Wavelet-based denoised and feature extraction of NMR spectroscopy based on pattern recognition

According to the shortages of application of MRS and MRI to the clinical cancer diagnosis, this paper bring forward a effective method to analyze and process the raw data of nuclear magnetic resonance based on wavelet transform and pattern recognition technologies. Aiming to the characteristics of FID signals and MRS, the paper performed de-nosing of FID and MRS data using wavelet threshold to obtain the better MRS spectrums firstly, and then extracted the feature of certain cancer from MRS spectrums based on Independent Component Analysis (ICA) and Support Vector Machine (SVM). Comparing with the de-nosing effect of conventional wavelet basis functions, a new designed wavelet filter set showed better performance. Experiments were carried out on small amount of low SNR dataset and the results were given, which showed the improved effect on de-nosing and feature extraction.