Raman Spectroscopy Based Non-Destructive Biomedical Diagnosis

1. Introduction

Early and accurate diagnosis is a key factor in disease management. It guides better treatment decisions and improves outcome[1]. There have been rapid advances in field of medical imaging and laboratory technologies in detecting abnormalities, however, even now, cancer is detected late, especially in the developing nations. This leads to a higher incidence and mortality.

Imaging techniques such as computed tomography scan, magnetic resonance imaging and positron emission tomography may provide detailed anatomical images but they are still costly and time consuming and may subject patients to ionizing radiations[2]. They also require specialized expertise in terms of operation and image assessment, and sophisticated infrastructure[3]. Many diseases including cancers, neurodegenerative and metabolic disorders begin with biochemical changes that precede structural alterations and are still a challenge to identify at an early stage.

In case of cancers, histopathology remains the diagnostic gold standard[4]; tissue evaluation typically relies on hematoxylin and eosin (H&E) staining and/or immunohistochemical staining of thin sections followed by pathological assessment. This again requires specialized expertise, making it prone to subjective errors[5]. Moreover, these are not practical for early screening in populated nations owing to time constraints and paucity of experts[6,7]. Other analytical techniques which assay biomolecules in blood, urine or tissue extracts for proteomics and metabolomics often rely on destructive sampling, multiple reagents and time-consuming workflows[8]. These limitations underscore the need for diagnostic methods that are non-destructive, rapid, and painless and can detect early molecular changes without extensive sampling or subjective interpretation.

Optical techniques are gaining popularity in biomedical research and diagnosis due to their high sensitivity, spatial and temporal resolution, lack of ionizing radiation, and ability to perform real-time, multiplexed imaging while being nondestructive[9,10]. Raman spectroscopy (RS) is based on Raman effect, discovered by Sir C. V. Raman (1928)[11]. It provides a fast, non-destructive, and reliable analytical approach that leverages inelastic light scattering to reveal molecular structures and changes in chemical composition through characteristic Raman shifts. Details on basic principles of RS is provided in Section 2.

Cancer diagnostics and prognostics represent a promising application area, offering minimally invasive, real-time, in vivo assessment and detailed molecular profiling. Literature indicates RS has achieved sensitivities of 77–100% and specificities of 45–100% for identifying malignancies[12].Together, RS and machine learning offer an important adjunct to conventional cytological evaluation for diagnosis[13]. It is to be noted that the current article is not a comprehensive literature review or a summary of advancements over a defined time period. In this review, we have discussed earliest and most recent applications of traditional Raman spectroscopy with a goal to highlight the nondestructive attribute that makes RS suitable for exploring cancers and some non-cancerous conditions.

2. Raman Spectroscopy: Basic Principles

Biochemical changes in living systems frequently occur before visible morphological changes, which in turn are also apparent in the optical properties. Optical techniques can be manipulated to exploit the changes in optical properties in response to the biochemical changes that occur due to pathological events in living systems. Often, the optical techniques utilize ultraviolet (UV), visible, near infrared (NIR) and infrared (IR) regions to measure absorption, scattering and/or fluorescence in biological samples.

Optical techniques typically utilize an excitation source, often a laser, to illuminate the sample. The resulting signals are captured by detectors, with modern instruments commonly employing charge-coupled device (CCD) detectors for signal acquisition. In contemporary systems, fiber optic probes are frequently used to transmit light from the excitation source to the sample and to collect the emitted or scattered signals for delivery to the detectors. These probes generally incorporate a combination of optical components—such as lenses, filters, and mirrors—to efficiently guide and isolate the desired signals.

RS is based on the scattering of photons by vibrating molecular bonds and is therefore categorized as a form of vibrational spectroscopy. Depending on the change in energy or frequency, scattered light can be classified as either elastic (no change in energy) or inelastic (with a change in energy). Inelastic scattering corresponds to Raman scattering and is further divided into anti-Stokes scattering, where the scattered photon gains energy, and Stokes scattering, where the scattered photon loses energy. In recent years, RS has shown considerable promise as a nondestructive technique for distinguishing between diseased and healthy tissues. Additionally, multimodal approaches that combine two or more techniques to obtain complementary information simultaneously are increasingly being explored. The Jablonski diagram (Figure 1) represents vibrational spectroscopy based optical phenomena.

