Recent Advances in AI-Integrated SERS Sensors for Biomedical Diagnostics

1. INTRODUCTION

Raman spectroscopy has attracted significant attention in the biomedical sensing field owing to its ability to provide molecular fingerprint information through non-destructive and label-free analysis [1,2]. Since Raman signals originate from the intrinsic vibrational modes of molecules, Raman spectroscopy enables highly selective chemical identification of complex biological samples. However, the inherently weak Raman scattering efficiency often limits its practical sensitivity for trace-level biomarker detection and clinical diagnostics [3,4]. To overcome this limitation, surface-enhanced Raman scattering (SERS) has been extensively investigated as an advanced spectroscopic sensing platform capable of dramatically amplifying Raman signals through localized surface plasmon resonance generated on noble metal nanostructures such as gold (Au) and silver (Ag) [5].

Owing to its ultrasensitive detection capability, SERS has been widely applied to various biomedical applications, including cancer diagnostics, pathogen detection, and biofluid analysis [6-8]. Recent advances in nanofabrication technologies have enabled the development of high-performance plasmonic substrates with improved signal enhancement and reproducibility [5,9]. Nevertheless, conventional SERS-based biosensors still face several limitations that hinder their practical clinical translation. Biological samples typically exhibit highly complex and overlapping spectral features, making accurate interpretation challenging. In addition, signal variability caused by substrate heterogeneity, environmental conditions, and instrument-dependent spectral differences often reduces analytical reliability [10,11]. Consequently, conventional spectral interpretation frequently relies on expert-dependent analysis, limiting rapid and automated diagnosis.

Recently, artificial intelligence (AI)-assisted spectral analysis has emerged as a promising strategy for overcoming the limitations of conventional SERS sensing systems. Machine learning and deep learning algorithms can efficiently extract hidden spectral patterns from complex Raman datasets and enable automated classification of biological samples with high accuracy [12,13]. By integrating AI with SERS platforms, intelligent sensing systems capable of rapid, reproducible, and user-independent diagnosis have become increasingly feasible. Furthermore, recent studies have demonstrated the potential of portable Raman devices combined with AI-driven spectral processing for point-of-care biomedical applications [14-16].

Among emerging biomedical applications, cerebrospinal fluid (CSF) rhinorrhea diagnosis has recently attracted attention as a representative target for AI-SERS sensing. CSF rhinorrhea is characterized by leakage of cerebrospinal fluid through the nasal cavity and is often difficult to distinguish from normal nasal secretions because of their visually similar characteristics [17,18]. Delayed diagnosis can increase the risk of severe complications such as meningitis due to pathogen penetration into the central nervous system [19]. Therefore, rapid and accurate discrimination between CSF and nasal secretions is clinically important. Recent studies have demonstrated that AI-integrated SERS platforms can effectively classify these highly similar biofluids through intelligent spectral analysis [20].

In this review, recent advances in AI-integrated SERS sensors for biomedical diagnostics are summarized, with particular emphasis on machine learning-assisted spectral analysis and plasmonic substrate engineering strategies. In particular, cerebrospinal fluid (CSF) rhinorrhea is highlighted in this review as a representative case study because it involves the discrimination of two visually similar and biochemically overlapping biofluids, CSF and nasal secretion. This diagnostic challenge clearly illustrates the strength of AI-SERS sensing, which can extract subtle molecular differences from complex spectra and provide rapid classification beyond conventional visual or peak-based interpretation.

2. CORE COMPONENTS OF AI-SERS SENSORS

2.1 Principle of Surface-Enhanced Raman Scattering (SERS)

Surface-enhanced Raman scattering (SERS) is an advanced spectroscopic sensing technique capable of significantly amplifying Raman signals through plasmonic interactions occurring on metallic nanostructures [21]. As shown in Fig. 1(a), Raman signals inherently have low scattering efficiency, which limits their direct use as a sensing technology. To overcome this, SERS amplifies the local electromagnetic field near nanostructured metal surfaces, particularly surfaces composed of gold (Au) and silver (Ag), enabling high-sensitivity molecular detection. When incident light excites localized surface plasmon resonance (LSPR) on metallic nanostructures, strong electromagnetic fields are generated in nanoscale junctions, commonly referred to as “hotspots,” where Raman signal enhancement can increase by several orders of magnitude [22].

