Adjustable spot wide-field Raman spectroscopy combined with machine learning for accurate classification of breast cancer cells.

Cell heterogeneity presents significant challenges for the accurate diagnosis and classification of breast cancer at the single-cell level using Raman spectroscopy. Traditional Raman spectroscopy systems are limited by their small laser spot sizes, which restrict them to capturing localized biochemical information within cells. To address this limitation, we propose a wide-field Raman spectroscopy system with an adjustable spot size (WFRS-AS), capable of collecting Raman signals from entire cells. This approach provides a more comprehensive biochemical fingerprint and effectively reduces the impact of cell heterogeneity. Using supervised classification methods, we compared breast cell spectra acquired by conventional Raman systems and the WFRS-AS system. The results indicate that, when combined with the Support Vector Machine (SVM) algorithm, WFRS-AS achieves 98.18 % accuracy in breast cancer cell diagnosis, representing an improvement of approximately 6.95 %, and 99.26 % accuracy in classifying five breast cell lines, representing an improvement of about 2.83 %. This indicates that integrating WFRS-AS technology with machine learning algorithms offers a powerful and efficient strategy for more accurate and effective breast cancer diagnosis at the single-cell level.