicated that, when combined with the adaptive weight k-local hyperplane (AWKH) method, the miniature Raman spectrometer achieved higher classification accuracy. Recently, Nicolson et al. delved into the clinical application potential of handheld Raman spectrometers. They utilized a handheld Raman spectrometer in conjunction with surface-enhanced spatially offset resonance Raman spectroscopy (SESORRS) for multiple imaging of breast cancer in in vivo tumor models. Three nanotags were discerned at a tissue depth of 10 mm, highlighting the capability of handheld Raman spectrometer in detecting and classifying multiple Raman fingerprints in vivo [43]. Another study by the same group employed SESORRS for 2D tissue imaging on a living tumor model using pig samples. They employed a red-shifted thiophene-based Raman reporter to detect pig tissues at a depth of up to 25 mm, constructing a false-colour 2D heat map to pinpoint the tumor model’s location. The portability and reliable detection performance of this technology make it promising for application in the noninvasive diagnosis of breast cancer in humans, with the potential for promotion in outpatient medical testing.

Raman spectroscopy in liquid biopsy of breast cancer

Liquid biopsy is a diagnostic technique that analyzes tumor-derived materials such as circulating tumor cells, circulating tumor DNA, tumor extracellular vesicles, etc., in body fluids for disease diagnosis. This approach offers advantages such as noninvasiveness, robust early diagnostic capabilities, and dynamic monitoring [51, 80–82]. When combined with RS analysis, liquid biopsy has the potential to overcome the heterogeneity of breast cancer tumor cells and comprehensively reflect the biochemical information encompassed by tumor tissue [11, 13, 83]. The combined strategy of liquid biopsy with RS, leveraging its minimally invasive nature and capacity for detailed global information recognition, holds significant promise for applications in the early diagnosis, treatment monitoring, and surgical evaluation of breast cancer.

The biomolecular information derived from RS based on blood samples proves valuable for screening and diagnosing breast cancer, identifying breast cancer staging, and evaluating surgical outcomes. Nargis et al. utilized RS to characterize plasma samples from patients at different stages of breast cancer. Their results demonstrated the capability to distinguish all stages of breast cancer progression from normal breast samples. The use of PCA facilitated the identification of biochemical components in the second and third stages of breast cancer, while the Raman spectroscopic biochemical information characterization of the fourth stage exhibited significant differences between different patients [13]. In another study, Bilal et al. proposed a breast cancer screening approach based on Raman detection of whole blood samples. The samples were obtained from both breast cancer-positive patients and healthy individuals. A multivariate Partial Least Squares (PLS) regression model was established to assess molecular structure changes associated with disease progression. The results were confirmed to be applicable to clinical practice [84]. These findings underscore the potential of RS in blood-based diagnostics for breast cancer screening and assessment.

While traditional RS and liquid biopsy have demonstrated reliable detection performance in breast cancer diagnosis, the appeal of Raman variant technology with signal enhancement capabilities is noteworthy for researchers. Nargis et al. employed surface-enhanced Raman spectroscopy (SERS) and conventional RS to classify breast cancer in serum samples from breast cancer patients and healthy individuals. The results indicated that SERS exhibited higher sensitivity and specificity (90%, 98.4%) compared to RS (88.2%, 97.7%), suggesting that SERS is a preferable choice for serum sample detection. The researchers established spectral features that can serve as markers for the diagnosis and classification of breast cancer. The diagnostic performance of SERS and RS was compared using partial least squares discriminant analysis (PLS-DA) [67]. In another study, Lin et al. utilized SERS to analyze peripheral blood samples from breast cancer patients before and after surgery, as well as healthy volunteers. The results from multivariate diagnostic algorithms demonstrated that the accuracy of this method in identifying blood samples before and after breast cancer surgery and distinguishing preoperative samples from normal populations reached 95% and 100%, respectively. These findings suggest that this method can provide reliable blood analysis for the surgical evaluation and tumor screening of breast cancer [85]. The application of surface-enhanced RS introduces enhanced signal capabilities, enhancing its potential in the realm of breast cancer diagnostics.

The auxiliary role of RS based on serum exosomes in the diagnosis and staging of breast cancer has been explored. Exosomes are extracellular lipid vesicles found in all body fluids, participating in various pathophysiological processes by delivering bioactive molecules [74, 86]. They can serve as noninvasive biomarkers for the early diagnosis and monitoring of cancer.

