Method for removing noise from nuclear magnetic resonance signals and images

A nuclear magnetic resonance signal processing circuit, comprising:

• (a) means for estimating a noise variance associated with a nuclear magnetic resonance (NMR) signal;
• (b) means for sampling the nuclear magnetic resonance signal;
• (c) means for digitizing the nuclear magnetic resonance signal;
• (d) means for selecting a set of analysis sampling rates, each of the rates varying in the number of samples in frequency so that each element of the set has a compact signal representation for at least one component according to differing component relaxation times of the signal;
• (e) means for performing a stable linear time-frequency (SLTF) transform for each element of the set of sampling rates in order to determine transform coefficients for the element;
• (f) means for applying threshold criteria to the transform coefficients and setting the coefficients below a noise threshold value equal to zero for each element of the set of sampling rates in order to denoise the coefficients;
• (g) means for performing an inverse SLTF transform on the denoised coefficients for each element of the set of sampling rates in order to form a denoised signal;
• and
• (h) means for averaging the denoised SLTF transform signals to form a denoised nuclear magnetic resonance signal.

A computer software product that includes a medium readable by a processor, the medium having stored thereon a set of instructions for removing noise from a nuclear magnetic resonance signal, comprising:

• (a) a first sequence of instructions which, when executed by the processor, causes said processor to estimate a noise variance associated with a nuclear magnetic resonance (NMR) signal;
• (b) a second sequence of instructions which, when executed by the processor, causes said processor to sample the nuclear magnetic resonance signal;
• (c) a third sequence of instructions which, when executed by the processor, causes said processor to digitize the nuclear magnetic resonance signal;
• (d) a fourth sequence of instructions which, when executed by the processor, causes said processor to select a set of analysis sampling rates, each of the rates varying in the number of samples in frequency so that each element of the set has a compact signal representation for at least one component according to differing component relaxation times of the signal;
• (e) a fifth sequence of instructions which, when executed by the processor, causes said processor to perform a stable linear time-frequency (SLTF) transform for each element of the set of sampling rates in order to determine transform coefficients for the element;
• (f) a sixth sequence of instructions which, when executed by the processor, causes said processor to apply threshold criteria to the transform coefficients for each element of the set of sampling rates, setting coefficients below a noise threshold value equal to zero in order to denoise the coefficients;
• (g) a seventh sequence of instructions which, when executed by the processor, causes said processor to perform an inverse SLTF transform on the denoised coefficients for each element of the set of sampling rates in order to form a denoised signal;
• and
• (h) an eighth sequence of instructions which, when executed by the processor, causes said processor to average the denoised SLTF transform signals to form a denoised nuclear magnetic resonance signal.