Fig. 1.

Schematic diagram of the SERS mechanism. Difference between (a) Raman signal and (b) SERS signal is highlighted with electromagnetic and chemical enhancement.

The signal enhancement mechanism of SERS is generally explained by two major contributions: electromagnetic enhancement and chemical enhancement (Fig. 1(b)). Electromagnetic enhancement primarily originates from localized plasmonic field amplification around metallic nanostructures and is considered the dominant mechanism responsible for high signal enhancement. Chemical enhancement, although relatively smaller in magnitude, arises from charge-transfer interactions between target molecules and metal surfaces, contributing to improved Raman scattering efficiency. In biomedical sensing applications, the synergistic effects of these mechanisms enable ultrasensitive detection of biomolecules even at trace concentrations [1].

To maximize sensing performance, considerable efforts have been devoted to engineering plasmonic substrates capable of generating dense and reproducible hotspots [23]. Various nanostructures, including nanoparticles, nanopillars, nanogaps, and hierarchical hybrid architectures, have been developed to optimize plasmonic coupling and signal reproducibility. In particular, Au and Ag-based hybrid nanostructures have attracted increasing attention owing to their ability to combine strong plasmonic enhancement with improved structural stability, thereby facilitating reliable sensing performance in complex biological environments [24].

2.3 Spectral Preprocessing for AI-SERS Sensors

Reliable AI-SERS sensing requires not only robust machine learning algorithms but also appropriate preprocessing of Raman spectra. Raw Raman data often contain fluorescence background, noise, baseline drift, and intensity variation caused by experimental conditions or instrument settings. Without suitable preprocessing, these unwanted spectral variations may reduce model performance and classification reliability.

Accordingly, preprocessing techniques such as baseline correction, smoothing, normalization, denoising, and dimensionality reduction are commonly applied before machine learning analysis [30]. As shown in Fig. 2, baseline correction is frequently used to remove fluorescence interference, whereas normalization techniques help minimize intensity fluctuations among measurements. In addition, dimensionality reduction approaches such as PCA can simplify complex spectral datasets and facilitate efficient classification.

Fig. 2.

Schematic representation of representative spectral preprocessing procedures for AI-SERS analysis, including baseline correction, smoothing, normalization, denoising, and dimensionality reduction. These preprocessing strategies improve spectral quality and enhance feature extraction for reliable machine learning-assisted classification of Raman signals.

Another emerging challenge in AI-SERS sensing involves instrument-dependent variability, including differences in spectral range, resolution, and acquisition conditions among Raman systems. Such inconsistencies may hinder model transferability and limit practical deployment across different sensing platforms. Therefore, recent studies have increasingly focused on cross-instrument spectral standardization strategies to improve robustness and enable portable point-of-care Raman diagnostics [31,32]. By combining optimized preprocessing with AI-assisted analysis, AI-SERS sensors are evolving into reliable and intelligent platforms for biomedical sensing applications.

3. BIOMEDICAL APPLICATIONS OF AI-SERS SENSORS

Artificial intelligence-assisted surface-enhanced Raman scattering (AI-SERS) sensors have recently emerged as powerful platforms for biomedical diagnostics by combining ultrasensitive molecular fingerprinting with automated spectral interpretation. In contrast to conventional Raman analysis, which often requires expert interpretation and suffers from complex spectral overlap, AI-assisted approaches enable rapid classification of biological samples through machine learning-driven pattern recognition [33]. Owing to these advantages, AI-SERS sensors have been increasingly investigated for cancer diagnosis, pathogen detection, and biofluid-based disease screening.

3.3 Biofluid-Based Diagnostics

Beyond cancer and pathogen detection, AI-SERS sensing has been increasingly explored for biofluid-based diagnostics because many diseases induce biochemical changes that can be reflected in blood, urine, saliva, tears, and nasal secretions [39]. Since Raman spectroscopy is minimally affected by water interference and requires limited sample preparation, SERS-based biofluid analysis provides a promising route toward non-invasive and rapid diagnosis.

Table 1.

