Rapid and precise detection of cancers via label-free SERS and deep learning

Diagnosis of neuropsychiatric systemic lupus erythematosus by label-free serum microsphere-coupled SERS fingerprints with machine learning

Surface-enhanced Raman spectroscopy (SERS) is a powerful optical technique for non-invasive and label-free bioanalysis of liquid biopsy, facilitating to diagnosis of potential diseases. Neuropsychiatric systemic lupus erythematosus (NPSLE) is one of the subgroups of systemic lupus erythematosus (SLE) with serious manifestations for a high mortality rate. Unfortunately, lack of well-established gold standards results in the clinical diagnosis of NPSLE being a challenge so far. Here we develop a novel Raman fingerprinting machine learning (ML-) assisted diagnostic method. The microsphere-coupled SERS (McSERS) substrates are employed to acquire Raman spectra for analysis via convolutional neural network (CNN). The McSERS substrates demonstrate better performance to distinguish the Raman spectra from serums between SLE and NPSLE, attributed to the boosted signal-to-noise ratio of Raman intensities due to the multiple optical regulation in microspheres and AuNPs. Eight statistically-significant (p-value <0.05) Raman shifts are identified, for the first time, as the characteristic spectral markers. The classification model established by CNN algorithm demonstrates 95.0% in accuracy, 95.9% in sensitivity, and 93.5% in specificity for NPSLE diagnosis. The present work paves a new way achieving clinical label-free serum diagnosis of rheumatic diseases by enhanced Raman fingerprints with machine learning.

Cervical cancer biomarker screening based on Raman spectroscopy and multivariate statistical analysis

Cervical cancer (CC) stands as one of the most prevalent malignancies among females, and the examination of serum tumor markers(TMs) assumes paramount significance in both its diagnosis and treatment. This research delves into the potential of combining Surface-Enhanced Raman Spectroscopy (SERS) with Multivariate Statistical Analysis (MSA) to diagnose cervical cancer, coupled with the identification of prospective serum biomarkers. Serum samples were collected from 95 CC patients and 81 healthy subjects, with subsequent MSA employed to analyze the spectral data. The outcomes underscore the superior efficacy of Partial Least Squares Discriminant Analysis (PLS-DA) within the MSA framework, achieving predictive accuracy of 97.73 %, and exhibiting sensitivities and specificities of 100 % and 95.83 % respectively. Additionally, the PLS-DA model yields a Variable Importance in Projection (VIP) list, which, when coupled with the biochemical information of characteristic peaks, can be utilized for the screening of biomarkers. Here, the Random Forest (RF) model is introduced to aid in biomarker screening. The two findings demonstrate that the principal contributing features distinguishing cervical cancer Raman spectra from those of healthy individuals are located at 482, 623, 722, 956, 1093, and 1656 cm-1, primarily linked to serum components such as DNA, tyrosine, adenine, valine, D-mannose, and amide I. Predictive models are constructed for individual biomolecules, generating ROC curves. Remarkably, D-mannose of V (C-N) exhibited the highest performance, boasting an AUC value of 0.979. This suggests its potential as a serum biomarker for distinguishing cervical cancer from healthy subjects.

Direct comparison of different protocols to obtain surface enhanced Raman spectra of human serum

Label-free Surface Enhanced Raman Spectroscopy (SERS) is a rapid technique that has been extensively applied in clinical diagnosis and biomedicine for the analysis of biofluids. The purpose of this approach relies on the ability to detect specific "metabolic fingerprints" of complex biological samples, but the full potential of this technique in diagnostics is yet to be exploited, mainly because of the lack of common analytical protocols for sample preparation and analysis. Variation of experimental parameters, such as substrate type, laser wavelength and sample processing can greatly influence spectral patterns, making results from different research groups difficult to compare. This study aims at making a step toward a standardization of the protocols in the analysis of human serum samples with Ag nanoparticles, by directly comparing the SERS spectra obtained from five different methods in which parameters like laser power, nanoparticle concentration, incubation/deproteinization steps and type of substrate used vary. Two protocols are the most used in the literature, and the other three are "in-house" protocols proposed by our group; all of them are employed to analyze the same human serum sample. The experimental results show that all protocols yield spectra that share the same overall spectral pattern, conveying the same biochemical information, but they significantly differ in terms of overall spectral intensity, repeatability, and preparation steps of the sample. A Principal Component Analysis (PCA) was performed revealing that protocol 3 and protocol 1 have the least variability in the dataset, while protocol 2 and 4 are the least repeatable.

