Intelligent Atlas for Automatic Image Analysis of Magnetic Resonance Imaging

Abstract

A non-invasive imaging system, including an imaging scanner suitable to generate an imaging signal from a tissue region of a subject under observation, the tissue region having at least one substructure; a signal processing system in communication with the imaging scanner to receive the imaging signal from the imaging scanner; and a data storage unit in communication with the signal processing system, wherein the data storage unit stores an anatomical atlas comprising data encoding spatial information of the at least one substructure in the tissue region, and a pathological atlas corresponding to an abnormality of the tissue region, wherein the signal processing system is adapted to automatically identify, using the imaging signal, the anatomical atlas, and the pathological atlas, a presence of the abnormality or a precursor abnormality in the subject under observation.

1 . A non-invasive imaging system, comprising:
 an imaging scanner suitable to generate an imaging signal from a tissue region of a subject under observation, the tissue region having at least one substructure;   a signal processing system in communication with said imaging scanner to receive the imaging signal from said imaging scanner; and   a data storage unit in communication with said signal processing system,   wherein said data storage unit is configured to store an anatomical atlas comprising spatial information of said at least one substructure in the tissue region, and a pathological atlas corresponding to an abnormality of said tissue region,   wherein said signal processing system is adapted to automatically identify, based on said imaging signal, said anatomical atlas, and said pathological atlas, a presence of said abnormality or a pre-cursor abnormality thereof in said tissue region of said subject under observation.   
     
     
         2 . The non-invasive imaging system according to  claim 1 , wherein said imaging scanner is a magnetic resonance imaging (MRI) scanner, a computed tomography (CT) imaging scanner, a positron emission tomography (PET) imaging scanner, a single positron emission computed tomography (SPECT) imaging scanner, or a combination thereof. 
     
     
         3 . The non-invasive imaging system according to  claim 1 , wherein said imaging signal is capable of providing at least one contrast mechanism that favorably delineates at least a portion of said at least one substructure of said tissue region. 
     
     
         4 . The imaging system according to  claim 3 , wherein said signal processing system is further adapted to automatically identify the presence of said abnormality by utilizing said at least one contrast mechanism. 
     
     
         5 . A workstation, comprising:
 a receiving engine adapted to receive   an image data representing a tissue region of a subject,   an anatomical atlas comprising spatial information of at least one anatomical substructure in the tissue region, and   a pathological atlas corresponding to an abnormality of said tissue region, wherein the pathological atlas comprises:   spatial information of a portion of said at least one anatomical substructure;   statistical quantities associated with said portion of the at least one anatomical substructure;   a normalizing engine constructed to provide a normalized image data by normalizing the image data, via a transformation, to the anatomical atlas;   a computing engine configured compute a statistical quantity from image voxels in the normalized image data corresponding to said portion of the at least one anatomical substructure of said tissue region; and   an analyzing engine configured to determine whether said abnormality or a pre-cursor abnormality thereof is present in said subject by analyzing a statistical relationship between the statistical quantity computed from said image data and the statistical quantities in the pathological atlas.   
     
     
         6 . The workstation according to  claim 5 , further comprising:
 a visualization engine adapted to superimpose said pathological atlas on the normalized image data.   
     
     
         7 . A method of generating a pathological atlas corresponding to an abnormality, comprising:
 receiving, from one of an imaging system, a workstation, or a first data storage device, a first image data representing a tissue region of at least one patient having said abnormality, wherein the first image data comprises a plurality of image voxels, wherein said tissue region has a plurality of anatomical substructures, and wherein the abnormality affects at least one of said plurality of anatomical substructures,   providing a normalized first image data by normalizing the first image data, via a transformation, to an anatomical atlas corresponding to said tissue region;   wherein the anatomical atlas comprises data encoding spatial information of the plurality of anatomical substructures, and   wherein the anatomical atlas is from one of the first data storage device, or a second data storage device;   creating the pathological atlas corresponding to said abnormality based on the normalized first image data;   wherein the pathological atlas comprises data encoding spatial information of the at least one of said plurality of anatomical substructures affected by said abnormality; and   storing the pathological atlas corresponding to said tissue abnormality on the data storage system.   
     
     
         8 . The method of  claim 7 , further comprising:
 computing a statistical quantity based on a portion of the plurality of image voxels corresponding to each of said plurality of anatomical substructures;   identifying a portion of said plurality of said anatomical structures wherein the statistical quantity computed is capable of providing a statistically significant differentiation between a group of patients having said abnormality and a normal control group of subjects; and   storing the statistical quantity computed in the pathological atlas.   
     
     
         9 . The method of  claim 8 , wherein said statistically significant differentiation is determined via a Linear Discriminatory Analysis (LDA) to substantially improve a classification power of using said statistic quantity. 
     
     
         10 . The method of  claim 8 , wherein the Linear Discriminatory Analysis (LDA) further comprises a plurality of permutation tests and a Receiver Operating Characteristic (ROC) analysis. 
     
     
         11 . The method of  claim 8 , further comprising:
 receiving a second image data representing the tissue region of a subject;   providing a normalized second image data by normalizing the second image data, via a transformation, to an anatomical atlas corresponding to said tissue region; and; and   applying the pathological atlas to automatically identify a presence of said abnormality or a pre-cursor abnormality thereof in the subject by:   computing said statistical quantity from a portion of image voxels in the normalized second image data corresponding to said portion of the plurality of anatomical substructures, and   determining whether said abnormality or said pre-cursor abnormality thereof is present by analyzing a statistic relationship between the statistical quantity computed from the normalized second image data and the statistical quantity stored in the pathological atlas.   
     
     
         12 . The method of  claim 11 ,
 wherein said second image data is capable of providing a plurality of contrast mechanisms for representing said tissue region, and   wherein said statistic relationship combines classification powers from each of the plurality of contrast mechanisms.   
     
     
         13 . The method of  claim 8 , wherein the statistical quantity is one of an average value, a median value, or equivalents thereof. 
     
     
         14 . The method of  claim 8 ,
 wherein said spatial information comprises location information, shape information, and transformation information, and   wherein the transformation information comprises atrophy information and expansion information.   
     
     
         15 . The method of  claim 7 , wherein the pathological atlas further comprises information capable of encoding at least one of fractional anisotropy, mean diffusivity, parallel diffusion, T 2  relaxation, or ′Yχ relaxation. 
     
     
         16 . The method of  claim 7 , wherein the first image data comprises diffusion tensor imaging (DTI) data. 
     
     
         17 . The method of  claim 7 , further comprising revising the atlas by
 receiving a third image data representing the tissue region,   providing a normalized third image data by normalizing the third image data, via a transformation, to an anatomical atlas corresponding to said tissue region; and   incorporating the normalized third image data to the pathological atlas, wherein the third image data is associated with a patient different from any of the at least one patient associated with the first image data.   
     
     
         18 . The method of  claim 7 , further comprising:
 superimposing the pathological atlas on the normalized first image data.   
     
     
         19 . The method of  claim 7 , wherein the transformation is a Large Deformation Deffeomorphic Metric Mapping. 
     
     
         20 . The method of  claim 7 , wherein said abnormality is one of a brain disease, a liver disease, a kidney disease, a muscle abnormality, or a joint abnormality. 
     
     
         21 . The method of  claim 20 , wherein said brain disease is a neurodegenerative disease.