Representative studies of biomedical AI-SERS sensors.

In particular, urine- and serum-based AI-SERS systems have demonstrated promising performance for disease screening through multivariate statistical analysis and machine learning classification [40]. Biofluid-derived molecular fingerprints combined with AI analysis allow discrimination between healthy and disease states while reducing user dependency and diagnostic subjectivity. More importantly, recent progress in portable Raman instrumentation suggests the possibility of translating AI-SERS sensing into point-of-care diagnostic systems.

Overall, recent studies suggest that AI-SERS sensors can significantly improve the interpretability and diagnostic performance of conventional Raman biosensing systems by enabling automated feature extraction and intelligent classification. While current applications are primarily focused on cancer screening, pathogen identification, and biofluid analysis, the continued advancement of portable Raman systems and cross-platform AI processing is expected to further accelerate clinical translation. In this context, CSF rhinorrhea diagnosis represents an emerging example in which AI-SERS sensing can effectively distinguish highly similar biofluids for rapid clinical decision-making.

4. AI-SERS SENSOR FOR CEREBROSPINAL FLUID RHINORRHEA DIAGNOSIS

In this section, CSF rhinorrhea diagnosis is discussed not merely as an individual application but as a representative example demonstrating the practical value of AI-SERS sensing for distinguishing highly similar biological matrices. This case is particularly relevant because the diagnostic problem requires rapid, minimally invasive, and objective discrimination between CSF leakage and normal nasal secretion, which directly matches the core advantages of AI-integrated SERS platforms.

4.2 AI-SERS Strategy for Biofluid Discrimination

Recent studies have demonstrated that integrating SERS with machine learning enables discrimination of highly similar biofluids through the recognition of subtle spectral variations that are difficult to resolve by conventional inspection [20]. In the context of CSF rhinorrhea, an AI-SERS sensing workflow has been proposed in which patient-derived samples are directly deposited onto plasmonic SERS substrates, followed by Raman acquisition and automated spectral classification. Rather than relying on a single diagnostic biomarker, the approach interprets collective spectral patterns generated from multiple biochemical constituents present in biological fluids, including proteins, glycoproteins, lipids, and nucleic acid-related components.

As a representative strategy, signal stability was enhanced under physiologically suitable conditions using a corrosion-resistant Au@Ag bimetallic nanopillar substrate, as shown in Fig. 3(a). Compared with conventional metallic SERS structures, hybrid plasmonic architectures may provide improved robustness and stronger signal amplification through synergistic electromagnetic coupling. Importantly, this platform enabled reproducible spectral acquisition from clinical CSF and nasal secretion specimens while maintaining compatibility with portable Raman systems.

Fig. 3.

Schematic illustration of the AI-integrated SERS platform for cerebrospinal fluid (CSF) rhinorrhea diagnosis. (a) SERS-based molecular fingerprint acquisition using an Au@Ag nanopillar substrate. (b) AI-assisted spectral preprocessing and machine learning-based classification of CSF and nasal secretions.

4.4 Toward Point-of-Care AI-SERS Diagnosis

One of the major advantages of AI-SERS sensing for CSF rhinorrhea diagnosis lies in its compatibility with portable diagnostic systems. As Raman instrumentation continues to evolve toward miniaturized and portable formats, recent studies have explored cross-instrument spectral standardization strategies to compensate for discrepancies in spectral range and resolution between benchtop and portable Raman devices. As illustrated in Fig. 4(a), although the same target analyte (4-aminothiophenol, 4-ATP) measured using benchtop and portable Raman systems exhibits identical characteristic Raman peaks, noticeable differences remain in spectral range and the number of data points corresponding to individual Raman bands. Such inconsistencies may reduce model transferability and limit the direct applicability of machine learning-assisted spectral classification across different devices.

Fig. 4.

Description of the CISP (Cross-Instrument Spectrum Preprocessing) algorithm and platform validation using a portable Raman spectrometer. (a) SERS spectra and spectral resolution of 4-ATP measured with benchtop and portable Raman spectrometers. (b) Effect of CISP on SERS spectra of 4-ATP measured with different instruments. (c) SERS spectra of pre-treated cerebrospinal fluid (CSF) and physiological saline (NS).

