System and method for statistical mapping between genetic information and facial image data

Abstract

A method and system for statistical mapping between genetic information and facial image data including collecting a multiplicity of sets of genetic information and matching facial image data representing a multiplicity of individuals, representing the genetic information of each of the multiplicity of individuals as a first multidimensional representation, representing the facial image data of each of the multiplicity of individuals as a second multidimensional representation; and inferring correlative, non-causal, statistical relationships between the first multidimensional representations and the second multidimensional representations. A system and method for estimating the likelihood of donor-recipient transplant compatibility using facial images of potential donors, the method including inferring correlative, non-causal, statistical relationships, indicative of transplant compatibility, between multidimensional representations of facial image data of potential donors and a multidimensional representation of information relating to a potential recipient.

1 - 70 . (canceled) 
     
     
         71 . A system for statistical mapping between genetic information and facial image data, the system comprising:
 a processor; and   a data storage device operative to store a multiplicity of sets of genetic information and matching facial image data representing a multiplicity of individuals, said genetic information of each of said multiplicity of individuals being stored as a first multidimensional representation and said facial image data of each of said multiplicity of individuals being stored as a second multidimensional representation,   said processor comprising a non-causal statistical correlation engine operative to generate inferred models based on correlative, non-causal, statistical relationships between said first multidimensional representations and said second multidimensional representations; and   said processor being operative to store said inferred models in said data storage device.   
     
     
         72 . A system for mapping between genetic information and facial image data according to  claim 71  and wherein said correlation engine is operative to exclude causal statistical relationships. 
     
     
         73 . A system for mapping between genetic information and facial image data according to  claim 71  and wherein said processor is operative to generate said first representation using at least one of the following methods:
 representing nucleotide sequences of length n as vectors of length 4n wherein each element depicts the number of occurrences of each nucleotide at a specific sequence location in said genetic information; 
 representing STR data of n loci as vectors of length 2n wherein each element depicts either the lower or the higher repetition number at one locus in said genetic information; 
 representing allele information of n loci, each having a multiplicity of γ i , i=1 . . . n alleles as vectors of length 
 
       
         
           
             
               
                 ∑ 
                 
                   i 
                   = 
                   1 
                 
                 n 
               
                
               
                   
               
                
               
                 γ 
                 i 
               
             
           
         
       
       wherein each vector element denotes the number of times each allele is present in said genetic information; and
 representing said genetic information implicitly by providing a similarity function between said genetic information arising from every two individuals. 
 
     
     
         74 . A system for mapping between genetic information and facial image data according to  claim 71  and wherein said processor is operative to generate said second representation using at least one of the following algorithms:
 detection of a face region within said facial image data by using a learned template; and 
 alignment of said face region by the detection of a set of fiducial points. 
 
     
     
         75 . A system for generating a facial image based on genetic information, the system comprising:
 a processor; and   a data storage device operative to store said genetic information;   said processor being operative to:
 provide results of statistical mapping between said genetic information and facial image data including face shape and facial texture information produced by inferring correlative, non-causal, statistical relationships between first multidimensional representations of said genetic information and second multidimensional representations of said facial image data for a multiplicity of persons; 
 apply said results of said statistical mapping to genetic information obtained from an individual not necessarily being an individual whose genetic information is used for inferring said statistical relationships, thereby to obtain facial image data including at least face shape information for said individual; and 
 employ said facial image data including face shape and facial texture information to generate said facial image of said individual. 
   
     
     
         76 . A system for generating a facial image based on genetic information according to  claim 75  and wherein said facial image data of each of said multiplicity of persons also includes facial texture information. 
     
     
         77 . A system for generating a facial image based on genetic information according to  claim 75  and wherein:
 said processor is operative to generate said second multidimensional representation using at least one of the following algorithms:
 detection of a face region within said facial image data by using a learned template; and 
 alignment of said face region by the detection of a set of fiducial points, and 
 
 said processor is operative to generate said second multidimensional representation of said facial image data of each of said multiplicity of persons and a multidimensional representation of facial image data of said individual includes at least one of the following representations:
 a shape vector comprising a multiplicity of elements each indicating a location of a feature point or a contour point in said facial image data; and 
 a texture vector comprising a multiplicity of elements each indicating an image measurement at one image location of said facial image data. 
 
 
     
     
         78 . A system for generating a facial image according to  claim 77  and wherein said processor is operative to generate said statistical mapping using at least one of the following algorithms:
 regularized linear regression of said shape vector; and 
 regularized linear regression of said texture vector. 
 
     
     
         79 . A system for generating a facial image according to  claim 77  and wherein said processor is operative to apply the results of said mapping using at least one of the following algorithms:
 use of linear regression models for recovering said shape vector of said individual; and 
 use of linear regression models for recovering said texture vector of said individual. 
 
