Raman Spectroscopic Signatures of Hepatic Carcinoma: Progress and Future Prospect

TL;DR

Raman spectroscopy is being explored as a non-invasive tool for detecting liver cancer by analyzing molecular changes in biological samples.

Contribution

The paper reviews recent advances in Raman spectroscopy for liver cancer diagnosis and discusses future directions for clinical application.

Findings

Raman spectroscopy can distinguish malignant from non-malignant liver conditions using biochemical signatures in biofluids and tissues.

Machine learning and chemometric methods have enhanced the diagnostic accuracy of Raman spectroscopy for liver cancer detection.

Complementary techniques like SERS and ROA are expanding the clinical potential of Raman-based diagnostics.

Abstract

Liver cancer continues to be a predominant cause of cancer-related mortality globally, primarily attributable to late diagnosis and a scarcity of dependable biomarkers for early identification. Raman spectroscopy has emerged as a valuable analytical instrument for liver cancer detection, providing rapid, label-free, and non-destructive molecular profiling of biological specimens. Raman-based methodologies can discern malignant from non-malignant conditions by analyzing small biochemical alterations in biofluids, including blood, urine, and exosomes, as well as in liver tissue, yielding unique spectrum fingerprints. Progress in chemometric analysis, including machine learning models and multivariate statistical methods, has significantly improved the diagnostic precision of Raman spectroscopy, attaining elevated sensitivity and specificity across numerous studies. Furthermore, the integration of complementary techniques, such as surface-enhanced Raman spectroscopy (SERS) and Raman optical activity (ROA) has broadened its prospects for clinical application. This review article elucidates the contemporary applications of Raman spectroscopy in the diagnosis of liver cancer, presents pivotal findings across various sample types, and examines the challenges and future prospects of building Raman-based platforms as dependable diagnostic instruments in oncology.

Figures

Full text

1. Introduction

Liver cancer rates rank sixth among all cancers globally, and is the third major cause of cancer-related mortality [1]. Early-stage liver cancer treatment can ensure a favorable prognosis and comparatively greater survival rate. Late diagnosis, on the other hand, leads to a poor prognosis where the survival rate is relatively low. Absence of nerves in the liver leads to mainly asymptomatic cancer during the early stages [2]. Currently, there are two prevalent screening methods for liver cancer: imaging and serological biomarker assessments [3]. Current imaging modalities comprise ultrasound imaging, magnetic resonance imaging, and computed tomography (Figure 1) [3].

However, these methods possess numerous drawbacks. Sensitivity of ultrasound imaging is highly variable, contingent upon the operator’s expertise and precision of the equipment. Computed tomography and magnetic resonance imaging have low sensitivity for small tumors (less than 1 cm), resulting in a propensity for misdiagnosis or failure to detect [4,5]. Further, an elevated level of the alpha-fetoprotein (AFP) is extensively utilized as a serum biomarker for liver cancer diagnosis [6]. However, even this method suffers from a lack of high sensitivity and specificity for early detection of liver cancer, restricting its clinical utility. In contrast, a punctured biopsy of liver cancer yields not only a definitive pathological diagnosis, significantly aiding in prognosis determination, albeit inducing discomfort to patients. Accordingly, the development of an alternative, complementary, essentially non-invasive, highly sensitive, precise, real-time, and economically viable technique for early detection of liver cancer would be highly beneficial to the field.

In recent years, surface-enhanced Raman spectroscopy (SERS) has been extensively used in biomedicine, yielding significant outcomes in disease diagnosis and screening [7]. Raman spectroscopy is a sensitive optical technique that probes molecular structure through the inelastic scattering of incident photons by vibrational modes in atoms, molecules or their aggregates, such as crystals, leading to secondary photons—scattered light—which can display changes in phase, polarization, and even energy [8,9] (Figure 2A).

