Prostate cancer sensor : combining Raman spectroscopy and tactile resonance technology

Abstract: Prostate cancer (PCa) is the most common cancer type among men in Sweden. The most prevalent curative treatment for PCa is radical prostatectomy (RP), i.e., complete surgical removal of the prostate. Unfortunately, cancer cells are found near the resection surface in 35 % of the RP patients. This indicates an increased risk of PCa reccurence.Our main objective is to develop a novel medical instrument for detecting PCa. The idea is to combine the techniques of Raman spectroscopy (RS) and the tactile resonance method (TRM) into one integrated instrument. The TRM would provide a swift and gentle survey of the region of interest, while the RS adds detailed information of the molecular content where malignancy is suspected. The dual mode instrument could be well suited for detecting and locating tumour cells in the surgical margin during RP. The studies included in this thesis are important steps towards this objective.Paper A investigated how the two data sets from each of the technologies could be compared and combined for tissue characterisation. The data set of RS was a spectrum with peaks characteristic to the sample's molecular content. The TRM output variable was a scalar value related to the sample stiness. The data sets could be compared and combined by applying principal component analysis (PCA) to the RS spectra followed by an hierarchical cluster analysis (HCA). A linear regression analysis showed that the PCs explained 67% of the stiffness variations. HCA was used to classify each RS measurement into groups consisting of similar measurements. The TRM's sensitivity and specificity of classifying these groups were evaluated by ROC curves and the area under the curve (AUC). The harder group could successfully be discriminated from the softer groups (AUC = 0.99).Paper B used support vector machines (SVM) as a method to classify and differentiate porcine and human prostate tissue types using the combined data sets. Prostate tissue is highly inhomogenous, with streaks of small anatomical structures. The analysis was evaluated within areas of three levels of homogenity, to avoid mismatching the measured tissue. The tissue homogenity was evaluated within the RS measurement area and the tissue type was set to the main histological content. Areas in which no single tissue type surpassed the threshold level were excluded from the analysis. The cross-validation accuracy for determining the tissues types within homogenous (main tissue type > 83%) porcine samples was 82% using TRM data alone. It increased to 87% while using the combined data sets of TRM and RS. For discerning healthy and cancerous human prostate tissue, the cross-validation accuracy was 67% and 77% for TRM alone versus TRM and RS combined.Paper C covered a number of design considerations which have to be addressed during the combination of TRM and RS. The effects of attaching an RS probe into a tubular TRM element were investigated. We investigated the temperature increase caused by the laser illumination from the RS and its eect on the TRM measurement parameter Δf. We also investigated if and how RS could be performed under ambient light. A thin RS probe and a small amount of rubber latex was preferable for attaching the RS probe inside the TRM sensor. The temperature rise of the TRM sensor due to a fibreoptic NIR-RS at 270 mWduring 20 s was less than 2ºC. The variation of Δf during a 5ºC temperature change was approximately 20 Hz. This is small compared to previous in vitro TRM studies. Fibre-optic NIR-RS was feasible in a dimmed bright environment using a small light shield and automatic subtraction of a pre-recorded contaminant spectrum. The results of these studies indicate how the hardware and and software could be combined into one integrated probe for prostate cancer detection.