Xie et al. evaluated the role of SERS analysis of serum exosomes in the diagnosis and surgical evaluation of breast cancer [32]. Results from a deep learning algorithm demonstrated 100% prediction accuracy in patients with different subtypes of breast cancer who did not undergo surgery. Additionally, by combining PCA similarity analysis, it was shown that this method could evaluate the surgical efficacy of different molecular sub-types of breast cancer. In the study by Diao et al., ‘hot spot’ rich 3D plasmonic AuNPs nanomembranes were used as substrates to study the precise fuzzy recognition ability of label-free SERS based on machine learning for exosomes in serum samples. Their results indicated that the prediction accuracy of exosomes derived from different cell lines based on machine learning algorithms could reach 91.1%, and the prediction accuracy of clinical samples using the SERS spectral training model of cell-derived exosomes could reach 93.3%. Figure 5 shows the process of label-free SERS analysis method driven by machine learning, typical spectra and evaluation of recognition accuracy [33]. Similar to exosome detection, the comprehensive analysis of exosome proteins in clinical samples has been a challenging problem in the personalized diagnosis and treatment of breast cancer. Su et al. recently developed a paper-based SERS-vertical flow biosensor for multiplex quantitative analysis of exosome proteins in clinical serum samples [34]. MUC1, HER2, and CEA from different subtypes of breast cancer cells could be quantitatively analyzed simultaneously. These studies highlight that RS of serum sample exosomes is becoming an integral com-ponent of noninvasive breast cancer typing and longitudinal treatment monitoring.

Figure 5. (a) Machine learning-based SERS detection for classification and prediction of different cell line-derived exosomes. (b) The averaged SERS spectra collected from three cell lines (H8, MCF-7, and HeLa cells)-derived exosomes with trilayered AuNPs-NMs as substrates. (c) PCA-LDA score plot of SERS signals of exosomes derived from H8 cells (green), HeLa cells (red), and MCF7 cells (orange), respectively. (d) LDA models developed for predicting exosomes derived from three different cell lines. (e) ROC curves of the LDA model for predicting exosomes derived from three different cell lines. Reprinted with permission from [33]. Copyright [2023] American Chemical Society.

In addition to common blood components, other body fluid components, such as saliva, have also become the focus of research in RS. Biomarkers present in saliva are also widely distributed in body fluids and participate in many metabolic processes, holding great potential for assessing the degree of malignancy and metastasis of tumors. Saliva sampling, being completely noninvasive and operationally simple, has advantages over blood sampling. Feng et al. purified salivary proteins and assessed the feasibility of using SERS analysis of salivary proteins to detect benign and malignant breast tumors [87]. The spectral data of saliva proteins were analyzed by Partial Least Squares Discriminant Analysis (PLSDA), resulting in diagnostic sensitivities of 75.75%, 72.73%, and 74.19%, and specificities of 93.75%, 81.25%, and 86.36%, respectively. Additionally, researchers evaluated the effectiveness of sialic acid levels as a biomarker for breast cancer. They employed SERS to determine the correlation between the level of sialic acid and the positive detection of breast cancer, comparing it with biopsy results. The method demonstrated a sensitivity of 80% and specificity of 93% [88]. In another study, confocal SERS imaging was employed to analyze the expression and distribution of sialic acid in breast tissue. These studies have validated the effectiveness of sialic acid as a biomarker for breast cancer detection. Detecting sialic acid expression and distribution is beneficial for early diagnosis and real-time, in vivo studies of breast cancer [89].

To identify the optimal source of liquid biopsy fluids, urine and tears were also included in the study of RS for breast cancer. In a 2015 study, the feasibility of urine-based RS as an adjunct diagnostic method for breast cancer was evaluated [90]. Untreated urine and concentrated urine from healthy rats and breast cancer rats were analyzed using RS. Spectral processing involved PCA and Principal Component-Linear Discriminant Analy-sis (PC-LDA). The classification efficiency of untreated urine from healthy rats and breast cancer rats was 80% and 72%, respectively, and for concentrated urine, it was 78% and 91%, respectively. Furthermore, the feasibility of urine-based RS for distinguishing breast cancer progression was investigated, revealing a sensitivity of 72.5% and specificity of 83% for cancerous urine classification in the early stage of cancer. In a recent study, urine components, specifically nucleosides, were explored as potential biomarkers for breast cancer screening [91]. Human urine nucleosides were separated and purified using affinity chromatography, and their biomolecular characteristics were characterized by SERS. The combination of PCA and LDA for spectral classification of breast cancer resulted in a sensitivity of 76.7% and specificity of 87.5%.

Tears, as a surprising source of liquid biopsy fluids, have also been shown to play a significant role in cancer diagnosis. A portable Raman system was utilized to detect tear samples, and PC-LDA was employed to identify RS components. This method demonstrated a sensitivity of 92% and specificity of 100% [92]. The exploration of alternative liquid biopsy sources, such as urine and tears, underscores the versatility and potential of RS in the diagnosis of breast cancer.