Non-invasive screening and subtyping for breast cancer by serum SERS combined with LGB-DNN algorithms

Early detection of breast cancer and its molecular subtyping is crucial for guiding clinical treatment and improving survival rate. Current diagnostic methods for breast cancer are invasive, time consuming and complicated. In this work, an optical detection method integrating surface-enhanced Raman spectroscopy (SERS) technology with feature selection and deep learning algorithm was developed for identifying serum components and building diagnostic model, with the aim of efficient and accurate noninvasive screening of breast cancer. First, the high quality of serum SERS spectra from breast cancer (BC), breast benign disease (BBD) patients and healthy controls (HC) were obtained. Chi-square tests were conducted to exclude confounding factors, enhancing the reliability of the study. Then, LightGBM (LGB) algorithm was used as the base model to retain useful features to significantly improve classification performance. The DNN algorithm was trained through backpropagation, adjusting the weights and biases between neurons to improve the network's predictive ability. In comparison to traditional machine learning algorithms, this method provided more accurate information for breast cancer classification, with classification accuracies of 91.38 % for BC and BBD, and 96.40 % for BC, BBD, and HC. Furthermore, the accuracies of 90.11 % for HR+/HR- and 88.89 % for HER2+/HER2- can be reached when evaluating BC patients' molecular subtypes. These results demonstrate that serum SERS combined with powerful LGB-DNN algorithm would provide a supplementary method for clinical breast cancer screening.

Evaluating the accuracy of Raman spectroscopy in differentiating leukemia patients from healthy individuals: a systematic review and meta-analysis

PURPOSE

To assess the accuracy of Raman spectroscopy in distinguishing between patients with leukemia and healthy individuals.

METHOD

PubMed, Embase, Web of Science, Cochrane Library, and CNKI databases were searched for relevant articles published from inception of the respective database to November 1, 2023. The pooled sensitivity (SEN), specificity (SPE), diagnostic odds ratio (DOR), positive likelihood ratio (PLR), negative likelihood ratio (NLR), were calculated along with their corresponding 95% confidence intervals (CI). A summary comprehensive receiver operating characteristic curve (SROC) was constructed and the area under the curve (AUC) was calculated. The degree of heterogeneity was tested and analyzed.

RESULTS

Fifteen groups of original studies from 13 articles were included. The pooled SEN and SPE were 0.93 (95% CI, [0.92 -0.93]) and 0.91(95% CI, [0.90-0.92]), respectively. The DOR was 613.01 (95%CI, [270.79-1387.75]), and the AUC was 0.99. The Deeks' funnel plot asymmetry test indicated no significant publication bias among the included studies (bias coefficient, 40.80; P = 0.13 <0.10). The meta-regression analysis findings indicated that the observed heterogeneity could be attributed to variations in sample categories and Raman spectroscopy techniques.

CONCLUSION

We confirmed that Raman spectroscopy has good accuracy in differentiating patients with leukemia from healthy individuals, and may become a means of leukemia screening in clinical practice. In the case of analysis based on live cells using surface-enhanced Raman spectroscopy (SERS) improved diagnostic efficacy was observed.

