Prostate cancer is the most frequent diagnosed cancers among men. When prostate cancer occurs, the cancer does not result in only one or few localized malignant tumor, but is generally spread within the whole prostate.
In order to counteract the very high level of heterogeneities exhibited by prostate tissues, we developed a method for high-resolution co-registration of Raman spectroscopy with prostate cancer diagnosis.
Raman spectra were acquired on fresh ex vivo prostate within 2 hours after radical prostatectomy using a multi-wavelength hand-held contact probe. After the measurements, the prostate was reintegrated to the usual pathological workflow: formalin fixated and paraffin embedded (FFPE), and prepared for microscope histopathological analyses. The precise reconstruction of the prostate slice with hematoxylin and eosin (H and E) tissue allows the spatial correlation of the measured area (0.2 mm2) with the correspondent histopathological information, for point-by-point diagnosis determination. The tissue was classified into groups (normal/cancer) and subgroups according to the percentage of benign glands, stroma or cancer.
Different machine learning algorithms were tested to classify the spectra with increasing levels of categorization. Preliminary results showed that Raman spectroscopy is capable of detecting prostate cancer with an accuracy >90%. In addition, high percentages of stroma (vs. glands) have been correlated with spectral signature of collagen, which is the main constituent of extracellular matrix.