Raman spectroscopy in single-cell detection of breast cancer

Single Cell Raman Spectrum (SCRS) offers the capability to unveil the intrinsic biochemical molecular composition of cells without the need for labeling or invasiveness. It allows the visualization and analysis of single cells at sub-micron spatial resolution [93–96]. Cells serve as the fundamental units of the structure and function of organisms. The cell biomolecular fingerprints provided by single-cell Raman spectroscopy, including proteins, carbohydrates, nucleic acids, lipids, and carotenoids, offer detailed information for interpreting the physiological state and metabolic activities of breast tissues [93, 94]. Recently, SCRS has found extensive applications in the diagnosis of breast cancer. These applications include cell identification and cell type identification, Raman spectroscopy imaging, characterization of cell metabolism, and clinical diagnosis. The advancements in Raman technology and software capabilities have facilitated the rapid acquisition of biological information in milliseconds, allowing the analysis of tens of thousands of data points in minutes. This technology can monitor tissue metabolic activities while simultaneously detecting several biological components [97]. The ability of SCRS to provide detailed molecular information at the single-cell level has proven valuable in advancing our understanding of breast cancer and improving diagnostic capabilities.

As a spectral detection technology, RS offers the ability to evaluate biomarkers of tumor cells without the need for labeling. Manciu et al. conducted an assessment of the biological activity of epidermal growth factor receptors on the surface of breast cancer cells using RS [97]. Their findings revealed a significant difference in Raman labeling between non-tumorigenic (MCF-10A) and tumorigenic (MCF-7) breast epithelial cells. Furthermore, they compared the Raman profile and image display of specimens with or without epidermal growth factor, noting important differences in the fingerprint regions of lipid, protein, and nucleic acid vibrations. The generation of new Raman features was found to be related to the presence of epidermal growth factor. Beyond identifying biomarkers on the surface of cancer cells, various research groups have explored the isolation and detection of cancer cells. Yarbakht et al. proposed a novel strategy for the selective separation and detection of breast cancer cell lines (MCF-7 and BT-20) based on SERS [98]. They developed a simplified scheme based on cell-aptamer interaction. In this method, core-shell (Au@Fe₃O₄) nanoparticles (CSNs) were functionalized by mucin 1 (MUC1)-specific aptamer (Apt1), and the cells were separated through the interaction between Apt1 and the tumor cell surface-overexpressed protein (MUC1). Simultaneously, Apt1 was coupled to the surface of bovine serum albumin (BSA)-coated 4-mercapto pyridine (4-Mpy)-functionalized gold nanoparticles to synthesize SERS nanotags for the detection of isolated cells.

Numerous studies have delved into the application of RS for cell analysis, and the results obtained suggest that SCRS has significant clinical potential and could, to some extent, replace existing detection techniques. Rotter et al. argued that nanoparticle-based SERS provides more comprehensive molecular imaging capabilities and can potentially replace current clinical immunohistochemical analysis [99]. Their study explored the potential of antibody-conjugated Surface-Enhanced Resonance Raman Scattering nanoparticles (SERRS-NPs) to depict and quantify the expression of high and low tumor surface markers in the brain and peripheral environment. The Raman signal intensities of surface markers such as epidermal growth factor receptor (EGFR) and human epidermal growth factor receptor 2 (HER2) were compared within and between tumors. SERRS-NPs demonstrated the ability to provide clear Raman spectra reflecting the distribution of targeted surface markers compared to immunohistochemistry. The signal intensity of SERRS-NP accurately distinguished high and low expression of surface markers between tumors and different regions of the tumor. In addition, Wang et al. investigated the potential substitution of RS for flow cytometry. They utilized a portable SERS detection method for classifying the surface immunophenotypes of multiple cells [100]. Customizable SERS nanotags were synthesized and functionalized for cell labeling. Two different cell models, namely red blood cells and breast cancer cells, were employed to evaluate and verify the analytical performance of portable SERS immunophenotyping. Their results indicated that the coefficient of variation of portable SERS immunoassay for detecting red blood cells and breast cancer cells was comparable to that of flow cytometry, ranging from 21.89% to 23.33% for red blood cells and 6.88% to 17.32% for breast cancer cells. Figure 6 shows the flow diagram of the portable SERS immunophenotyping assay they used and the Raman spectra to analyze the expression of different cell surface markers.