The strength of this method is rooted in the uniqueness of each Raman spectrum for every specific molecule, thus providing a basis for molecular fingerprinting [10]. SERS is a powerful advance over conventional Raman spectroscopy considering its enhancement in inelastic Raman scattering. SERS substrates employ nanostructured metallic surfaces, typically composed of silver or gold, to significantly enhance the light intensity and hence amplify the weak Raman scattering signal from analyte molecules proximal to the surface. Optical illumination of a metal surface or a material with a high density of free charges leads to collective oscillation of surface electrons, a phenomenon known as surface plasmon resonance (SPR) [11]. Under specific conditions, these SPR-induced surface charge oscillations couple with electromagnetic waves, leading to the production of surface plasmon polaritons; the quantum of these oscillations is denoted SPP. The excitation of SPPs represents a fundamental step in SPR-based biosensing [12,13,14] (Figure 2B) where the high intensity of the localized light markedly enhances the surface sensitivity of metallic substrates by increasing the probability of Raman scattering vis-à-vis the target analytes [15,16]. This heightened sensitivity has been widely utilized in virus detection technologies [17,18].

When light interacts with metal nanoparticles instead of a continuous thin film, the resulting plasmonic effect is termed localized surface plasmon resonance (LSPR) [19] (Figure 2C). LSPR induces an intense electric field around the nanoparticle surface [19]. Moreover, when nanoparticles are positioned in close proximity or form aggregates, their coupled plasmonic fields generate a substantially stronger electromagnetic field between particles, leading to the electromagnetic enhancement of Raman scattering from certain chemical species [20,21]. This enhancement occurs through two primary mechanisms: the electromagnetic effect, which is associated with LSPRs in the metallic nanostructures, and the chemical effect, which entails electronic interactions between the analyte and the metal surface [22] (Figure 2B). These combined effects enable SERS to detect even single molecules, providing unparalleled sensitivity and specificity for biomedical and diagnostic applications [23,24,25]. Nowadays, SERS-derived data are combined with machine learning (ML) techniques, which lead to ultra-sensitive detection of molecular species. Thus, the non-invasive characteristics and speed of SERS position it to be an optimal instrument for a manifold of screening applications. In the present context, initial phases, liver cancer, and other malignancies in the incipient stage frequently elicit structural alterations in the associated biomolecules circulating in blood [26]. The variations in the SERS spectra of biofluids can thus signify alterations in the associated tissues and hence enable early detection of illnesses. While Raman and SERS offer promising analytical performance and the potential for cost-effective diagnostics, challenges related to inter-laboratory reproducibility, protocol standardization, and regulatory approval remain important considerations for future clinical translation.

The objective of the present review is to explore potential applications of various Raman spectroscopy techniques in the detection and diagnosis of liver cancer, particularly hepatocellular carcinoma (HCC), while also highlighting avenues for future research and practical implementation. The study selection process adhered to a systematic screening methodology. Initially, 125 research articles were identified using keywords associated with Raman technology, Raman spectroscopy, HCC, intrahepatic cholangiocarcinoma (ICC), and liver cancer. The investigation was performed utilizing PubMed, Google Scholar, and Scopus. Following the screening of titles and abstracts, 85 articles were retained, whereas 40 were excluded for their lack of relevance to the research topic or absence of key terms. The comprehensive evaluation eliminated 32 studies due to absent methodological details, irrelevance to the technique, or unreliable outcomes. This resulted in 53 studies that satisfied the inclusion criteria and were examined in the review. This process guaranteed the inclusion of only studies with adequate methodological and scientific significance.

