For a study, researchers used Surface-enhanced Raman spectroscopy (SERS) and multivariate statistical techniques to identify human plasma for screening for prostate cancer (PCa) and benign prostatic hyperplasia (BPH).

The test was designed to identify 106 plasma samples from 39 normal persons, 26 PCa patients, and 41 BPH patients. The differential spectrogram helps to initially separate the cancer group from the normal group since it shows significant variations in peak intensity at 495, 636, 1,135, 1,205, and 1,675 cm-1. The spectrum data was then analyzed using multivariate statistical approaches, such as the principle component analysis (PCA) and linear discriminant analysis (LDA) diagnostic algorithms, as well as the recursive weighted partial least square (PLS) method and support vector machine (SVM) algorithm.

The classification accuracy of PCA-LDA was 96.80% and 97.50% for PCa vs. normal group and BPH vs. normal group, respectively, whereas the classification accuracy of PLS-SVM was 100.00% and 100.00%, respectively. PCA-LDA had a sensitivity, specificity, and accuracy of 65.40%, 75.60%, and 71.06%, respectively, in diagnosing PCa and BPH. Its area under the curve (AUC) value of the receiver operating characteristic (ROC) curve was 0.788. In contrast, PLS-SVM had a sensitivity, specificity, and accuracy of 88.46%, 87.80%, and 88.06%, respectively. Its AUC value was 0.881.

The diagnostic outcomes of PLS-SVM were superior to those of PCA-LDA, supporting the claim that the PLS-SVM algorithm had more potential for PCa pre-diagnosis and screening than the PCA-LDA algorithm.