To address these challenges, preprocessing and AI-based calibration strategies have been introduced to standardize instrument-dependent spectral variation, as shown in Fig. 4(b). By harmonizing spectral resolution, aligning data point density, and minimizing inter-device variability, these correction algorithms substantially improve the suitability of Raman spectra as training datasets for machine learning models. When applied to CSF and nasal secretion samples, the proposed framework reduced spectral variability and enhanced classification robustness, thereby facilitating more reliable discrimination between highly similar biological matrices (Fig. 4(c)).

Collectively, these findings highlight CSF rhinorrhea diagnosis as a representative example demonstrating how AI-assisted SERS sensing can bridge ultrasensitive molecular analysis with practical clinical translation. More broadly, this case underscores the increasingly important role of intelligent biosensing systems in addressing diagnostic challenges involving highly similar biological matrices that are difficult to differentiate using conventional analytical approaches.

5. CONCLUSIONS

The integration of AI with SERS has opened new opportunities for intelligent biomedical sensing by overcoming several longstanding limitations of conventional Raman analysis, including spectral complexity, signal variability, and user-dependent interpretation. By coupling ultrasensitive molecular fingerprint acquisition with automated data-driven classification, AI-SERS platforms enable rapid, objective, and highly reliable analysis of complex biological samples. As discussed throughout this semi-review, recent advances in plasmonic substrate engineering, spectral preprocessing strategies, and machine learning-assisted analysis have significantly improved the robustness and diagnostic capability of SERS-based sensing systems.

Particularly, biomedical applications involving cancer diagnosis, pathogen detection, and biofluid-based analysis highlight the growing utility of AI-SERS sensors for non-invasive and label-free diagnostics. Among these, CSF rhinorrhea diagnosis represents a compelling example demonstrating how AI-assisted Raman analysis can successfully distinguish highly similar biological matrices that remain difficult to classify through conventional analytical approaches. The integration of compelling Raman instrumentation and cross-platform spectral standardization further underscores the translational potential of AI-SERS sensing toward real-world clinical implementation.

Despite these advances, several challenges remain before widespread clinical adoption becomes feasible. The reproducibility and large-scale manufacturability of SERS substrates, standardization of spectral acquisition across instruments, availability of sufficiently diverse clinical datasets, and interpretability of AI-driven decisions continue to require further investigation. In particular, future research should move beyond isolated proof-of-concept demonstrations toward clinically robust sensing frameworks capable of maintaining diagnostic reliability across heterogeneous populations and measurement environments.

Looking forward, AI-SERS sensors are expected to evolve into next-generation intelligent biomedical sensing platforms capable of real-time, portable, and personalized diagnostics. Future systems may integrate miniaturized Raman hardware, adaptive machine learning models, cloud-based data analysis, and multimodal biosensing strategies to enable continuous health monitoring and point-of-care clinical decision support. Rather than functioning solely as analytical devices, AI-SERS technologies are likely to become intelligent diagnostic ecosystems that autonomously interpret molecular information and assist precision medicine. In this context, the convergence of AI and SERS is anticipated to play an increasingly important role in shaping future biomedical sensing technologies and accelerating the transition toward smarter and more accessible healthcare systems.

Acknowledgments

This research was funded by the National Research Foundation of Korea (NRF) under Grant Nos. RS-2023-NR077066, RS-2024-00438542 and RS-2025-00554830.