Molecular separation-assisted label-free SERS combined with machine learning for nasopharyngeal cancer screening and radiotherapy resistance prediction

Nasopharyngeal cancer (NPC) is a malignant tumor with high prevalence in Southeast Asia and highly invasive and metastatic characteristics. Radiotherapy is the primary strategy for NPC treatment, however there is still lack of effect method for predicting the radioresistance that is the main reason for treatment failure. Herein, the molecular profiles of patient plasma from NPC with radiotherapy sensitivity and resistance groups as well as healthy group, respectively, were explored by label-free surface enhanced Raman spectroscopy (SERS) based on surface plasmon resonance for the first time. Especially, the components with different molecular weight sizes were analyzed via the separation process, helping to avoid the possible missing of diagnostic information due to the competitive adsorption. Following that, robust machine learning algorithm based on principal component analysis and linear discriminant analysis (PCA-LDA) was employed to extract the feature of blood-SERS data and establish an effective predictive model with the accuracy of 96.7% for identifying the radiotherapy resistance subjects from sensitivity ones, and 100% for identifying the NPC subjects from healthy ones. This work demonstrates the potential of molecular separation-assisted label-free SERS combined with machine learning for NPC screening and treatment strategy guidance in clinical scenario.

Precision classification and quantitative analysis of bacteria biomarkers via surface-enhanced Raman spectroscopy and machine learning

The SERS spectra of six bacterial biomarkers, 2,3-DHBA, 2,5-DHBA, Pyocyanin, lipoteichoic acid (LTA), Enterobactin, and β-carotene, of various concentrations, were obtained from silver nanorod array substrates, and the spectral peaks and the corresponding vibrational modes were identified to classify different spectra. The spectral variations in three different concentration regions due to various reasons have imposed a challenge to use classic calibration curve methods to quantify the concentration of biomarkers. Depending on baseline removal strategy, i.e., local or global baseline removal, the calibration curve differed significantly. With the aid of convolutional neural network (CNN), a two-step process was established to classify and quantify biomarker solutions based on SERS spectra: using a specific CNN model, a remarkable differentiation and classification accuracy of 99.99 % for all six biomarkers regardless of the concentration can be achieved. After classification, six regression CNN models were established to predict the concentration of biomarkers, with coefficient of determination R2 > 0.97 and mean absolute error (MAE) < 0.27. The feature of important calculations indicates the high classification and quantification accuracies were due to the intrinsic spectral features in SERS spectra. This study showcases the synergistic potential of SERS and advanced machine learning algorithms and holds significant promise for bacterial infection diagnostics.

Multi-branch attention Raman network and surface-enhanced Raman spectroscopy for the classification of neurological disorders

Surface-enhanced Raman spectroscopy (SERS), a rapid, low-cost, non-invasive, ultrasensitive, and label-free technique, has been widely used in-situ and ex-situ biomedical diagnostics questions. However, analyzing and interpreting the untargeted spectral data remains challenging due to the difficulty of designing an optimal data pre-processing and modelling procedure. In this paper, we propose a Multi-branch Attention Raman Network (MBA-RamanNet) with a multi-branch attention module, including the convolutional block attention module (CBAM) branch, deep convolution module (DCM) branch, and branch weights, to extract more global and local information of characteristic Raman peaks which are more distinctive for classification tasks. CBAM, including channel and spatial aspects, is adopted to enhance the distinctive global information on Raman peaks. DCM is used to supplement local information of Raman peaks. Autonomously trained branch weights are applied to fuse the features of each branch, thereby optimizing the global and local information of the characteristic Raman peaks for identifying diseases. Extensive experiments are performed for two different neurological disorders classification tasks via untargeted serum SERS data. The results demonstrate that MBA-RamanNet outperforms commonly used CNN methods with an accuracy of 88.24% for the classification of healthy controls, mild cognitive impairment, Alzheimer's disease, and Non-Alzheimer's dementia; an accuracy of 90% for the classification of healthy controls, elderly depression, and elderly anxiety.