2. Raman Spectroscopy: What Are the Modes and What Are the Applications?

Raman spectroscopy assesses the inelastic scattering of monochromatic light, which produces an ensemble of molecular “fingerprints” of tissues and biofluids where sample preparation requirements are minimal. In biomedicine, Raman spectroscopy and its variants have been investigated for diagnostic and prognostic purposes, considering the ability to detect biochemical changes in lipids, proteins, and nucleic acids, which, combined with multivariate analysis, enables swift categorization [27]. Over time, various improved or modified versions of Raman have been developed to overcome its intrinsically weak signal strength and broaden its application to intricate biological settings. Currently, around 25 distinct types of Raman spectroscopy techniques are used, which include spontaneous Raman, coherent anti-Stokes Raman scattering (CARS) [28], SERS, and tip-enhanced Raman scattering (TERS) [29].

Raman and SERS analyses often necessitate meticulous preprocessing of spectra, including cosmic ray elimination [30], baseline correction [31], and normalization [32], prior to subsequent chemometric or ML evaluation. Multivariate techniques, including principal component analysis (PCA) [33], linear discriminant analysis (LDA) [34], support vector machines (SVM) [35], and deep learning (DL) [36], are currently prevalent, particularly in the analysis of intricate clinical samples. A significant challenge in the application of SERS is the reproducibility of Raman spectra [37], which necessitates the use of standardized substrates and ratiometric methodologies.

Applications of Raman-based technologies in oncology have experienced rapid growth in recent years. Raman spectroscopy has been employed in cancer diagnosis through the analysis of unique chemical compositions [38,39,40]. Several promising pilot studies have shown that Raman spectra can effectively differentiate between malignant and benign skin [41], bladder [42], breast [43], and head and neck [44] tissues with high specificity and sensitivity.

Stimulated Raman histology (SRH), a therapeutic application of stimulated Raman scattering (SRS), generates hematoxylin and eosin-like images of fresh tissue within minutes [45], enabling near-real-time intraoperative identification of brain tumors, and is currently under evaluation for other cancers, including gastrointestinal and urogenital malignancies [46]. Conversely, SERS has become a significant technique for liquid biopsy, as nanoparticles enhance spectral signals from trace biomolecules in serum or plasma samples from patients, and ML classifiers exhibit promising accuracy for early cancer detection [47]. Serum SERS enables the early identification and staging of several malignancies [48], while biomarker-level SERS recognizes proteins, nucleic acids, and cell-surface markers [49], and screening, which together can be extended to personalized and precision medicine.

Consistent with the applications mentioned above, Raman spectroscopy has also been extensively utilized to identify various forms of liver cancer, primarily for the diagnosis of HCC. Here, we explore and summarize the application of Raman spectroscopy and SERS in the diagnosis and treatment of liver cancer vis-à-vis blood serum, liver tissue, blood plasma, and other samples.

3. Sample-Based Raman Application in Liver Cancer Treatment/Diagnosis

3.1. Blood Serum

The utilization of Raman and SERS methodologies for liver cancer diagnosis has advanced from initial proof-of-concept investigations to highly refined strategies using nanostructures and artificial intelligence (AI), demonstrating consistent enhancement in sensitivity, specificity, and clinical relevance. Most of the contemporary applications of Raman spectroscopy for liver cancer research employ blood serum as the sample (Table 1) [50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74]. In 2013, the initial application of serum micro-Raman spectroscopy for HCC diagnosis was documented to differentiate sera from cirrhotic patients with and without HCC, achieving approximately 90% accuracy with SVM models but proving ineffective with PCA alone, underscoring the necessity for robust computational methodologies [62]. Soon after, other researchers utilized Ag-colloidal SERS in conjunction with sophisticated classifiers (Partial Least Squares (PLS)-SVM, Artificial Neural Networks (ANN), and orthogonal partial least squares discriminant analysis (OPLS-DA)), resulting in enhanced classification accuracies (>90%) and the identification of metabolic fingerprints, including tryptophan, valine, and nucleic acid peaks associated with HCC [63,64].