Figure 6. (a) Schematic of the portable SERS immunophenotyping assay for characterizing cell surface proteomes. (b) Tetraplex immunophenotyping of different marker expressions on the cell surface of SKBR3, MCF7, and MDA-MB-231, respectively. The gray curve corresponds to the IgG control. Reprinted with permission from [100]. Copyright [2022] American Chemical Society.

The application of SCRS in breast cancer research is proving invaluable in overcoming tumor heterogeneity. SCRS enables a detailed understanding of the metabolic characteristics of individual tumor cells, aiding in the early detection of breast cancer. Zhao et al. introduced the concept of Metabolically Active Phenotype (MAP), utilizing confocal Raman microscopy for high-throughput quantitative measurement of lipid and protein synthesis activity at the single-cell level [101]. This approach allows the classification of different breast cancer cell subtypes by exploring metabolic differences between single cells. The study also evaluated the metabolic heterogeneity of different drugs in the treatment of breast cancer at the single-cell level. In another study by Zhang et al., the regulation mechanism of cytokines in the tumor microenvironment was explored using noninvasive RS based on machine learning [102]. Single-cell Raman microscopy was employed to analyze changes in molecular composition and biochemical characteristics induced by granulocyte colony-stimulating factor (G-CSF) at the cellular level. PCA was used to identify Raman bands with the most significant changes in the control group and the G-CSF group. The RS results exhibited a high degree of correlation with the verification of tumor-promoting cytokines and gene analysis. This research demonstrates the potential of SCRS in unraveling the complexities of breast cancer at the single-cell level, contributing to a more precise understanding of tumor behavior and response to treatment.

The investigation of redox state changes and biological mechanisms of cytochrome C in cancer cells, particularly using RS, has proven to be beneficial for noninvasive grading and differential diagnosis of breast cancer. Abramczyk et al. conducted a study utilizing RS to detect the redox state of mitochondrial cytochromes in breast tissues at different wavelengths (532, 633, and 785 nm), and Raman imaging was employed to visualize the cytochromes in the breast [103]. By analyzing cytochrome C vibrations at specific wavenumbers (750, 1126, 1337, and 1584 cm⁻¹) as a function of malignancy grade, they observed that the concentration of reduced cytochrome C in breast cancer became abnormally high and correlated with the grade of cancer. Building on this research, the same group visualized the distribution of cytochromes in the main organelles of cancer cells. By comparing the concentrations of cytochrome C from the nucleus, mitochondria, cytoplasm, lipid droplets, and cell membranes, they demonstrated that the 1584 cm⁻¹ ferric low-spin heme in cytochrome C can serve as a sensitive indicator for evaluating cancer invasiveness [104]. This innovative approach provides valuable insights into the cellular and molecular characteristics associated with breast cancer, offering potential applications for noninvasive grading and diagnosis.

Raman spectroscopy in microcalcifications of breast cancer

Microcalcification in breast cancer involves the deposition of calcium salts in breast tissue, often appearing as tiny white spots or spots in mammography. While not an early marker of breast cancer, the detection of microcalcifications is a crucial clinical indicator for evaluating breast cancer progression [37, 105]. Currently, breast biopsy is the primary method for determining the nature of microcalcifications, with clinicians classifying them based on appearance and morphological characteristics. Two main types of microcalcifications, Type I (benign, composed of calcium oxalate dihydrate) and Type II (composed of hydroxyapatite), can be observed in both benign and malignant lesions [106]. The structures of calcium oxalate and hydroxyapatite effectively scatter photons under laser irradiation, enabling the classification of microcalcifications by identifying Raman characteristic peaks. RS analysis of needle aspiration biopsy tissue has shown promise in reducing the occurrence of non-diagnostic and false-negative biopsies. Saha et al. conducted an analysis of 159 sites of breast stereotactic needle biopsy using RS, including 54 normal sites, 75 microcalcified lesions, and 30 non-microcalcified lesions. The positive predictive value of microcalcification detected by the Raman technique was reported to be 97%. Building on this study, the researchers developed a technique capable of identifying microcalcification status in real-time during stereotactic core needle biopsy, facilitating the diagnosis of breast lesions and distinguishing normal tissues, fibrocystic changes (FCC), fibroadenoma, and breast cancer based on microcalcification detection [107]. Figure 7 shows typical Raman spectra and histopathological observations of breast lesions (fibrocystic changes) with type I and type II microcalcification. This real-time identification has significant implications for improving the accuracy and efficiency of breast cancer diagnosis during biopsy procedures [108].

