There is a real need for improvements in cancer detection. Significant problems are encountered when utilising the gold standard of excisional biopsy combined with histopathology. This can include missed lesions, perforation and high levels of inter- and intra-observer discrepancies. The clinical requirements for an objective, non-invasive real time probe for accurate and repeatable measurement of tissue pathological state are overwhelming. This study has evaluated the potential for Raman spectroscopy to achieve this goal. The technique measures the molecular specific inelastic scattering of laser light within tissue, thus enabling the analysis of biochemical changes that precede and accompany disease processes. Initial work has been carried out to optimise a commercially available Raman microspectrometer for tissue measurements; to target potential malignancies with a clinical need for diagnostic improvements (oesophagus, colon, breast, and prostate) and to build and test spectral libraries and prediction algorithms for tissue types and pathologies. This study has followed rigorous sample collection protocols and histopathological analysis using a board of expert pathologists. Only the data from samples with full agreement of a homogeneous pathology have been used to construct a training data set of Raman spectra. Measurements of tissue specimens from the full spectrum of different pathological groups found in each tissue have been made. Diagnostic predictive models have been constructed and optimised using multivariate analysis techniques. They have been tested using cross-validation or leave-one-out and demonstrated high levels of discrimination between pathology groups (greater than 90% sensitivity and specificity for all tissues). However larger sample numbers are required for further evaluation. The discussions outline the likely work required for successful implementation of in vivo Raman detection of early malignancies.
You have access to this article
Please wait while we load your content...
Something went wrong. Try again?