Although AFP and AFP-L3 are crucial in HCC diagnoses, numerous SERS investigations have notably improved their detection, showcasing the capacity to enhance traditional tests. Ma et al. (2017) and Ren et al. (2022) demonstrated that functionalized immunochips and antibody-based nanostructures can assess AFP-L3% with high consistency and sensitivity, thereby tackling a persistent clinical problem [60,65]. These focused strategies leverage clinical familiarity and direct translational relevance; nonetheless, they are fundamentally constrained by the moderate sensitivity of AFP in early HCC. Conversely, metabolite- and protein-based profiling [56,64,67] and miRNA-focused assays [55,72] expand the biomarker repertoire, attaining elevated diagnostic accuracies (>95% in certain instances) and providing insights into tumor metabolism and progression beyond AFP alone. Consequently, AFP/AFP-L3 SERS assays serve as a conduit for clinical implementation, whereas multi-omic and AI-enhanced SERS methodologies may delineate the next era of precision diagnostics.

3.2. Blood Plasma

Plasma-based spectroscopy offers multiple approaches for liver cancer diagnostics, as summarized in Table 2. Magnetic bead-assisted SERS facilitated multiplex and ultra-sensitive detection of AFP, Carcinoembryonic Antigen (CEA), and Ferritin, achieving pg/mL-level limits of detection and 86.7% accuracy, albeit requiring a multi-step preparation process [76]. Raman spectroscopy and ROA, in conjunction with multivariate statistics, demonstrated modified biomolecular plasma composition, facilitating cancer detection and differential diagnosis of gastrointestinal malignancies; nevertheless, specificity among cancer types was constrained [77]. In obese cirrhotic patients, the amalgamation of infrared (IR), Raman, electronic circular dichroism (ECD), and ROA with sophisticated multivariate models demonstrated robust differentiation of HCC from non-HCC (Area under the receiver operating characteristic (AUROC) 0.961; sensitivity 0.81; specificity 0.857), significantly surpassing individual modalities, although necessitating protracted fluorescence quenching and photobleaching procedures [78]. Benchmarking of IR, Raman, and ROA shows that preprocessing selections significantly influence classification accuracy, offering guidance for reproducibility and underscoring the necessity for uniform data pipelines for clinical application [79].

3.3. Liver Tissue

Raman spectroscopy has also been applied to liver tissue samples, with details provided in Table 3 [80,81,82,83,84,85,86,87,88,89]. Lipid signatures detected using Raman imaging demonstrated significant diagnostic accuracy, with Random Forest classification achieving around 86% (sensitivity 76%, specificity 93%) [88]. Optimized pipelines markedly enhanced results: DL models trained on large datasets (>12,000 spectra) attained over 92% accuracy in distinguishing cancer from surrounding tissue and demonstrated consistent effectiveness in differentiating HCC from intrahepatic cholangiocarcinoma (ICC), a clinically challenging task [84]. These clinical investigations frequently surpassed preliminary ex vivo research that depended on limited cohorts and logistic regression, which failed to correctly stratify HCC subclasses. The incorporation of Raman spectroscopy with supplementary methodologies like Matrix-Assisted Laser Desorption/Ionization (MALDI)-Imaging Mass Spectrometry (IMS) enhanced resolution, facilitating both cancer detection and precise grading of HCC, a feat unattainable by Raman alone [81].

Preclinical murine research, conversely, advanced the limits of sensitivity and functional imaging by nanoparticle-based SERS [86]. Gold nanostars and peptide-modified probes achieved signal increases of up to 12-fold relative to unmodified nanostructures, facilitating the detection of microscopic tumors (~250 μm) and early-stage fibrosis that would often elude traditional histology [86]. Fluorescence-guided SERS facilitated the in situ classification and spatial mapping of collagen subtypes, yielding molecular-level fibrosis staging [83]. The most groundbreaking advancement was from the amalgamation of SERS with CT imaging, which not only enhanced sensitivity but also facilitated the swift identification of sub-2 mm lesions within minutes of probe injection, with an accuracy above 91% [80]. Although these findings underscore the significant promise of nanoparticle-assisted SERS for intraoperative navigation and early illness identification, their application is presently constrained by biosafety issues, brief circulation durations, and the absence of extensive human validation.

