Raman Spectroscopy, Surface-Enhanced Raman Spectroscopy (SERS) and Tip-Enhanced Raman Spectroscopy (TERS) for advanced diagnostics

Cancer is a global health burden, highlighting the limitations of traditional diagnostic approaches, which are often invasive, subjective, and limited in their capacity for timely detection. Addressing these challenges through the development of innovative diagnostic tools that enable accurate, non-invasive and objective assessment is crucial for advancing early diagnosis, optimizing treatment strategies, and improving patient outcomes. This thesis explores the transformative potential of advanced Raman spectroscopy techniques as innovative technologies capable of revolutionizing cancer diagnostics, emphasizing their ability to provide unique biochemical fingerprinting in a non-invasive and real-time manner. Raman spectroscopy is introduced as a label-free technique, with its fundamental principles, instrumentation, and essential data pre-processing steps described in detail. The technique is widely used for the analysis of biological samples—from cells to tissues to biofluids—due to its ability to provide detailed molecular and spectral fingerprints, revealing biochemical composition and enabling deeper understanding of biological systems. Owing to these strengths, Raman spectroscopy can identify subtle biochemical alterations that differentiate healthy cells from cancerous ones, while simultaneously providing insights into the distinct metabolic and genetic profiles characteristic of cancer cells. The discussion also emphasizes the critical role of advanced machine learning algorithms—including both unsupervised and supervised approaches—in interpreting complex spectral data for accurate disease classification. Additionally, the principal applications of Raman spectroscopy in cancer diagnostics are presented across in vitro, ex vivo, and in vivo settings. The initial part of the research focuses on the application of artificial intelligence–enhanced Raman spectroscopy in two distinct cancer case studies—hepatocellular carcinoma and non- small cell lung cancer—highlighting the broad applicability and versatility of this combined diagnostic strategy. The first study assesses primary human non-tumor and tumor liver cells from a patient with hepatocellular carcinoma using Raman analysis, which successfully differentiates these cellular samples by revealing distinct molecular fingerprints that reflect altered nucleic acid content in cancer cells. For tumor cell identification and discrimination, different linear discriminant analysis (LDA)- and convolutional neural network–long short- term memory (CNN-LSTM)-based models were developed. The LDA-based models achieved classification accuracies close to 90%, while the best CNN-LSTM model reached about 93% at the single-spectrum level. Raman spectroscopy is further applied in the context of the liquid biopsy, using different suspended non-small cell lung cancer (NSCLC)-derived cell lines as models of circulating tumor cells (CTCs). This methodological approach provides a powerful platform to simulate CTC dissemination in the bloodstream and to reliably distinguish them from healthy leukocytes. Spectral analysis reveals lipid enrichment in cancer cells, while healthy cells exhibit higher levels of nucleic acids and proteins, reflecting their distinct cellular origins and functions. Beyond differentiating malignant from healthy cells, the research provides novel insights into the heterogeneity, and the metabolic and genetic landscapes of different cancer cell lines. By integrating Raman spectroscopy with a PCA-LDA algorithm, the study achieves accurate discrimination between four NSCLC cell lines and leukocytes, reaching a classification accuracy of 96.40%, highlighting its significant potential in supporting lung cancer diagnostics. Expanding on diagnostic innovations, the thesis also presents a comprehensive analysis of surface-enhanced Raman spectroscopy (SERS), detailing its electromagnetic and chemical enhancement mechanisms, as well as the strategic design of SERS substrates and functionalized nanotags to optimize sensitivity and specificity. The study further explores SERS applications in biosensing and bioimaging, enabling precise identification of biomarkers in cells, tissues, and even in vivo, with particular emphasis on the use of SERS optical fibers as an innovative platform for healthcare applications. In this context, a novel aspect of the research involves the development and evaluation of Au-coated ZnO nanopillar arrays integrated onto optical fiber tips as SERS-active platforms. The study systematically examines how fabrication strategies, structural ordering, and excitation wavelengths affect SERS performance, aiming to optimize sensitivity and reproducibility. These platforms are investigated for their potential in real-time, minimally invasive biomedical applications, highlighting their promise for broader diagnostic use. The research also extends to the application of Tip-enhanced Raman spectroscopy (TERS), outlining its fundamental principles and diagnostic applications. Specifically, a TERS as proof-of-concept study is conducted for nanoscale analysis of cancer cell membranes and, in combination with SERS, for cell surface biomarkers, providing simultaneous biochemical and topographical information. The research demonstrates label-free TERS for characterizing HepG2 hepatocellular carcinoma cell membranes, highlighting their heterogeneity. A hybrid strategy combining TERS with SERS tags is presented for the targeted detection of glypican- 3 (GPC-3), a key HCC biomarker, on cell membranes. Functionalized SERS tags specifically bind to overexpressed GPC-3, and gap-mode TERS amplifies the signals, enabling precise nanometric localization of the biomarker. In conclusion, this thesis underscores the pivotal role of advanced Raman platforms as powerful and reliable diagnostic tools, marking a significant step forward in cancer diagnostics. Their integration with artificial intelligence and nanotechnology represents a key advancement, opening the way to real-time, non-invasive, and highly personalized diagnostic strategies. These developments not only enhance the translational relevance of Raman-based technologies but also lay the foundation for their future implementation in clinical practice, where they hold great promise for improving patient outcomes and advancing the vision of precision oncology.