System for magnetic resonance spectroscopy of brain tissue for pattern-based diagnostics

Abstract

A system and method for preprocessing magnetic resonance spectroscopy (MRS) data of brain tissue for pattern-based diagnostics is disclosed. The MRS preprocessing system includes an MRS preprocessing module that executes an operation that normalizes MRS spectrum data, recalibrates and scales the normalized MRS spectrum data, and then renormalizes the scaled MRS spectrum data. The resulting preprocessed MRS data is used to assist in identifying abnormalities in tissues shown in MRS scans.

1 .- 22 . (canceled) 
     
     
         23 . An MRS preprocessing system for preprocessing raw MRS spectrum data comprising:
 one or more processors; and   a plurality of modules configured to be executed by the one or more processors, the plurality of modules comprising:
 a variance-weighting module to scale the raw MRS spectrum data by using a plurality of weighting constants to generate a weighted MRS spectrum data, thereby generating a preprocessed MRS spectrum data. 
   
     
     
         24 . The system of  claim 23 , wherein each of the plurality of weighting constants correspond to a particular nuclear magnetic resonance frequency. 
     
     
         25 . The system of  claim 24 , wherein scaling the raw MRS spectrum data includes dividing each of the summary of signals at a particular nuclear magnetic resonance frequency by one of the plurality of weighting constants corresponding to that same particular nuclear magnetic resonance frequency. 
     
     
         26 . The system of  claim 23 , wherein each of the plurality of weighting constants at a particular nuclear magnetic resonance frequency represents a standard deviation from a mean value of the raw MRS spectrum data associated with a single type of tissue at the particular nuclear magnetic resonance frequency. 
     
     
         27 . The system of  claim 23 , wherein each of the plurality of weighting constants at a particular nuclear magnetic resonance frequency represents an average standard deviation taken from an average of raw MRS spectrum data associated with a plurality of types of tissue at the particular nuclear magnetic resonance frequency. 
     
     
         28 . A non-transitory machine-readable media encoded with an MRS preprocessing system to process raw MRS spectrum data, the MRS preprocessing system comprising machine readable instructions executable by at least one processor to perform the steps of:
 receiving raw MRS spectrum data at the MRS preprocessing system executing on at least one processor;   scaling the raw MRS spectrum data by using a plurality of weighting constants to generate a weighted MRS spectrum data, thereby generating a preprocessed MRS spectrum data.   
     
     
         29 . The machine-readable media of  claim 28 , the raw MRS spectrum data comprises a summary of signals produced by a tissue sample at a range of nuclear magnetic resonance frequencies. 
     
     
         30 . The machine-readable media of  claim 29 , wherein scaling the MRS spectrum data using the plurality of weighting constants further comprises enhancing the at least one of the summary of signals being representative of a particular biomarker relative to the at least another one of the summary of signals representative of random noise. 
     
     
         31 . The machine-readable media of  claim 28 , wherein each of the plurality of weighting constants corresponds to a particular nuclear magnetic resonance frequency. 
     
     
         32 . The machine-readable media of  claim 31 , wherein scaling the MRS spectrum data includes dividing each of the summary of signals at a particular nuclear magnetic resonance frequency by one of the plurality of weighting constants corresponding to that same particular nuclear magnetic resonance frequency. 
     
     
         33 . The machine-readable media of  claim 29 , wherein each of the plurality of weighting constants at a particular nuclear magnetic resonance frequency represents a standard deviation from a mean value of the normalized MRS spectrum data associated with a single type of tissue at the particular nuclear magnetic resonance frequency. 
     
     
         34 . The machine-readable media of  claim 33 , wherein each of the plurality of weighting constants at a particular nuclear magnetic resonance frequency represents an average standard deviation taken from an average of raw MRS spectrum data associated with a plurality of types of tissue at the particular nuclear magnetic resonance frequency. 
     
     
         35 . The MRS preprocessing system of  claim 23 , further comprising a normalization module and a renormalization module, wherein the normalization module normalizes the raw MRS spectrum data to generate normalized MRS spectrum data, wherein the variance-weighting module scales the normalized MRS spectrum data, and wherein the renormalization module renormalizes the weighted MRS spectrum data to generate a preprocessed MRS spectrum data. 
     
     
         36 . The machine-readable media of  claim 28 , further comprising normalization and renormalization steps, wherein the normalizing step comprises normalizing the raw MRS spectrum data to generate normalized MRS spectrum data, wherein the scaling step comprises scaling the normalized MRS spectrum data, and wherein renormalizing step comprises renormalizing the weighted MRS spectrum data to generate a preprocessed MRS spectrum data. 
     
     
         37 .- 50 . (canceled) 
     
     
         51 . A method for detecting a tissue abnormality comprising:
 providing to at least one computing device a preprocessed MRS spectrum data of a tissue sample comprising a summary of signals produced by the tissue sample at a range of nuclear magnetic resonance frequencies, wherein the preprocessed MRS spectrum data of the tissue sample further comprises a raw MRS spectrum data subjected to a preprocessing method including scaling by variance-weighting;   providing to the at least one computing device a set of preprocessed comparison MRS spectrum data, comprising a plurality of preprocessed comparison MRS spectrum data, wherein each preprocessed comparison MRS spectrum data of the set comprises a summary of signals produced by an abnormal tissue sample having a known abnormality at a range of nuclear magnetic resonance frequencies, and wherein each preprocessed comparison MRS spectrum data of the set further comprises a raw comparison MRS spectrum data subjected to a preprocessing method including scaling by variance-weighting;   comparing, by the at least one computing device, the preprocessed MRS spectrum data of the tissue sample to the set of preprocessed comparison MRS spectrum data using a pattern recognition method; and,   identifying the tissue abnormality as the known abnormality of the preprocessed comparison MRS spectrum data that most closely matches the preprocessed MRS spectrum data of the tissue sample.   
     
     
         52 . (canceled) 
     
     
         53 . The method of  claim 51 , wherein the known abnormality comprises normal tissue, one or more malignant tumor tissues, one or more benign tumor tissues, or one or more non-cancerous tissue abnormalities. 
     
     
         54 . The method of  claim 51 , wherein the a raw MRS spectrum data and the raw comparison MRS spectrum data are subjected to a preprocessing method further comprising normalization and renormalization.