REFERENCES

• R. Beeram, K.R. Vepa, V.R. Soma, Recent trends in SERS-based plasmonic sensors for disease diagnostics, biomolecules detection, and machine learning techniques, Biosensors 13 (2023) 328. [https://doi.org/10.3390/bios13030328]
• R. Pilot, R. Signorini, C. Durante, L. Orian, M. Bhamidipati, L. Fabris, A review on surface-enhanced Raman scattering, Biosensors 9 (2019) 57. [https://doi.org/10.3390/bios9020057]
• Y. Liu, Y. Wu, J. Wang, J. Qi, C. Zhou, Y. Xue, Recent Advances in Raman Spectral Classification with Machine Learning, Sensors 26 (2026) 341. [https://doi.org/10.3390/s26010341]
• J. Luo, J. Sun, H. Chen, C. Zhao, M. Ye, A Review of High-Sensitivity SERS-Active Photonic Crystal Fiber Sensors for Chemical and Biological Detection, Sensors 25 (2025) 6982. [https://doi.org/10.3390/s25226982]
• L.L. Lin, R. Alvarez-Puebla, L.M. Liz-Marzan, M. Trau, J. Wang, L. Fabris, et al., Surface-enhanced Raman spectroscopy for biomedical applications: recent advances and future challenges, ACS Appl. Mater. Interfaces 17 (2025) 16287–16379. [https://doi.org/10.1021/acsami.4c17502]
• K. Yuan, B. Jurado-Sánchez, A. Escarpa, Nanomaterials meet surface-enhanced Raman scattering towards enhanced clinical diagnosis: a review, J. Nanobiotechnology 20 (2022) 537. [https://doi.org/10.1186/s12951-022-01711-3]
• M. Constantinou, K. Hadjigeorgiou, S. Abalde-Cela, C. Andreou, Label-free sensing with metal nanostructure-based surface-enhanced Raman spectroscopy for cancer diagnosis, ACS Appl. Nano Mater. 5 (2022) 12276–12299. [https://doi.org/10.1021/acsanm.2c02392]
• N. Lyu, A. Hassanzadeh-Barforoushi, L.M. Rey Gomez, W. Zhang, Y. Wang, SERS biosensors for liquid biopsy towards cancer diagnosis by detection of various circulating biomarkers: current progress and perspectives, Nano Converg. 11 (2024) 22. [https://doi.org/10.1186/s40580-024-00428-3]
• K. Milenko, F.T. Dullo, P.C. Thrane, Z. Skokic, C.A. Dirdal, UV-nanoimprint lithography for predefined SERS nanopatterns which are reproducible at low cost and high throughput, Nanomaterials 13 (2023) 1598. [https://doi.org/10.3390/nano13101598]
• I.A. Bratchenko, L.A. Bratchenko, Overestimation of the classification model for Raman spectroscopy data of biological samples, Msystems 10 (2025) e00063–25. [https://doi.org/10.1128/msystems.00063-25]
• S. Lee, R. Park, H.S. Jung, AI‐enhanced surface‐enhanced raman scattering for accurate and sensitive biomedical sensing, Adv. Intell. Disc. 2 (2026) e202500030. [https://doi.org/10.1002/aidi.202500030]
• A. Lebrun, H. Fortin, N. Fontaine, D. Fillion, O. Barbier, D. Boudreau, Pushing the limits of surface-enhanced Raman spectroscopy (SERS) with deep learning: identification of multiple species with closely related molecular structures, Appl. Spectrosc. 76 (2022) 609–619. [https://doi.org/10.1177/00037028221077119]
• F. Ebrahimi, A. Kumari, K. Dellinger, Integration of nanoengineering with artificial intelligence and machine learning in surface‐enhanced Raman spectroscopy (SERS) for the development of advanced biosensing platforms, Adv. Sensor Res. 4 (2025) 2400155. [https://doi.org/10.1002/adsr.202400155]
• C. Soliman, D. Tu, S. Mabbott, G. Coté, K. Maitland, Portable, multi-modal Raman and fluorescence spectroscopic platform for point-of-care applications, J. Biomed. Opt. 27 (2022) 095006. [https://doi.org/10.1117/1.JBO.27.9.095006]
