Abstract:
Early cancer detection remains a major challenge, with patient outcomes strongly dependent on identifying disease at its earliest stages. Raman spectroscopy offers a promising, label-free approach for blood-based diagnostics, but its clinical translation is limited by baseline distortions and variability that compromise reproducibility.
In this dissertation, I introduce the Baseline Harmonization Algorithm (BHA), a new preprocessing method that standardizes baseline variability across entire datasets while preserving true biochemical signals. I also establish a robust end-to-end workflow—from plasma collection and handling, through Raman acquisition, to baseline correction and PCA–SVM classification—designed to maximize reproducibility in realistic clinical conditions. Using extensive synthetic and clinical datasets from breast and colon cancer patients, this pipeline achieves diagnostic performance with sensitivity and specificity of ~0.95.
In addition, I develop 3D-nanoprinted hollow-core waveguides enabling direct Raman interrogation of whole blood via integrated corpuscular filtration, eliminating the need for centrifugation and enabling rapid, microscope-compatible liquid biopsy measurements.
Together, these computational, analytical, and photonic advances establish a pathway toward reliable, minimally invasive, and clinically scalable plasma-based cancer diagnostics.

Bio:
Elisa Grassi holds a Bachelor's degree in Biomedical Engineering from the University of Pavia and a Master's in Bioengineering from KAUST, where she worked on stimulated Raman scattering for biomedical applications. Currently a Ph.D. candidate in Prof. Carlo Liberale’s lab at KAUST, her research focuses on Raman spectroscopy for early cancer detection via liquid biopsies, with a dual emphasis on algorithm development and microfabrication. She develops advanced baseline correction algorithms to improve spectral interpretation and builds 3D-printed optical structures via two-photon lithography to enhance Raman sensitivity. Her interdisciplinary work spans signal processing, optics, and diagnostic engineering.