Breast imaging reporting and data system (bi-rads) tissue composition

Abstract

Breast density is a significant breast cancer risk factor measured from mammograms. Disclosed is a methodology for converting continuous measurements of breast density and calibrated mammograms into a four-state ordinal variable approximating the BI-RADS ratings. In particular, the present disclosure is directed to a calibration system for a specific full field digital mammography (FFDM) technology. The calibration adjusts for the x-ray acquisition technique differences across mammograms resulting in standardized images. The approach produced various calibrated and validated measures of breast density, one of which assesses variation in the mammogram referred to as Vc (i.e. variation measured from calibrated mammograms). The variation in raw mammograms [i.e. Vr] is a valid breast density risk factor in both FFDM in digitized film mammograms.

1 . A method of assessing breast density for breast cancer risk applications, the method comprising:
 receiving raw digital image data including a plurality of pixels;   performing a statistical analysis with optimization on the raw digital image data by determining a measured variation of the raw digital image data based on a deviation of the pixel values within a breast region;   determining four-state ordinal variables (BR pg ) from the measured variation; and   associating the four-state ordinal variables (BR pg ) with a measure of risk for breast cancer.   
     
     
         2 . The method of  claim 1 , further comprising approximating a portion of a breast represented within the raw digital image data that was in contact with a compression paddle during image acquisition. 
     
     
         3 . The method of  claim 1 , wherein determining the four-state ordinal variables comprises applying differential evolution optimization. 
     
     
         4 . The method of  claim 3 , wherein applying the differential evolution optimization comprises maximizing or minimizing a fitness function by a repeated processing of image case-control datasets with plural random vectors for a given breast density measurement determination for a number of generations, wherein the number of generations is less than predetermined value when a preset convergence condition is met. 
     
     
         5 . The method of  claim 2 , further comprising:
 generating an integrated histogram from the raw digital image data;   performing the statistical analysis with optimization on the generated integrated histogram; and   determining four-state ordinal variables based on the analyzed integrated histogram data with a measure of risk for breast cancer.   
     
     
         6 . The method of  claim 2 , further comprising applying differential evolution (DE) optimization to determine the four-state ordinal variables (BR pg ). 
     
     
         7 . The method of  claim 6 , applying the differential evolution optimization further comprising maximizing or minimizing a fitness function by a repeated processing of image case-control datasets with plural random vectors for a given breast density measurement determination for a number of generations, wherein the number of generations is less than a predetermined value when a preset convergence condition is met. 
     
     
         8 . The method of  claim 1 , further comprising:
 calibrating the raw digital image data to generate calibrated digital image data, wherein each pixel is mapped into a normalized percent glandular representation; and   assessing the measured variation from the calibrated digital image data to further determine the four-state ordinal variables.   
     
     
         9 . A method of assessing breast density using full field digital mammography (FFDM), comprising:
 receiving mammograms as raw digital image data;   converting the mammograms into a four-state ordinal variable as an approximation for the BI-RADS measurements using the histograms for each image;   assessing a variation in the mammograms; and   determining a breast density risk factor from the variation.   
     
     
         10 . The method of  claim 9 , wherein determining the four-state ordinal variables comprises applying differential evolution optimization. 
     
     
         11 . The method of  claim 10 , wherein applying the differential evolution optimization comprises maximizing or minimizing a fitness function by a repeated processing of image case-control datasets with plural random vectors for a given breast density measurement determination for a number of generations, wherein the number of generations is less than a predetermined value when a preset convergence condition is met. 
     
     
         12 . The method of  claim 11 , wherein the present convergence condition is met when |AZ Maximum −AZ Minimum ≤0.01 in a given generation. 
     
     
         13 . The method of  claim 12 , wherein the present convergence condition is met when |Δ Maximum −Δ Minimum ≤0.001 in a given generation. 
     
     
         14 . The method of  claim 13 , further comprising:
 generating an integrated histogram from the raw data digital image data;   performing a statistical analysis with optimization on the generated integrated histogram; and   determining four-state ordinal variables based on the analyzed integrated histogram data with a measure of risk for breast cancer.   
     
     
         15 . The method of  claim 14 , further comprising comparing the four state breast density variables based on the integrated histogram with the four state breast density variables based on the determined measured variation. 
     
     
         16 . The method of  claim 9 , further comprising:
 calibrating the raw digital image data to generate calibrated digital image data, wherein each pixel is mapped into a normalized percent glandular representation; and   assessing the measured variation from the calibrated digital image data to further determine the four-state ordinal variables.   
     
     
         17 . A method of assessing breast density for breast cancer risk applications, the method comprising:
 receiving raw digital image data including a plurality of pixels;   calibrating the digital image data, wherein each pixel is mapped into a normalized percent glandular representation;   performing a statistical analysis with optimization on the raw digital image data by determining a first measured variation of the raw digital image data based on a deviation of the pixel values within a breast region;   performing a statistical analysis with optimization on the calibrated digital image data by determining a second measured variation of the calibrated digital image data based on a deviation of the pixel values within a breast region;   determining four-state ordinal variables (BR pg ) from the first measured variation and the second measured variation; and   associating the four-state ordinal variables (BR pg ) with a measure of risk for breast cancer.   
     
     
         18 . The method of  claim 17 , wherein determining the four-state ordinal variables comprises applying differential evolution optimization. 
     
     
         19 . The method of  claim 18 , wherein applying the differential evolution optimization comprises maximizing or minimizing a fitness function by a repeated processing of image case-control datasets with plural random vectors for a given breast density measurement determination for a number of generations, wherein the number of generations is less than predetermined value when a preset convergence condition is met. 
     
     
         20 . The method of  claim 17 , further comprising:
 generating an integrated histogram from the raw digital image data and the calibrated digital image data;   performing the statistical analysis with optimization on the generated integrated histogram; and   determining four-state breast ordinal variables based on the analyzed integrated histogram data with a measure of risk for breast cancer.