     
     
         80 . A system for generating a facial image according to  claim 77  and wherein said processor is operative to employ thin plate spline warping of said texture vector of said individual with in accordance with said locations of said shape vector of said individual. 
     
     
         81 . A system for generating a facial image according to  claim 75  and wherein said processor is operative to generate said statistical mapping using at least one of the following algorithms:
 regularized linear regression of said face shape information; and 
 regularized linear regression of said facial texture information. 
 
     
     
         82 . A system for generating a facial image according to  claim 75  and wherein said processor is operative to apply the results of said mapping using at least one of the following algorithms:
 use of linear regression models for at least one of recovering, generating and predicting said face shape information included in said facial image data of said individual; and 
 use of linear regression models for recovering said facial texture information included in said facial image data of said individual. 
 
     
     
         83 . A system for generating a facial image according to  claim 75  and wherein said processor is operative to employ thin plate spline warping of said facial texture information of said individual in accordance with said face shape information of said individual. 
     
     
         84 . A system for selecting a facial image corresponding to an individual from facial images of a plurality of persons based on genetic information of said individual, the system comprising:
 a processor; and   a data storage device operative to store said facial images of a plurality of persons;   said processor being operative to:
 provide results of statistical mapping between genetic information and facial image data produced by inferring non-causal (correlative) statistical relationships between first multidimensional representations of said genetic information and second multidimensional representations of said facial image data for a multiplicity of individuals, not necessarily including said individual and said plurality of persons; and 
 apply said results of said statistical mapping to genetic information obtained from said individual and to said facial images of said plurality of persons, thereby to select said facial image of said individual from among said facial images of said plurality of persons. 
   
     
     
         85 . A system for selecting a facial image according to  claim 84  and wherein said facial image data of each of said plurality of persons is represented as a multidimensional representation. 
     
     
         86 . A system for selecting a facial image according to  claim 85  and wherein:
 said processor is operative to generate said multidimensional representation of said facial image data of each of said plurality of persons using at least one of the following algorithms:
 detection of a face region within said facial image data by using a learned template; and 
 alignment of said face region by the detection of a set of fiducial points, and 
 
 said multidimensional representation of said facial image data of each of said plurality of persons includes at least one of the following representations:
 a vector containing a multiplicity of histograms of image gradients at a multiplicity of image locations; and 
 a vector containing a multiplicity of histograms of local-binary-patterns obtained at a multiplicity of image locations. 
 
 
     
     
         87 . A system for selecting a facial image according to  claim 84  and wherein said processor is operative to infer statistical relationships using at least one of the following algorithms:
 Canonical Correlation Analysis; 
 Kernel Canonical Correlation Analysis; 
 Nearest Neighbor Transfer; 
 Generalized Nearest Neighbor Transfer; 
 Inverse Nearest Neighbor Transfer; 
 Inverse Generalized Nearest Neighbor Transfer; 
 identification based on the structure of recovered phylogenetic-trees; 
 statistical classifier learning of multidimensional representations each containing information from said first multidimensional representation and said second multidimensional 
 Common Discriminant Feature Extraction; 
 Maximal Margin Robot; and 
 Preference Margin Optimization. 
 
     
     
         88 . A system for selecting a facial image according to  claim 84  and wherein said processor is operative to select said facial image of said individual from among said facial images of said plurality of persons using at least one of the following methods:
 transformation of said facial image data and said genetic information using transformations obtained by employing Canonical Correlation Analysis; 
 transformation of said facial image data and said genetic information using transformations obtained by employing Kernel Canonical Correlation Analysis; 
 selection based on minimal D L     1    distance; 
 selection based on minimal D CoL     1    distance; 
 selection based on minimal D L     2    distance; 
 selection based on minimal D CoL     2    distance; 
 selection based on minimal D C  distance; 
 selection based on minimal D CC  distance; 
 selection based on the Nearest Neighbor Transfer decision rule; 
 selection based on the Generalized Nearest Neighbor Transfer decision rule; 
 selection based on the Inverse Nearest Neighbor Transfer decision rule; 
 selection based on the Inverse Generalized Nearest Neighbor Transfer decision rule; 
 selection based on examining said genetic information obtained from said individual and said facial images of said plurality of persons within the context of phylogenetic trees constructed from said first multidimensional representations of said genetic information and said second multidimensional representations of said facial image data for said multiplicity of individuals; 
 selection based on applying classification rules to multidimensional representations each containing information from said first multidimensional representation and said second multidimensional representation; 
 transformation of said facial image data and said genetic information using transformations obtained by employing Common Discriminant Feature Extraction; 
 selection based on the Maximal Margin Robot matching score; and 
 selection based on the Preference Margin Optimization matching score.