Because water—one of the major constituents of living cells and tissues—produces minimal interference in Raman measurements, Raman spectroscopy is well suited for in vivo biomedical investigations. However, since Raman scattering is inherently a weak phenomenon, its early applications in biological systems were quite limited. Advances in instrumentation, including the development of high-power lasers, efficient Rayleigh rejection filters, and improved detection systems such as charge-coupled devices (CCDs), have significantly expanded its use in biological research. In addition, many portable and transportable Raman instruments are now available at comparatively lower costs, further facilitating broader applications. Figure 2 shows a schematic for RS.

RS has found applications in a diverse range of fields including food adulterations, quality control, environmental health and sample characterization[14,15,16,17,18,19,20,21]. The method is versatile and can be applied to a wide range of [22,23]samples, including tissues, individual cells, urine, serum, and can also be applied in vivo using customized probes[19,24,25,26,27].

Variants of conventional RS, such as Surface-Enhanced Raman Spectroscopy (SERS), Coherent Anti-Stokes Raman Scattering (CARS), Spatially Offset Raman Spectroscopy (SORS), and Stimulated Raman Scattering (SRS), further enhance the capabilities of the technique by improving one of more of following attributes: sensitivity, signal strength, imaging speed, and penetration depth. We refer to the cited literature to gain more understanding of these techniques[28,29,30,31,32,33,34,35,36].

3. Methodology

Raman methodology includes sample selection, appropriate parameters for Raman measurements, preprocessing of raw spectra and analysis of the preprocessed spectra to obtain diagnostic information[37,38,39]. Additionally, multimodal approaches may be undertaken to compare and correlate Raman data with other optical/spectroscopic and biological techniques for additional insights[40].

3.1. Sample Selection

Sample selection is an important step, and it depends on the research questions and many practical considerations. Samples may include freshly excised tissues, tissue sections, cultured cells, biofluids (e.g., serum, plasma, urine), hard tissues (bones and teeth), and in vivo measurements in patients and volunteers using fiber-optic or handheld probes[41,42]. Further steps may depend on sample type and their biological and pathological variations of interest with sufficient. For example, tissue may be processed and sectioned for appropriate thickness while cultured cells may be subjected to wash and synchronizations steps before Raman measurements. Another consideration is substrate selection. A substrate is the surface on which sample is placed before measurement. Commonly used substrates include glass, quartz and calcium fluoride.

3.2. Spectral Acquisition Parameters

Once the appropriate sample is selected, Raman spectra are acquired on key parameters which include excitation wavelength, laser power, integration time, spectral range, number of accumulations, and optical configuration. Optimizing laser wavelength, laser power and the type of objectives lead to optimal signal-to-noise ratio (SNR), reduced autofluorescence and minimized sample damage. Usually higher wavelength (near infrared) is preferred to reduce tissue autofluorescence as well as avoid potential tissue damage (for soft tissues and in vivo applications; higher wavelength has lesser energy, E = hc/λ). 532 nm with water immersion objective has been shown to give better results from dental samples[18]. The spectral range generally covers fingerprint and/or high wavenumber regions, capturing vibrational modes of proteins, lipids, nucleic acids, and other biomolecules[42].

3.3. Spectra Preprocessing and Analyses

Preprocessing of spectra is an essential process as raw Raman spectra may contain unwanted components such as spectral noise, cosmic ray spikes, baseline drift due to fluorescence, and intensity variations unrelated to biochemical differences. Preprocessing commonly involves all or a combination of the following steps: removal of cosmic spikes, background/baseline corrections, normalization and smoothing[43]. These steps make spectra comparable across samples[44].

The preprocessed spectra are suitable for further statistical and multivariate analysis through unsupervised as well as supervised methodologies. Commonly applied are principal component analysis (PCA), T-distributed stochastic neighbor embedding (tSNE), linear discriminant analysis (LDA), support vector machine (SVM), partial least squares (PLS), and recently, deep learning based approaches including autoencoders, convolutional neural network (CNN) and artificial neural network (ANN) that can extract diagnostic information from complex spectral datasets[41,45,46,47,48]. For details on preprocessing, multivariate analysis and machine/deep learning classifiers, readers can refer to the cited articles.

4. Applications of Raman Spectroscopy in Clinical Diagnosis

RS has been widely applied in cancer diagnosis. It is being used for tissue classification, cytological exploration, healthy mucosa characterization, intraoperative margin assessment, and liquid-biopsy style screening. It spans from usage of ex vivo tissue section and fresh surgical specimens, in vivo fiber-optic/endoscopic measurements, and biofluids for non-invasive screening[19,49,50,51]. Table 1 summarizes the founding study for non destructive application of RS on major cancers.

