Key points are not available for this paper at this time.
Due to the high intrusiveness of pathological diagnosis and the elusiveness of liquid biopsy, breast cancer (BC) is still in a dilemma between robustness and invasiveness. In our study, a molecular-specific diagnostic strategy was introduced for screening BC at an early stage, which utilizes surface-enhanced Raman spectroscopy (SERS) based on Ag NPs at 50–60 nm to acquire the fingerprint SERS spectra of fine needle aspiration (FNA) samples and machine learning for data mining. The SERS spectra of FNA samples from 78 patients were analyzed. Multiple machine learning algorithms including principal component analysis (PCA), principal component analysis–linear discriminant analysis (PCA-LDA), partial least-squares discriminant analysis (PLS-DA), and support vector machine (SVM) models were applied to deconstruct those SERS spectra for discrimination of different types of breast disease. Significant biochemical differences were found in SERS spectra of breast fibroadenoma, breast hyperplasia, and BC. With the SVM algorithm, the diagnostic sensitivity and specificity of BC, breast fibroadenomas, and breast hyperplasia can reach 94.74%, 83.33%, 81.82% and 86.96%, 100%, 94.00%, respectively. The hyphenated method of SERS and machine learning would re-energize FNA and enable FNA diagnosis of breast disease early and precisely, benefiting patients' treatment efficacy and patient life cycle.
Wang et al. (Fri,) studied this question.