A machine learning-driven Raman spectroscopy approach for non-invasive diagnosis of non-puerperal mastitis.

Early and rapid diagnosis of non-puerperal mastitis (NPM), as well as elucidation of its specific pathological features, is of important clinical and scientific value. Peripheral blood mononuclear cells (PBMCs), which are key mediators in the inflammatory response, contribute substantially to disease onset, progression, and therapeutic effect, making them promising biomarkers for the early identification and management of inflammatory processes. Nevertheless, novel approaches for the detection and analysis of PBMCs remain urgently needed to facilitate the development of liquid biopsy strategies. In this study, we employed Raman spectroscopy to characterize molecular alterations in PBMCs derived from two distinct groups of NPM patients and healthy controls. Additionally, several machine learning algorithms, including principal component analysis (PCA), linear discriminant analysis (LDA), partial least squares discriminant analysis (PLSDA), and support vector machine (SVM), were applied to establish diagnostic prediction models for NPM, yielding area under the curve (AUC) values exceeding 0.93. Our findings indicate that PBMC-based liquid biopsy coupled with Raman spectroscopy and machine learning provides novel opportunities for the diagnosis of NPM.