4.1. Oral Cancer

Oral cancer is widely explored using RS. In spite of advancements and general access of oral cavity, the survival rate for oral cancers is surprisingly low, especially in the South Asia where oral cancer is endemic[52,53,54]. One of the earliest studies by Venkatkrishna et al in 2001 explored differences between oral and malignant tissues [55]. Raman spectroscopy has been a promising technology in assessing oral cancer progression, identifying differences between cancer stages in patients and healthy volunteers, and diagnostic exploration using cells, tissues, and liquid biopsy approaches[56,57,58,59,60,61,62]. Some of the founding studies in Raman oral cancers has come from Murali Krishna group involving tissues, in vivo studies, exfoliated cells, biofluids and animal studies[10,14,16,63,64,65,66,67,68,69,70,71,72,73,74,75].

4.2. Breast and Cervical Cancers

Breast and cervical cancers are the most common cancers in females. Breast cancer is also the second most common cancer overall[76]. Thus, early intervention can decrease the incidence and mortality in both cases. Mammogram and fine needle aspiration and cytology are the gold standard for breast cancer diagnosis while Pap smear is primarily used for screening cervical cancers. Raman studies on breast cancers have been reviewed by Katie et al[77]. Major biochemical distinctions between normal and abnormal breast tissue are in terms of lipids, proteins and carotenoids. The earliest Raman studies of breast tissues were by Liu et al (1991)[78]. Recent studies have explored tissues[79] using machine learning and cells using deep learning classifiers[80].

Cervical cancers have been explored through tissues as well as in vivo approaches. The earliest studies were by Mahadevan-Jansen et al to distinguish cervical precancers[81] and other confounding factors[82,83,84]. A new study has tried to integrate Raman, AFM, and electron microscopy to gain insights on biochemical signatures as well as surface morphology from cervical tissues[85].

4.3. Colorectal and Gastric Cancer

Various studies have demonstrated the potential of in vivo Raman spectroscopy being a valuable tool in colon cancer diagnosis. The non- invasive technique allows fast and accurate detection, potentially allowing early intervention. Several studies have reported high accuracy in distinguishing normal mucosa, adenomas and adenocarcinoma, while coupling Raman spectroscopy with advanced machine learning[86]. Gastric cancer studies similarly cover serum SERS assays, tissue section mapping, and endoscopic in vivo Raman, with reviews summarizing both in vitro/ex vivo and in vivo results[87]. Biofluid based studies are minimally invasive in these cases and have been explored using serum by several groups[88,89,90].

4.4. Brain Cancer

Several types of brain cancers explored using RS have been reviewed by Kumar[15]. Mizuno eta (1994) applied Fourier transformed RS spectra to distinguish normal, edematous gray, and white matter using ex vivo approach. In vivo studies under operative conditions have been explored by F Leblond group[91,92]. The same group continues to explore different aspects of Raman exploration in different brain tumors including adenoma and glioblastoma[93,94,95,96]. Glioblastoma can be highly aggressive and often recur. Kaur et al have explore unique spectral markers for recurrent Glioblastoma cells[97]. Sahu et al have also looked at discriminatory potential of serum samples for meningioma detection [98].

4.5. Other Cancers

Skin: Real-time, in vivo Raman spectroscopy have been applied to skin cancer diagnosis for non-invasive differentiation of melanoma, basal cell carcinoma, squamous cell carcinomas actinic keratoses, atypical nevi, melanocytic nevi, blue nevi, and seborrheic keratoses. The findings suggested that RS can be applied in real time to distinguishing malignant and benign skin lesions[99]. A recent report has shown utility of multimodal tools besides RS, including optical coherence tomography, photoacoustic tomography and ultrasound[100].

Liver carcinoma: A recent study by Huang et al used a handheld probe for intraoperative liver cancer detection. After surgery the tissues were also collected to detect metabolic differences via LC-MS. In the same study, 240 tissue samples from 120 patients were also employed[101].

4.6. Neurological/Neurodegenerative Disorders

Raman spectroscopy has been explored for diagnosis of neurodegenerative disease from biofluids[13]. Studies have analyzed CSF and plasma/serum Raman spectra for Alzheimer’s Disease classification, with machine-learning aiding discrimination of disease states[102]. Separately, engineered intracranial Raman probes (preclinical) have been piloted for real-time Traumatic brain injury (TBI) monitoring[103]. Alix et al did a follow up study on Amyotrophic lateral sclerosis (ALS) patients to identify spectral patterns [104]. In another Raman serum investigation, patients with ADHD, bipolar disorder, and healthy controls were identified with 97.4% accuracy[105].