Figure 7. Typical Raman spectra and histopathology of breast lesions (fibrocystic change) with type I and II microcalcifications. The Raman spectrum of the breast lesion with type I microcalcifications in (a) shows prominent bands at 912 cm⁻¹ and 1477 cm⁻¹ (arrows) characteristic of calcium oxalate; the calcium oxalate crystals comprising the type I microcalcifications (b) do not bind H&E (left panel) and appear as colorless crystals (arrows) that are birefringent when viewed under polarized light (right panel). In contrast, the Raman spectrum of the breast lesion with type II microcalcifications in (c) shows a prominent band at 960 cm⁻¹ (arrow) characteristic of calcium hydroxyapatite; the calcium hydroxyapatite rich type II microcalcifications appear as basophilic concretions on the H&E stain (d) and are non-birefringent. Reprinted with permission from [107]. Copyright [2011] Optica publishing group.

The real-time detection of tissues in vivo using RS holds great promise for improving diagnostic procedures. Stone et al. conducted groundbreaking research by analyzing human breast tissue using Kerr-gated Raman spectroscopy. This technique had previously demonstrated its ability to detect Raman characteristics of calcium phosphate and calcium oxalate in chicken breast and adipose tissue at a depth of 0.96 mm. The study highlighted that the measurement depth of calcified components achieved with spatially shifted Raman spectroscopy was one to two orders of magnitude greater than the detection depth of traditional Raman methods [38, 39]. This enhancement in measurement depth is crucial for effectively probing tissues in vivo. In another study, transmission Raman spectroscopy was employed to perform RS on calcified substances buried in 16 mm of chicken breast tissue. While this technique could assess calcified substances, it was noted that RS is not sensitive to the geometry of these substances. As a result, the researchers suggested that this method might serve as a complementary tool to existing mammography or ultrasonography, providing additional insights [40]. The same research group applied transmission RS to detect 2.7 cm microcalcifications in pig tissues, demonstrating that the technology has reached the detection depth and sensitivity required for clinical applications [45]. These advancements in real-time, in vivo RS have the potential to significantly impact the field of diagnostic medicine, particularly in breast cancer detection and evaluation.

In recent years, researchers have utilized RS to analyze and detect the components of microcalcifications, providing valuable insights into breast cancer diagnosis. Shell isolated nanoparticles enhanced Raman spectroscopy (SHINERS) has emerged as a technique to enhance the sensitivity of traditional Raman methods. Studies using SHINERS have demonstrated its effectiveness in identifying the characteristics of type II microcalcifications in various breast tissues, including fibroadenoma (FB), atypical ductal hyperplasia (ADH), and ductal carcinoma in situ (DCIS) [37]. Vanna et al. observed that microcalcifications in malignant samples tend to be more uniform and exhibit a higher number of crystals compared to benign samples. Moreover, there were statistically significant differences in Raman characteristics between the phosphate band and the carbonate band of benign and malignant components [109]. These findings highlight the potential of RS, especially when enhanced by techniques like SHINERS, to provide detailed information on microcalcifications in breast tissues, aiding in the diagnosis and characterization of breast cancer.

Application of Raman spectroscopy in the treatment of breast cancer

Raman spectroscopy in surgeries of breast cancer

Breast-conserving surgery is a preferred treatment for early breast cancer, and the accurate detection of cancer cells at the tumor margin is crucial for its success. Traditional methods, such as rapid frozen section analysis, can be time-consuming and less accurate. In this context, an automated RS three-dimensional scanner is being explored as a potential tool for evaluating the margin of breast cancer during surgery within a clinically feasible time frame [110]. This technique aims to reconstruct a three-dimensional image model of the tumor edge tissue, providing surgeons with accurate anatomical information for re-resection areas. Additionally, portable near-infrared RS systems have demonstrated effectiveness in identifying cancer cell residues in breast surgical specimens [111]. These systems leverage specific spectral bands, typically using laser wavelengths of 1064 nm or 785 nm, to distinguish between healthy and malignant tissues. The application of such technologies holds promise for precise intraoperative detection, offering a cost-effective and accessible solution for patients. This advancement may contribute to improved outcomes in breast-conserving surgery.

Multimodal spectral imaging, combining tissue autofluorescence and RS, has emerged as a reliable tool to enhance the accuracy of intraoperative surgical margin evaluation, particularly for detecting residual tumors at the margin of breast tissue [112]. In a study involving 121 samples from 107 patients, this multimodal imaging technique demonstrated a sensitivity of 95% and specificity of 82%. Furthermore, a rapid intraoperative evaluation technique has been developed, integrating autofluorescence spectral imaging and RS to measure the nonfat area of breast cancer tissue. The fully automated and faster scanning recognition of this technique allows for the efficient completion of breast cancer specimen detection within 20 min, potentially reducing the exposure time of breast tissue during surgery [113]. Figure 8 illustrates the detection of breast conserving surgery (BCS) specimens with positive surgical margins using multimodal spectral histopathology (MSH) measurements [112]. These advancements in multimodal imaging contribute to more accurate and timely intraoperative assessments of surgical margins in breast cancer procedures.