• M. Fan, Y. Weng, Y. Liu, Y. Lu, L. Xu, J. Ye, et al., Surface‐Enhanced Raman Spectroscopy‐Based Optical Biosensor for Liquid Biopsy: Toward Precision Medicine, Laser Photonics Rev. 18 (2024) 2301072. [https://doi.org/10.1002/lpor.202301072]
• T.R. Szymborski, S.M. Berus, A.B. Nowicka, G. Słowiński, A. Kamińska, Machine learning for COVID-19 determination using surface-enhanced raman spectroscopy, Biomedicines 12 (2024) 167. [https://doi.org/10.3390/biomedicines12010167]
• A. Abuabara, Cerebrospinal fluid rhinorrhoea: diagnosis and management, Med. Oral Patol. Oral Cir. Bucal 12 (2007) 397–400.
• M. Xie, K. Zhou, S. Kachra, T. McHugh, D.D. Sommer, Diagnosis and localization of cerebrospinal fluid rhinorrhea: a systematic review, Am. J. Rhinol. Allergy 36 (2022) 397–406. [https://doi.org/10.1177/19458924211060918]
• M. Mantur, M. Łukaszewicz-Zając, B. Mroczko, A. Kułakowska, O. Ganslandt, H. Kemona, et al., Cerebrospinal fluid leakage—Reliable diagnostic methods, Clin. Chim. Acta 412 (2011) 837–840. [https://doi.org/10.1016/j.cca.2011.02.017]
• E. Park, H. Park, W. Kim, J. Park, K. Chai, G. Kim, et al., Ultrasensitive CSF rhinorrhea screening via machine learning-aided SERS on Au@ Ag nanopillars, J. Mater. Sci. Technol. 266 (2025) 236–249. [https://doi.org/10.1016/j.jmst.2025.11.047]
• M. Fleischmann, P.J. Hendra, A.J. McQuillan, Raman spectra of pyridine adsorbed at a silver electrode, Chem. Phys. Lett. 26 (1974) 163–166. [https://doi.org/10.1016/0009-2614(74)85388-1]
• W. Hou, S.B. Cronin, A review of surface plasmon resonance‐enhanced photocatalysis, Adv. Funct. Mater. 23 (2013) 1612–1619. [https://doi.org/10.1002/adfm.201202148]
• H. Park, W. Kim, K. Chai, G. Kim, C. Kang, J. Park, Advances in Raman spectroscopic detection of pesticide residues: focus on substrate innovations, J. Sens. Sci. Technol. 34 (2025) 387–397. [https://doi.org/10.46670/JSST.2025.34.4.387]
• H. Park, K. Chai, W. Kim, J. Park, W. Lee, J. Park, Asterias forbesi-inspired SERS substrates for wide-range detection of uric acid, Biosensors 14 (2023) 8. [https://doi.org/10.3390/bios14010008]
• Y. Ding, Y. Sun, C. Liu, Q.Y. Jiang, F. Chen, Y. Cao, SERS‐based biosensors combined with machine learning for medical application, ChemistryOpen 12 (2023) e202200192. [https://doi.org/10.1002/open.202200192]
• L. Pan, P. Zhang, C. Daengngam, S. Peng, M. Chongcheawchamnan, A review of artificial intelligence methods combined with Raman spectroscopy to identify the composition of substances, J. Raman Spectrosc. 53 (2022) 6–19. [https://doi.org/10.1002/jrs.6225]
• M. Kluz-Barłowska, T. Kluz, W. Paja, J. Sarzyński, M. Łączyńska-Madera, A. Odrzywolski, et al., FT-Raman data analyzed by multivariate and machine learning as a new methods for detection spectroscopy marker of platinum-resistant women suffering from ovarian cancer, Sci. Rep. 13 (2023) 20772. [https://doi.org/10.1038/s41598-023-48169-3]
• M. Yesmean, B.R. Shakya, M. Mannerkorpi, S. Saarakkala, M. Jansson, Predictive Performance of Raman Spectroscopy in Osteoarthritis: A Systematic Review, J. Med. Syst. 49 (2025) 166. [https://doi.org/10.1007/s10916-025-02304-x]
• N. Zhang, Z. Wang, X. Wang, G.A. Vázquez-Lizardi, P.P. Varela, D.J. de Aberasturi, et al., Machine Learning Integrated Near-Infrared Surface-Enhanced Raman Spectroscopy for Accurate Strain-Level Virus Identification, ArXiv., https://arxiv.org/abs/2509.09851, (2025).