4.7. Atherosclerosis

Animal studies have shown ability of RS to characterize plaque (cholesterol, lipid, calcification) and differentiate vulnerable and stable lesions[106]. Fiber-optic and catheter-based Raman probes (integrated with intravascular imaging modalities) have been developed for intravascular plaque assessment[107]. Recently Kesic et al have used serum for identification of atherosclerotic carotid stenosis[108].

4.8. Infections and Pathogen Identification

RS have been used for rapid, label-free identification of bacteria and other pathogens, with several groups demonstrating species-level discrimination and even antibiotic-resistance phenotyping involving large spectral databases as well as deep-learning models. Recent studies report promising clinical-sample workflows for bloodstream infections aiming to reduce culture time[109]. RS has been extensively explored for bacterial species characterization in Popp group[110,111].

Malaria & Dengue: Malaria and Dengue are tropical infectious diseases which affect RBCs and cause significant morbidity and mortality[112]. Circulating RBCs are affected by the infectious pathogen and are susceptible to RBC programmed eryptosis or RBC programmed necrosis. Raman features can distinguish RBCs in healthy and diseased states[113]. Healthy, dengue and malaria serum samples was used by Patel et al to achieve good classifications using PC-LDA [114]. Khan et al have used different multivariate methods to explore dengue serum samples[115,116,117].

COVID-19: The COVID-19 pandemic created an urgent need for tests that are fast, reliable, and easy to use at the point of care. RS based diagnosis is especially useful in view of frequent false negative results through rapid methods. Serum based RS has shown good classification in recent studies[118,119]. In an interesting study, besides Raman exploration, antibody levels were also measures in patients and the study reported 100% sensitivity in differentiating patients at different stages post infection[120].

5. Discussion

Raman spectroscopy has demonstrated tremendous potential as a non-destructive diagnostic technique across a wide range of clinical conditions. The ability to obtain label-free biochemical fingerprints from a wide range of specimens like tissues, biofluids, and even directly from patients in vivo makes RS a unique diagnostic method. Nevertheless, despite these advantages, several barriers still limit widespread clinical adoption.

There is still scope of improvements in signal to noise ratio, fluorescence background reduction, and data analysis pipelines. Recent improvements in instrumentation and machine learning, along with enhanced methods like SERS have helped address some of these limitations. Continued progress will likely expand its clinical use. Achieving full clinical translation will require multi-center and multidisciplinary collaborations between researchers, clinicians, optical engineers, data analysis and statistical[121,122]. Differences in instrument configuration, spectral pre-processing methods, and data labeling can hinder reproducibility across laboratories. Establishing standardized acquisition protocols, harmonized preprocessing pipelines, and shared spectral libraries will be crucial to advancing Raman-based diagnostics from exploratory studies to routine clinical use. While Raman studies in vivo are convenient at accessible sites like oral cavity and skin, integration with modalities like endoscopes and colonoscopes can be useful in gastrointestinal cases. In other cases, biofluid based diagnostic programs can be useful wherein samples can be collected and transported to a central facility for initial screening. Positive cases can be followed up and confirmed via traditional gold standards. Integration of Raman spectroscopy with complementary imaging modalities further strengthens diagnostic capability and reflects a broader move toward multimodal precision diagnostics.

6. Conclusions

Raman spectroscopy represents a rapidly evolving optical tool with strong applications for non-destructive biomedical diagnosis. By probing characteristic vibrational signatures of biomolecules, it enables detailed insight into cellular and extracellular composition without the need for exogenous labels or chemical processing or destructive methodologies. Applications across cancers, neurodegenerative disease, infections, and metabolic disorders consistently demonstrate its potential to detect biochemical alterations that precede gross structural changes.

Ongoing improvements in hardware, signal enhancement techniques, and advanced data analytics are steadily overcoming traditional limitations related to sensitivity, speed, processing time and interpretability. As clinical validation studies increase in scale and standardization improves, Raman spectroscopy is poised to transition from ‘bench-side to bedside’, acting as a practical adjunct in clinical diagnostics, surgical guidance, and personalized medicine. In summary, Raman spectroscopy when combined with enhanced modalities and machine learning offers a powerful, non-destructive approach to biomedical diagnosis.