RESUMO
Digital rectal examination (DRE) was the primary means to detect prostate diseases. The DRE has a high variability as it is based manly in the tactile sensitivity and expertise of the examiner. The prostate-specific antigen (PSA) test was initially developed for surveillance of prostate cancer and later it was also used as a diagnosis test. Raman spectroscopy is a powerful analytical technique that can measure the chemical composition of complex biological samples, such as body fluids. Biochemical changes caused by diseases can lead to significant changes in the Raman spectra. This study aimed to identify the differences in the Raman spectra of serum samples with normal and altered PSA values and correlate these differences by using multivariate techniques (principal component analysis - PCA and partial least squares regression - PLS). A total of 321 spectra were collected from 108 subjects. Two hundred and seventy were obtained from 91 non-altered PSA samples and 51 spectra from 17 samples with altered PSA. Each spectrum acquired was standardized to the area under the curve (1-norm). Discriminating and quantitative models employing PCA and PLS were developed. The PCA analyses showed 85.7% predictive power (87.41% sensitivity and 76.47% specificity). The PLS test showed a near-perfect sensitivity (98.51%) and an intermediate specificity (62.75%). The quantitative model through PLS regression showed a good correlation between PSA values and the spectral features (r = 0.605). This preliminary study suggests that Raman spectroscopy could be efficiently used for screening patients with altered PSA as well as for follow-up of the treatment of the prostate cancer by using initially the PLS to identify the possible presence of the prostate cancer and later on use de PCA to confirm the diagnosis.