New scanning techniques and statistical algorithms for early diagnosis of cancer

New techniques for the diagnosis of cancer are being developed on the world-leading ALICE accelerator at Daresbury using a state-of-the-art scanning near field optical microscope (SNOM) installed on an infrared free electron laser (IR FEL) by the Liverpool group. In this PhD project we will explore the potential of these techniques to yield a diagnostic for several types of cancer in collaboration with clinical colleagues from the Liverpool Women’s hospital. The project will also explore whether the analysis of IR FEL-SNOM images can yield common features in tissue. Novel statistical algorithms will be used to identify the set of infrared wavelengths that will be subsequently used to classify a set of benign/pre-cancer and cancer tissue.

Early detection of disease and accurate prediction of prognosis are essential to improve patient outcomes for the greatest chance of cure, as well as to optimise healthcare resources. These goals can be achieved by identifying key biomarkers that are a characteristic of a particular disease, and which can be observed from an early stage in the development of the disease. Knowledge of specific biomarkers can improve diagnostic accuracy and predict the most likely response to a particular treatment. Problems in identifying such biomarkers include non-trivial extraction from large amounts of multi-parameter data, signal extraction from noise, and determining specific correlations across parameters in data - which can itself be classed as a biomarker. For example, we have found that discrete wavelength pairing within infrared (IR) absorption spectroscopy data from cancerous tissues shows promise as a sensitive biomarker of disease.

This PhD project will investigate the use of variable selection algorithms to identify the set of infrared wavelengths that will be subsequently used to classify a set of benign/pre-cancer and cancer tissue. Fourier transform infrared spectroscopy (FTIR) and scanning near-field optical spectroscopy (IR-SNOM) data will be obtained and analysed. The correlation matrix generated from SNOM images for the selected range of wavelengths will be used for classification.

The long-term aim is to establish protocols to enhance the diagnosis of potentially malignant abnormal lesions identified in long term monitoring programmes and to develop instruments for in-theatre diagnostics with the potential to identify tumour cells during surgery.