Label-free discrimination analysis of breast cancer tumor and adjacent tissues of patients after neoadjuvant treatment using Raman spectroscopy: a diagnostic study

Background and Objective: Breast-conserving surgery (BCS) plays a crucial role in breast cancer treatment, with a primary focus on ensuring cancer-free surgical margins, particularly for patients undergoing neoadjuvant treatment. After neoadjuvant treatment, tumor regression can complicate the differentiation between breast cancer and adjacent tissues. Raman spectroscopy, as a rapid and non-invasive optical technique, offers the advantage of providing detailed biochemical information and molecular signatures of internal molecular components in tissue samples. Despite its potential, there is currently no research on using label-free Raman spectroscopy to distinguish between breast cancer tumors and adjacent tissues after neoadjuvant treatment. This study intends to distinguish between cancer and adjacent tissues after neoadjuvant treatment in breast cancer through label-free Raman spectroscopy. Methods: In this study, the intraoperative frozen samples of breast cancer tumor and adjacent tissue were collected from patients who underwent neoadjuvant treatment during surgery. The samples were examined using Raman confocal microscopy, and Raman spectra were collected by LabSpec6 software. Spectra were preprocessed by Savitz-Golay filter, adaptive iterative reweighted penalized least squares and MinMax normalization method. The differences in Raman spectra between breast cancer tumor and adjacent tissues after neoadjuvant treatment were analyzed by Wilcoxon rank-sum test, with a Bonferroni correction for multiple comparisons. Based on the support vector machine (SVM) method in machine learning, a predictive model for classification was established in the total group and subgroups of different hormone receptor (HR) status, human epidermal growth factor receptor 2 (HER2) status and Ki-67 expression level. The independent test set was used to evaluate the performance of the model, and the area under curve (AUC) of the receiver operating characteristic (ROC) curve, sensitivity, specificity and accuracy of different models were obtained. Result: This study comprised 4260 Raman spectra of breast cancer tumor and adjacent frozen tissue samples from 142 breast cancer patients treated with neoadjuvant treatment. The Raman peaks associated with nucleotides and their metabolites in the Raman spectra of breast cancer tumor tissues were higher in intensities than those of adjacent tissues after neoadjuvant therapy (676 cm −1 : Bonferroni adjusted P < 0.0001; 724 cm −1 : P < 0.0001; 754 cm −1 : P < 0.0001), and the Raman peaks from amide III bands were more intense (1271 cm −1 : P < 0.01). Multivariate curve resolution- alternating least squares (MCR-ALS) decomposition of Raman spectra revealed reduced lipid content and increased collagen and nucleic acid content in breast cancer tumor tissues compared to adjacent tissues following neoadjuvant therapy. The predictive model based on the Raman spectral signature of breast cancer tumor and adjacent tissues after neoadjuvant treatment achieved an AUC of 0.98, with accuracy, sensitivity, and specificity values of 0.89, 0.97, and 0.83, respectively. The AUC of subgroup analysis according to different status of molecular pathological biomarkers was stably around 99%. Conclusion: This study demonstrated that label-free Raman spectroscopy can differentiate cancer and adjacent tissues of breast cancer patients treated with neoadjuvant therapy thorough getting the panoramic perspective of the biochemical compounds for the first time. Our study provided a novel technique for determining the margin status in BCS in breast cancer following neoadjuvant treatment rapidly and precisely.