3.4. Other Potential Samples

As summarized in Table 4, blood, urine, and exosomes serve as promising non-invasive sample types for liver cancer detection [90,91,92]. Circulating tumor cells were among the initial targets, with nanoparticle-enhanced SERS tests exhibiting single-cell detection sensitivity (limit of detection: 1 cell/mL) [92]. Despite their technological sophistication, these spectroscopy methods are constrained by the intricacies of nanoprobe manufacturing [92]. Urine-based SERS has evolved as a more accessible option, capturing spectrum signatures of nucleic acids, amino acids, and metabolites, achieving 83–90% sensitivity and specificity for cirrhosis and approximately 85% for HCC, consistently surpassing serum AFP [57]. The low cost, simple processing requirements, and label-free characteristics render urine an appealing biofluid; nevertheless, bigger multicenter trials are necessary to validate its diagnostic reliability.

Exosome profiling has enhanced the liquid biopsy domain by utilizing tumor-derived vesicles as concentrated molecular repositories [90,91]. Nano-gold plasmonic substrates facilitated very reproducible exosomal SERS spectra, achieving diagnostic performance that surpasses AFP, with sensitivities and specificities approaching 95–100% in differentiating HCC from viral hepatitis cohorts [91]. The domain has evolved to sophisticated AI-assisted frameworks, wherein deep learning, coupled with large language models “ChatExosome”, amalgamated spectral and molecular characteristics to attain over 94% accuracy in a substantial patient cohort, notably maintaining robust performance in AFP-negative instances (87.5%) [90]. This signifies a pivotal advancement towards clinically pertinent solutions, tackling both diagnostic sensitivity and interpretability. These liquid biopsy studies illustrate a progressive evolution: from proof-of-concept blood assays to practical urine-based screening, culminating in exosome-focused platforms that integrate nanoscale sensitivity with AI-driven precision, positioning them at the forefront of translational potential.

3.5. Raman Application in Liver Cancer Cell Lines

Raman-based methodologies have been extensively utilized in in vitro liver cell models, offering regulated settings to analyze spectrum biomarkers and evaluate methodological advancements prior to application on patient samples. Single-cell investigations employing laser tweezers Raman spectroscopy integrated with deep neural networks successfully distinguished hepatocytes from several liver cancer cell lines, revealing metabolic markers associated with differentiation state, highlighting the promise of optical tweezers for high-resolution, label-free diagnostics [94]. Complementary investigations on freshly uncultured primary cells and mixed tumor/non-tumor populations revealed that AI-assisted Raman spectroscopy can attain classification accuracies nearing 90–93% in pure samples; however, performance diminished with a reduced proportion of tumor cells, highlighting the challenge posed by spectral heterogeneity [95]. SERS nanoprobes have been developed to investigate functional and molecular markers in cultured cells: dual-reporter nanoflower probes facilitated ratiometric quantification of carboxylesterase-1 activity in HepG2 cells with integrated internal normalization, while dual-nanoprobe systems specifically targeted lncRNA DAPK1-215, an oncogenic regulator of migration and invasion, achieving precise intracellular detection with minimal cytotoxicity [96]. In addition to cell culture, spiking experiments in blood confirmed the capability of identifying circulating tumor cells at concentrations as low as 1 cell/mL utilizing TiO_2_@Ag nanoprobes, demonstrating proof-of-principle for liquid biopsy applications [97]. These in vitro models collectively demonstrate the adaptability of Raman and SERS platforms, encompassing metabolic phenotyping, functional enzyme assays, and mutation-specific nucleic acid detection, thereby underscoring their significance as a link between mechanistic cellular investigations and clinically pertinent liquid biopsy diagnostics.