Spontaneous-stimulated Raman co-localization dual-modal analysis approach for efficient identification of tumor cells

The study of highly heterogeneous tumor cells, especially acute myeloid leukemia (AML) cells, usually relies on invasive analytical methods such as morphology, immunology, cytogenetics, and molecular biology classification, which are complex and time-consuming to perform. Mortality is high if patients are not diagnosed in a timely manner, so rapid label-free analysis of gene expression and metabolites within single-cell substructures is extremely important for clinical diagnosis and treatment. As a label-free and non-destructive vibrational detection technique, spontaneous Raman scattering provides molecular information across the full spectrum of the cell but lacks rapid imaging localization capabilities. In contrast, stimulated Raman scattering (SRS) provides a high-speed, high-resolution imaging view that can offer real-time subcellular localization assistance for spontaneous Raman spectroscopic detection. In this paper, we combined multi-color SRS microscopy with spontaneous Raman to develop a co-localized Raman imaging and spectral detection system (CRIS) for high-speed chemical imaging and quantitative spectral analysis of subcellular structures. Combined with multivariate statistical analysis methods, CRIS efficiently differentiated AML from normal leukocytes with an accuracy of 98.1 % and revealed the differences in the composition of nuclei and cytoplasm of AML relative to normal leukocytes. Compared to conventional Raman spectroscopy blind sampling without imaging localization, CRIS increased the efficiency of single-cell detection by at least three times. In addition, using the same approach for further identification of AML subtypes M2 and M3, we demonstrated that intracytoplasmic differential expression of proteins is a marker for their rapid and accurate classifying. CRIS analysis methods are expected to pave the way for clinical translation of rapid tumor cell identification.

Rapid and accurate identification of stem cell differentiation stages via SERS and convolutional neural networks

Monitoring the transition of cell states during induced pluripotent stem cell (iPSC) differentiation is crucial for clinical medicine and basic research. However, both identification category and prediction accuracy need further improvement. Here, we propose a method combining surface-enhanced Raman spectroscopy (SERS) with convolutional neural networks (CNN) to precisely identify and distinguish cell states during stem cell differentiation. First, mitochondria-targeted probes were synthesized by combining AuNRs and mitochondrial localization signal (MLS) peptides to obtain effective and stable SERS spectra signals at various stages of cell differentiation. Then, the SERS spectra served as input datasets, and their distinctive features were learned and distinguished by CNN. As a result, rapid and accurate identification of six different cell states, including the embryoid body (EB) stage, was successfully achieved throughout the stem cell differentiation process with an impressive prediction accuracy of 98.5%. Furthermore, the impact of different spectral feature peaks on the identification results was investigated, which provides a valuable reference for selecting appropriate spectral bands to identify cell states. This is also beneficial for shortening the spectral acquisition region to enhance spectral acquisition speed. These results suggest the potential for SERS-CNN models in quality monitoring of stem cells, advancing the practical applications of stem cells.

Microfluidic-SERS Technologies for CTC: A Perspective on Clinical Translation

Enumeration and phenotypic profiling of circulating tumor cells (CTCs) provide critical information for clinical diagnosis and treatment monitoring in cancer. To achieve this goal, an integrated system is needed to efficiently isolate CTCs from patient samples and sensitively evaluate their phenotypes. Such integration would comprise a high-throughput single-cell processing unit for the isolation and manipulation of CTCs and a sensitive and multiplexed quantitation unit to detect clinically relevant signals from these cells. Surface-enhanced Raman scattering (SERS) has been used as an analytical method for molecular profiling and in vitro cancer diagnosis. More recently, its multiplexing capability and power to create distinct molecular signatures against their targets have garnered attention. Here, we share our insights into the combined power of microfluidics and SERS in realizing CTC isolation, enumeration, and detection from a clinical translation perspective. We highlight the key operational factors in CTC microfluidic processing and SERS detection from patient samples. We further discuss microfluidic-SERS integration and its clinical utility as a paradigm shift in clinical CTC-based cancer diagnosis and prognostication. Finally, we summarize the challenges and attempt to look forward to what lies ahead of us in potentially translating the technique into real clinical applications.

