6533b85ffe1ef96bd12c2291

RESEARCH PRODUCT

Classification of spectra and search for biomarkers in prostate tumours from proton nuclear magnetic resonance spectroscopy

subject

description

Prostate cancer is the most common cancer in men over 50 years. Current detection methods either lack sensitivity or specificity or are unpleasant for the patient. Magnetic resonance spectroscopy allows the study of metabolism in vivo. The use of a high field machine (≥3T) has allowed us to dispense with the use of an endorectal coil, which is particularly uncomfortable for the patient. The objective of this thesis is to create an automatic method to detect cancer by processing data obtained through magnetic resonance spectroscopy MRS is a complex phenomenon, very sensitive to acquisition conditions. Firstly, we have studied how to improve and optimise signal acquisition. However, even with a very good quality signal, it must still undergo further post-processing to be analysed automatically by a classification method. Further work was therefore needed to investigate which postprocessing steps were required in order to optimize the spectra for classification. We then investigated the optimal classification method for this problem. A particular set of steps (signal acquisition, processing and spectral classification data) allows us to highlight the presence of prostate tumors with an overall error rate of less than 12%. In a second step, we searched for new biomarkers within the spectra. These biomarkers could be a metabolite or a specific frequency range corresponding to several metabolites. We did not find any additional significant attributes other than choline and citrate, however, some frequency bands seem to participate in improving the error rate. Finally, we expanded our investigation by attempting to apply these techniques to the rat. Technical constraints related to acquisition did not allow us to obtain a sufficient number of spectra in the pre-clinical cases. Nonetheless, we have validated the feasibility of MRS in rodents and its relevance in the brain. The technique, however, must be improved in order to be validated in the case of prostate cancer in rats.