Abstract:Aiming to assist the accurate diagnosis of cancer, the application research status of nuclear magnetic resonance and correlative data processing techniques were analyzed, and aiming at the problems of MRS, for example, complex technology, hard recognition by clinical doctors, low applied ability, a new scheme of NMR spectroscopy analysis and cancer feature extraction was put forward based on pattern recognition, wavelet transform and network techniques. Taking the data generated by nuclear magnetic resonance devices applied to cancer clinical diagnosis field as the research objects, including free induction decay (FID) signals and magnetic resonance spectrums, the new scheme implemented the signal decomposition and de-noising using new designed wavelet basis function and wavelet threshold de-noising techniques, and feature extraction of cancer by pattern recognition technique according to diagnosis experiences, and then integrated the final expert system on the platform of picture archiving and communication system (PACS). Effective solutions to the questions of system structure and key technologies were presented.

Nuclear magnetic resonance spectroscopy analysis and cancer feature extraction technologies