SERS sensing for cancer biomarker: Approaches and directions

These days, cancer is thought to be more than just one illness, with several complex subtypes that require different screening approaches. These subtypes can be distinguished by the distinct markings left by metabolites, proteins, miRNA, and DNA. Personalized illness management may be possible if cancer is categorized according to its biomarkers. In order to stop cancer from spreading and posing a significant risk to patient survival, early detection and prompt treatment are essential. Traditional cancer screening techniques are tedious, time-consuming, and require expert personnel for analysis. This has led scientists to reevaluate screening methodologies and make use of emerging technologies to achieve better results. Using time and money saving techniques, these methodologies integrate the procedures from sample preparation to detection in small devices with high accuracy and sensitivity. With its proven potential for biomedical use, surface-enhanced Raman scattering (SERS) has been widely used in biosensing applications, particularly in biomarker identification. Consideration was given especially to the potential of SERS as a portable clinical diagnostic tool. The approaches to SERS-based sensing technologies for both invasive and non-invasive samples are reviewed in this article, along with sample preparation techniques and obstacles. Aside from these significant constraints in the detection approach and techniques, the review also takes into account the complexity of biological fluids, the availability of biomarkers, and their sensitivity and selectivity, which are generally lowered. Massive ways to maintain sensing capabilities in clinical samples are being developed recently to get over this restriction. SERS is known to be a reliable diagnostic method for treatment judgments. Nonetheless, there is still room for advancement in terms of portability, creation of diagnostic apps, and interdisciplinary AI-based applications. Therefore, we will outline the current state of technological maturity for SERS-based cancer biomarker detection in this article. The review will meet the demand for reviewing various sample types (invasive and non-invasive) of cancer biomarkers and their detection using SERS. It will also shed light on the growing body of research on portable methods for clinical application and quick cancer detection.

Stratification of tumour cell radiation response and metabolic signatures visualization with Raman spectroscopy and explainable convolutional neural network

Reprogramming of cellular metabolism is a driving factor of tumour progression and radiation therapy resistance. Identifying biochemical signatures associated with tumour radioresistance may assist with the development of targeted treatment strategies to improve clinical outcomes. Raman spectroscopy (RS) can monitor post-irradiation biomolecular changes and signatures of radiation response in tumour cells in a label-free manner. Convolutional Neural Networks (CNN) perform feature extraction directly from data in an end-to-end learning manner, with high classification performance. Furthermore, recently developed CNN explainability techniques help visualize the critical discriminative features captured by the model. In this work, a CNN is developed to characterize tumour response to radiotherapy based on its degree of radioresistance. The model was trained to classify Raman spectra of three human tumour cell lines as radiosensitive (LNCaP) or radioresistant (MCF7, H460) over a range of treatment doses and data collection time points. Additionally, a method based on Gradient-Weighted Class Activation Mapping (Grad-CAM) was used to determine response-specific salient Raman peaks influencing the CNN predictions. The CNN effectively classified the cell spectra, with accuracy, sensitivity, specificity, and F1 score exceeding 99.8%. Grad-CAM heatmaps of H460 and MCF7 cell spectra (radioresistant) exhibited high contributions from Raman bands tentatively assigned to glycogen, amino acids, and nucleic acids. Conversely, heatmaps of LNCaP cells (radiosensitive) revealed activations at lipid and phospholipid bands. Finally, Grad-CAM variable importance scores were derived for glycogen, asparagine, and phosphatidylcholine, and we show that their trends over cell line, dose, and acquisition time agreed with previously established models. Thus, the CNN can accurately detect biomolecular differences in the Raman spectra of tumour cells of varying radiosensitivity without requiring manual feature extraction. Finally, Grad-CAM may help identify metabolic signatures associated with the observed categories, offering the potential for automated clinical tumour radiation response characterization.