Figure 8. Examples of multi-modal spectral histopathology (MSH) measurements of whole breast conserving surgery (BCS) specimens with positive margins confirmed by histopathological assessment. The surface measured by MSH is facing downward in the specimen images. MSH detected tumor on the surface of all specimens in 12–24 min. (a–c) invasive carcinoma (IC); (d,e) ductal carcinoma in situ (DCIS). Reprinted with permission from [112]. Copyright [2018] Springer nature.

Wen et al. have made significant strides in addressing the challenge of residual small tumors during surgery. They developed a nanoprobe equipped with optical imaging, SERS, and photothermal tumor ablation to target and eradicate microtumors in the surgical area [114]. This innovative technique involves image-guided surgery for tumor removal and the subsequent elimination of residual microtumors through photothermal ablation. The nanoprobe’s capabilities allow for precise tumor resection and effective removal of small tumors. Experimental results demonstrated that this technology achieved complete eradication of microtumors, resulting in a 100% tumor-free survivability. The development of such technology holds promise for significantly improving the outcomes of cancer removal surgeries.

The application of RS in axillary lymph node surgery has become a focal point, particularly concerning sentinel lymph nodes, which play a crucial role in breast cancer diagnosis, treatment planning, and prognosis evaluation. The accurate assessment of intraoperative sentinel lymph node status is pivotal for surgical decision-making and postoperative patient management [115]. The utilization of RS for the rapid and accurate evaluation of untreated lymph node tissues during surgery has gained attention. In 2003, Smith et al. pioneered the application of RS in assessing axillary lymph nodes in breast cancer patients [116]. Through the analysis of Raman spectra and false-color spectral images, a comparative study with standard tissue pathological sections could be conducted. RS probes have seen rapid development due to their reliable sensitivity and multi-level analysis capabilities. Recent studies have demonstrated that Raman spectroscopic probes not only provide reliable detection performance but also offer the advantage of completing evaluations within the intraoperative timeframe [115, 117]. In 2010, Horsnell et al. assessed the ability of commercially available Raman spectroscopic probes to detect axillary lymph nodes, achieving a high sensitivity of 92% and specificity of 100% in distinguishing normal and metastatic lymph nodes. Further advancements in the sampling method of Raman probes in 2012 showed that the sensitivity in distinguishing positive and negative lymph nodes was 81%, with a specificity of 97%. Comparative analysis with histopathological results revealed accuracies of 89% and 91% for the 5-sampling points group and the 10-sampling points group, respectively [117]. These findings underscore the potential of RS in enhancing the accuracy of assessing axillary lymph nodes during breast cancer surgery.

Raman spectroscopy in chemotherapy

Chemotherapy is a critical component of breast cancer treatment and is frequently employed as an adjuvant therapy before or after surgery. The goal of chemotherapy in breast cancer treatment is to reduce tumor volume, eliminate residual cancer cells, enhance the effectiveness of surgery, and diminish the risk of tumor recurrence and metastasis [118, 119]. Additionally, for cases where breast cancer has spread to other parts of the body, different chemotherapy regimens can be selected based on the specific subtype of breast cancer. This personalized approach aims to control and slow down the growth rate of the tumor, thereby extending the patient’s life expectancy [120, 121].

RS plays a crucial role in supporting traditional diagnostic methods and evaluating the efficacy of breast cancer chemotherapy. It is utilized for various purposes, including aiding in the diagnosis and classification of tumors, monitoring treatment progress, offering feedback on treatment outcomes, and contributing to the development and application of new drugs [11, 13, 37–39, 68, 107, 122–124]. Detailed information on the diagnosis and classification of breast cancer can be found in the third part of the text. RS can globally monitor the metabolic changes at the cellular and molecular levels caused by chemotherapy drugs in the treatment of breast cancer [11, 125–128]. The in-situ monitoring function of RS can monitor the time-dependent intracellular release of anticancer drugs without labeling [127]. In drug delivery systems, RS is used in combination with specifically modified nanorods to detect and monitor the distribution of chemotherapeutic drugs within cells. This enables researchers and clinicians to gain insights into drug delivery efficiency and optimize treatment strategies [126]. Specific cell markers, such as human epidermal growth factor receptor-2 (HER2), are crucial for understanding the pathogenesis and evaluating the therapeutic efficacy of breast cancer. A recent study introduced a machine learning-driven surface-enhanced RS detection strategy for label-free detection and stoichiometric analysis of intracellular HER2. This approach allows for the longitudinal monitoring of the therapeutic effect of HER2 at the cellular level, contributing to effective therapeutic monitoring of diseases [128]. Figure 9 illustrates the dynamic SERS spectrum analysis during breast cancer chemotherapy. Overall, RS serves as a powerful tool in the comprehensive evaluation and monitoring of breast cancer chemotherapy, providing valuable information for both research purposes and clinical applications.