4. Conclusions

Raman and SERS have established themselves as versatile and powerful tools in the study and diagnosis of liver cancer, spanning applications from tissue analysis to liquid biopsy and in vitro models. Tissue-based studies have shown that Raman spectroscopy, especially when used with AI classifiers or multimodal platforms, can accurately tell the difference between malignant and non-malignant liver tissue and even grade tumors [95]. These kinds of studies show how useful Raman can be as a supplement to histopathology and as a guide for making decisions during surgery. In addition, liquid biopsy techniques have expanded Raman’s use in non-invasive diagnostics: urine-based SERS [57] has been shown to be better than AFP for finding cirrhosis and HCC, exosome-derived spectra have come close to perfect accuracy [91], and AI-driven exosome platforms [90] now have strong diagnostic power even in patients who are negative for AFP—this is arguably the most clinically important breakthrough so far. Blood-based assays, including CTC detection [92], remain technically impressive but face challenges in scalability and standardization.

The biochemical changes identified by Raman spectroscopy and SERS in HCC can be understood within the larger context of tumor metabolic reprogramming. This metabolic reprogramming is a key characteristic that allows malignant liver cells to continue proliferating, adapt to their tumor microenvironment, and evade immune detection [98]. Core metabolic shifts in HCC include the dysregulation of glucose utilization, lipid biosynthesis, and amino acid metabolism. These shifts not only contribute to the progression of the tumor but also play a role in modulating the immune response. Raman-detectable molecular signatures, such as altered vibrations in lipids, proteins, and nucleic acids, reflect these metabolic changes [99]. These signatures provide label-free insights into the biochemical remodeling associated with cancer, observable across cells, tissues, and bodily fluids. Rather than serving as isolated biomarkers, these spectral features collectively represent the integrated metabolic and microenvironmental changes that occur during hepatocarcinogenesis. This supports the biological relevance of Raman-based diagnostics for characterizing the disease and monitoring therapeutic responses.

In vitro and cell-line studies continue to play an indispensable role, enabling precise exploration of Raman biomarkers, functional enzyme activity, and oncogenic nucleic acid signatures under controlled conditions [96]. These models serve as innovative testbeds, providing mechanistic insight and proof-of-principle evidence prior to translating it for patient-derived samples. However, their diagnostic accuracies, often high in homogeneous settings, must be interpreted cautiously, as they do not fully capture the complexity of clinical biofluids or tumor microenvironments.

Despite the frequently reported high diagnostic accuracy in Raman-based liver cancer studies, such findings must be interpreted cautiously due to methodological limitations. These include small or single-center cohorts, cohort imbalances, reliance on internal validation, and heterogeneous analytical pipelines. Such factors increase the likelihood of overfitting and reduce generalizability across different clinical populations. Particularly for AI- and deep learning-based approaches, it is crucial to assess dataset size, external and multicenter validation, and regulatory feasibility before considering any clinical translation to be reliable. Consequently, the performance metrics currently reported should be viewed as evidence of technical feasibility rather than definitive clinical effectiveness. In addition, cost–benefit justification and operational robustness under real-world conditions remain critical determinants for adoption. Addressing these factors through substrate standardization, harmonized analytical pipelines, external multicenter trials, and regulatory-grade validation frameworks will be essential to enable reliable clinical implementation of Raman-based diagnostics.

Importantly, tissue-based Raman diagnostics and emerging liquid-biopsy platforms appear closest to near-term clinical translation, whereas in vivo nanoparticle-enabled SERS approaches remain largely exploratory due to unresolved challenges related to biosafety, regulatory approval, and long-term biocompatibility [100].

In a nutshell, translating Raman/SERS technology into clinical applications can be a game-changer in diagnosis and precision medicine, recognizing its high sensitivity, real-time sensing, and low cost (>$10/sample) platform. Moreover, as discussed, it does not require complicated sample preparation or invasive methods to collect samples. The amount of sample needed is as low as 2 µL, and using a 785 nm laser typically does not harm biological samples.