Deep learning vs. robust federal learning for distinguishing adrenal metastases from benign lesions with multi-phase CT images

Background

Differentiating adrenal adenomas from metastases poses a significant challenge, particularly in patients with a history of extra-adrenal malignancy. This study investigates the performance of three-phase computed tomography (CT) based robust federal learning algorithm and traditional deep learning for distinguishing metastases from benign adrenal lesions.

Material and methods

This retrospective analysis includes 1187 instances who underwent three-phase CT scans between January 2008 and March 2021, comprising 720 benign lesions and 467 metastases. Utilizing the three-phase CT images, both a Robust Federal Learning Signature (RFLS) and a traditional Deep Learning Signature (DLS) were constructed using the Least Absolute Shrinkage and Selection Operator (LASSO) logistic regression. Their diagnostic capabilities were subsequently validated and compared using metrics such as the Areas Under the Receiver Operating Curve (AUCs), Net Reclassification Improvement (NRI), and Decision Curve Analysis (DCA).

Results

Compared with DLS, the RFLS showed better capability in distinguishing metastases from benign adrenal lesions (average AUC: 0.816 vs.0.798, NRI = 0.126, P < 0.072; 0.889 vs.0.838, NRI = 0.209, P < 0.001; 0.903 vs.0.825, NRI = 0.643, p < 0.001) in the four-testing cohort, respectively. DCA showed that the RFLS added more net benefit than DLS for clinical utility. Moreover, Comparison with state-of-the-art federal learning methods, the results once again confirmed that the RFLS significantly improved the diagnostic performance based on three-phase CT (AUC: AP, 0.727 vs. 0.757 vs. 0.739 vs. 0.796; PCP, 0.781 vs. 0.851 vs. 0.790 vs. 0.882; VP, 0.789 vs. 0.814 vs. 0.779 vs. 0.886).

Conclusion

RFLS was superior to DLS for preoperative distinguishing metastases from benign adrenal lesions with multi-phase CT Images.

Label-free screening of common urinary system tumors from blood plasma based on surface-enhanced Raman spectroscopy

BACKGROUND

The incidence of common urinary system tumors has been rising rapidly in recent years, and most urinary system-derived tumors lack specific biomarkers.

OBJECTIVES

To explore the efficacy of surface-enhanced Raman spectroscopy (SERS) of blood plasma in screening three common urinary system tumors, including bladder cancer (BC), prostate cancer (PCa), and renal cell carcinoma (RCC).

METHODS

SERS plasma spectra from 125 plasma samples, including 25 PCa, 38 RCC, 24 BC patients, and 38 normal volunteers, were collected. All candidates had no other comorbidities. The Diagnosis was based on the combination of Principal Component Analysis (PCA) and Linear Discriminant Analysis (LDA), and the effectiveness of the diagnostic algorithms was verified using the Receiver Operating Characteristic Curve (ROC).

RESULTS

There are significant differences in SERS signals between PCa, BC, RCC, and normal plasma, especially at 639, 889, 1010, 1136, and 1205 cm-1. The PCA-LDA results show that high sensitivity (100 %), specificity (100 %), and accuracy (100 %) could be achieved for screening the PCa, RCC, BC group vs. the normal group, the PCa group vs. the BC and RCC group, respectively. The diagnostic sensitivity, specificity, and accuracy for the BC group vs. the RCC group are 79.2 %, 71.1 %, and 75.15 %, respectively. The integrated area under the ROC curve (AUC) is 1.0, 1.0, and 1.0 for the PCa, RCC, and BC group vs. the normal group, respectively. The AUC of the PCa group vs. the BC group and RCC group and the BC group vs. the RCC group are 1.0, 1.0, and 0.842, respectively.

CONCLUSIONS

Label-free plasma-SERS technology with PCA-LDA analysis could be a useful screening method for detecting urinary system tumors (PCa, RCC, and BC) in this exploratory study.