Figure 9. Dynamic SERS spectroscopic analysis during the therapeutic treatment with HApt. (a) SERS spectra and (b) corresponding violin plot of the cHER2 expression of SKBR-3 cells before (control) and after treatment with HApt for 24, 48, and 96 h. Note: the shadow in each SERS dataset represents 1 s.d. Reprinted with permission from [128]. Copyright [2022] American Chemical Society.

Evaluating the efficacy of breast cancer treatment is crucial for informing clinical decision-making, and RS plays a role in studying drug responses to aid in personalized treatment decisions. Nam et al. utilized label-free SERS with plasmonic nanostructures to investigate the drug response of cancer cells under the influence of different doses of chemotherapeutic drugs [129]. This research aims to assist clinicians in selecting individualized drug treatment plans that are more tailored to each patient’s needs. While there is still limited research comparing the efficacy of different drugs in breast cancer treatment using RS, it is acknowledged that the heterogeneity of the disease and the complexity of tumor classification pose challenges. However, efforts have been made to address this gap. Zhang et al. developed a microfluidic chip designed to overcome background noise interference in Raman measurements, enabling the differentiation of cancer cells treated with the chemotherapeutic drug DOX from non-DOX-treated cells [130]. This microfluidic chip demonstrates good biocompatibility and maintains the biomechanical properties of cells, providing a noninvasive detection tool for Raman measurements. This technology has the potential to aid in screening anticancer drugs in in vitro experiments.

With the continuous advancement of the concept of precision medicine, many efforts have been devoted to the study of the combination of targeted drug delivery systems and RS [114, 131, 132]. Varzandeh et al. contributed to the field by developing a targeted drug delivery system based on silver-coated gold nanorods [132]. This system was designed to target epithelial cell adhesion, particularly for the treatment of triple-negative breast cancer. The study demonstrated that cancer cells treated with this system received significantly higher drug doses compared to the control group, showcasing the potential for enhanced drug absorption rates [133]. They have developed a drug delivery system based on silver-coated gold nanorods that target epithelial cell adhesion for targeted therapy of triple-negative breast cancer. Flow cytometry data showed that cancer cells received two orders of magnitude higher drug doses than the control group. In addition to improving the absorption rate of drugs, targeted drug nano-delivery systems with multiple functions have also been developed [126]. Nima et al. designed a multi-layer silver-modified gold nanorod for the delivery of chemotherapy drugs for breast cancer (MCF7). The composed nanoparticles have unique spectral characteristics in SERS, which can be used to detect the metabolic changes of cells and monitor the distribution of chemotherapeutic drugs in cells.

Raman spectroscopy in radiotherapy

Similar to chemotherapy drugs, radiotherapy is often used as a preoperative and postoperative adjuvant therapy for breast cancer to reduce the risk of local recurrence, reduce tumor volume, and control distant metastasis [134]. In the radiotherapy of breast cancer, RS can not only provide clinicians with information about tissues and biomolecules by detecting biochemical changes in tumor tissues but also help doctors make more accurate governance plans by monitoring radiation exposure and possible side effects of normal tissues [134–136].

RS has been proven to reveal the differences in biochemical changes between different breast cancer cell types and normal tissues. Meksiarun et al. studied the changes in biochemical metabolism in breast cancer HER2 subtype, Ki67 subtype, and normal breast epithelial cell line MCF10A under different radiation doses (0–50 Gray) [137]. The RS results show that there are unique Raman spectral features that can be identified between breast cancer cell lines, and the effects of radiation on MCF10A can be distinguished in different breast cancer cell lines.