• H.J. Butler, L. Ashton, B. Bird, G. Cinque, K. Curtis, J. Dorney, et al., Using Raman spectroscopy to characterize biological materials, Nat. Protoc. 11 (2016) 664–687. [https://doi.org/10.1038/nprot.2016.036]
• T.W. Bocklitz, S. Guo, O. Ryabchykov, N. Vogler, J. Popp, Raman based molecular imaging and analytics: a magic bullet for biomedical applications!?, Anal. Chem. 88 (2016) 133–151. [https://doi.org/10.1021/acs.analchem.5b04665]
• I. Pence, A. Mahadevan-Jansen, Clinical instrumentation and applications of Raman spectroscopy, Chem. Soc. Rev. 45 (2016) 1958–1979. [https://doi.org/10.1039/C5CS00581G]
• B. Chen, H. Sun, J. Gao, Z. Chen, X. Qiu, Machine learning-enhanced SERS diagnostics: Accelerating the AI-powered transition from laboratory discoveries to clinical practice, Comput. Biol. Med. 194 (2025) 110490. [https://doi.org/10.1016/j.compbiomed.2025.110490]
• M. Chen, M. Zhao, Y. Cai, Q. Zhang, Z. Peng, Q. Li, et al., Surface-enhanced Raman spectroscopy for label-free cancer liquid biopsy: from fundamentals to clinical analysis of biofluid, Front. Chem. 13 (2025) 1696979. [https://doi.org/10.3389/fchem.2025.1696979]
• H. Shin, B.H. Choi, O. Shim, J. Kim, Y. Park, S.K. Cho, et al., Single test-based diagnosis of multiple cancer types using Exosome-SERS-AI for early stage cancers, Nat. Commun. 14 (2023) 1644. [https://doi.org/10.1038/s41467-023-37403-1]
• J. Shi, R. Li, Y. Wang, C. Zhang, X. Lyu, Y. Wan, et al., Detection of lung cancer through SERS analysis of serum, Spectrochim. Acta A Mol. Biomol. Spectrosc. 314 (2024) 124189. [https://doi.org/10.1016/j.saa.2024.124189]
• M. H.-U. Rahman, R. Sikder, M. Tripathi, M. Zahan, T. Ye, E. Gnimpieba Z., et al., Machine learning-assisted raman spectroscopy and SERS for bacterial pathogen detection: clinical, food safety, and environmental applications, Chemosensors 12 (2024) 140. [https://doi.org/10.3390/chemosensors12070140]
• J. Ye, Y.-T. Yeh, Y. Xue, Z. Wang, N. Zhang, H. Liu, et al., Accurate virus identification with interpretable Raman signatures by machine learning, Proc. Natl. Acad. Sci. U.S.A. 119 (2022) e2118836119. [https://doi.org/10.1073/pnas.2118836119]
• V. Moisoiu, S.D. Iancu, A. Stefancu, T. Moisoiu, B. Pardini, M.P. Dragomir, et al., SERS liquid biopsy: An emerging tool for medical diagnosis, Colloids Surf. B Biointerfaces 208 (2021) 112064. [https://doi.org/10.1016/j.colsurfb.2021.112064]
• B.A. Buhas, V. Toma, J.-B. Beauval, I. Andras, R. Couți, L.A.-M. Muntean, et al., Label-free SERS of urine components: a powerful tool for discriminating renal cell carcinoma through multivariate analysis and machine learning techniques, Int. J. Mol. Sci. 25 (2024) 3891. [https://doi.org/10.3390/ijms25073891]
• A. Conger, F. Zhao, X. Wang, A. Eisenberg, C. Griffiths, F. Esposito, et al., Evolution of the graded repair of CSF leaks and skull base defects in endonasal endoscopic tumor surgery: trends in repair failure and meningitis rates in 509 patients, J. Neurosurg. 130 (2018) 861–875. [https://doi.org/10.3171/2017.11.JNS172141]
• N. Choi, H. Dang, A. Das, M.S. Sim, I.Y. Chung, J. Choo, SERS biosensors for ultrasensitive detection of multiple biomarkers expressed in cancer cells, Biosens. Bioelectron. 164 (2020) 112326. [https://doi.org/10.1016/j.bios.2020.112326]
• L. Vázquez-Iglesias, G.M.S. Casagrande, D. García-Lojo, L.F. Leal, T.A. Ngo, J. Pérez-Juste, et al., SERS sensing for cancer biomarker: Approaches and directions, Bioact. Mater. 34 (2024) 248–268. [https://doi.org/10.1016/j.bioactmat.2023.12.018]