Machine learning spectral ffr-ct

Abstract

One embodiment of the present invention includes a computer-implemented method that includes receiving spectral computed tomography (CT) volumetric image data. The spectral CT volumetric image data include data for at least two different energies and/or energy ranges. The spectral CT volumetric image data is processed with a machine learning engine configured to map spectrally enhanced features extracted from the spectral CT volumetric image data onto fractional flow reserve (FFR) values to determine a FFR value. The FFR value is then visually presented.

What is claimed is: 
     
         1 . A computer-implemented method, comprising:
 receiving spectral computed tomography (CT) volumetric image data, wherein the spectral CT volumetric image data include data for at least two different energies and/or energy ranges;   processing the spectral CT volumetric image data with a machine learning engine configured to map spectrally enhanced features extracted from the spectral CT volumetric image data onto fractional flow reserve (FFR) values to determine a FFR value; and   visually presenting the FFR value.   
     
     
         2 . The method according to  claim 1 , further comprising extracting a spectral morphological feature by detecting, segmenting and classifying using the spectral volumetric image data. 
     
     
         3 . The method according to  claim 1 , wherein the spectrally enhanced physiological feature includes a quantity related to coronary blood flow. 
     
     
         4 . The method according to  claim 3 , wherein the spectrally enhanced physiological feature includes one or more of a spectral myocardium deficit or a spectral collateral flow. 
     
     
         5 . The method according to  claim 1 , further comprising extracting a feature from the spectral CT volumetric image data, wherein the extracted feature includes at least one of a spectrally enhanced anatomical feature, a spectrally enhanced plaque morphological and functional feature, and a spectrally enhanced physiological feature. 
     
     
         6 . The method according to  claim 5 , wherein the spectrally enhanced anatomical feature includes a geometrical and topological feature. 
     
     
         7 . The method according to  claim 6 , further comprising enhancing an anatomical feature using different spectral images to determine a presence of different anatomical tissue of a coronary tree anatomy. 
     
     
         8 . The method according to  claim 6 , further comprising enhancing an anatomical feature using different mono-energetic images to find a boundary between different anatomical tissue of a coronary tree anatomy. 
     
     
         9 . The method according to  claim 6 , further comprising enhancing an anatomical feature using spectrally enabled regularization. 
     
     
         10 . The method according to  claim 1 , wherein the machine learning engine estimates the FFR value at a predetermined location of a coronary tree by applying a function on a feature describing the predetermined location. 
     
     
         11 . The method according to  claim 10 , wherein the function describes a statistical relationship between the feature and the FFR value.