Recently, many studies have used RS to investigate the radiosensitivity of normal breast cells under different doses of irradiation. Lasalvia et al. found that even at a dose of 0.5 Gray, the biological effects of proton irradiation can be detected in RS [138]. They showed that the damage of genetic material will increase with the increase of radiation time, and most cells have irreparable DNA/RNA structural damage. In another similar study, they showed that some Raman characteristic peaks associated with genetic materials showed dose-related systemic inhibition. At the same time, they proposed that the peak intensity of 784 cm⁻¹ in RS can be used as a spectral marker for radiation damage of genetic materials. [136]. In addition to the study of the radiosensitivity of normal breast cells, some scholars have investigated the radio resistance of cancer cells. Tipatet et al. used RS and machine learning to characterize the biochemical changes caused by acquired radiation resistance of breast cancer cells [135]. Their research method showed 100% accuracy in distinguishing cells which have acquired radioresistance from wild-type cells. RS studies have shown that anti-radiation cancer cells contain fewer lipids and proteins than ordinary cancer cells. Figure 10 shows the average Raman spectra of wild-type and radioresistant cell lines. The findings of these studies indicate that different doses of radiotherapy have different biological metabolic effects on the treatment of breast cancer. The information on lesions and tissue characteristics provided by RS can help doctors clarify the patient’s disease progression and develop a more appropriate treatment plan.

Figure 10. Average Raman spectra (offset for clarity) for wild-type and radioresistant cell lines. The difference spectrum is created by subtracting the average spectrum of the (WT) parental cells from the average spectrum of the radioresistant (RR) cells. Reprinted with permission from [135]. Copyright [2021] Royal Society of Chemistry.

Current challenges and future prospects

The technical characteristics of RS, such as fast, high sensitivity, and high specificity, make it have rich potential application directions in the clinical diagnosis and treatment of breast cancer. However, it still needs to overcome some obstacles to be widely used in clinical diagnosis and treatment activities.

The primary limitation of RS research is the lack of uniform standards, which affects the repeatability of research methods and the reliability of results. In the RS study, there are differences in the processing and preparation methods of specimens among different laboratories. This includes sample curing, cutting, dyeing, removal of autofluorescence, reduction of scattering, and so on. The lack of uniform specimen processing standards has led to the incomparability of research results between laboratories [139–145]. There is also a lack of uniform standards for data analysis and interpretation, which is mainly manifested in the fact that different laboratories may use different algorithms, data processing methods, and spectral line fitting techniques [83, 122–124, 144, 146–153]. This makes the interpretation of the data and the credibility of the results different, and it is not easy to compare between different studies. In addition, laboratory environmental conditions and different settings of RS instruments may lead to inconsistent data. Therefore, there is an urgent need to develop uniform and detailed standards and guidelines to ensure the repeatability of research and the comparability of data.

The weak biological information signal and the susceptibility to autofluorescence signal also pose a challenge to the application of RS. The weak scattered light intensity in the Raman effect and the autofluorescence signal that overlaps with the Raman signal led to the masking or interference of the Raman signal [154]. Factors such as light source fluctuations, optical path interference, and mechanical vibration in the experimental environment may also introduce additional signal interference. These interferences will affect the accuracy of RS detection and the authenticity of interpretation and will reduce the credibility of experimental data. To deal with the above problems, research investment in RS should be increased to reduce autofluorescence interference and improve signal-to-noise ratio. In addition, in-depth exploration of data processing technology and advanced data analysis methods will help to improve the accuracy of RS. At the same time, researchers should also carefully plan experimental design and carefully select data processing methods to minimize the impact of interference signals on the results.

The ease of use and convenience of RS largely determine whether it can be truly used in clinical practice. Raman spectroscopy detection and interpretation of data results require medical personnel to have a certain knowledge reserve and technical ability, which is complex and laborious for clinicians and even interferes with the standard workflow. It is worth noting that the introduction of machine learning and AI technology seems to have a positive effect. They can achieve automated integrated analysis of large-scale data, and quickly identify and classify features that may be hidden in complex spectral lines, thereby saving time and human resources. In addition, these technologies can also exert their powerful pattern recognition capabilities to help medical professionals more accurately distinguish different cases and diagnose the type and severity of breast cancer [22, 59, 155–158]. The introduction of machine learning and AI provides a new prospect for RS detection technology for breast cancer, which is expected to achieve higher accuracy and efficiency in clinical diagnosis, and further promote the application of RS in clinical diagnosis and treatment activities. For portability, the existing handheld Raman spectrometer shows good application advantages, such as a fast result detection cycle, multiplexing ability, and small space occupation. It is worth mentioning that the portable Raman spectrometer with a smaller volume and lighter weight has shown excellent performance in the detection of tumor models in vivo [43, 44]. Some research results show that it has the potential to replace traditional analytical techniques [100]. This indicates that the portable Raman spectrometer is expected to achieve a wider range of applications in clinical medical and biomedical research, especially in the field of tumor diagnosis.

Disclosure statement

The authors report there are no competing interests to declare.

Additional information

Funding

This work was supported by the Department of Science and Technology of Jilin Province under Grant [number 20230401091YY].