Document Type

Theses, Ph.D


Available under a Creative Commons Attribution Non-Commercial Share Alike 4.0 International Licence

Publication Details

Submitted in partial fulfilment for the degree of Doctor of Philosophy (PhD) Technological University Dublin, School of Physics January 2017.


Cervical cancer is the second most common cancer in women worldwide. For decades, most developed countries have applied organized cytology screening programs using the Papanicolau (PAP) test to identify abnormal cases. Despite a high specificity of 95-98%, Pap test sensitivity is reported to vary greatly from 74 to 96% with constant testing needed to achieve the highest values. Semi-automated cytology screening platforms, immunocytochemistry panels and other methodologies such as human papilloma virus (HPV) testing has been developed to help reduce false negative rates. More recently, HPV testing, thus far used for triage of abnormal cytology cases and test of cure, have been recommended for primary screening. However HPV testing only informs on the presence of the virus not adding on the abnormal transformation of the cells. Its suitability for screening of younger women (years) whom present the highest rates of transient viral infection is also questionable. The potential of Raman spectroscopy has also been acknowledged with its ability of detecting spectral changes in malignant and pre-malignant cells extensively reported. In this project the potential of Raman spectroscopy for cytological diagnosis of samples from a cervical cancer screening population was assessed. Routinely used ThinPrep® glass slides were used as spectroscopy substrates in order to minimize costs and simplify sample processing. Raman spectra were recorded from single cell nuclei and subjected to multivariate statistical analysis. Different approaches were tested to minimize the glass contribution to the sample spectra and a non-negative least squares method was found to provide the best results. Normal and abnormal ThinPrep® samples were discriminated based on their biochemical fingerprint using Principal Component Analysis (PCA). Principal Component Analysis – Linear Discriminant Analysis (PCA-LDA) was further employed to build classification models using both Cervical Intraepithelial Neoplasia (CIN) and Squamous Intraepithelial Lesion (SIL) terminology. Results showed that Raman spectroscopy can be successfully applied to the study of routine cervical cytology samples from a cervical screening programme and normal and abnormal samples could be discriminated with high sensitivity and specificity rates (>95%) when tested with leave one patient out cross validation. In addition, the results suggested that HPV infection and previous disease history might be inferred from negative samples and might influence the performance of classifiers. In summary this study has shown Raman spectroscopy has potential as a screening tool for Thinprep® cervical cytology samples.


Included in

Oncology Commons