Data reduction in spectral CT

A system for data reduction of multi-energy computed tomography (CT) projection data, comprising:

• a memory that stores a plurality of instructions;
• and
• a processor assembly coupled to the memory and configured to execute the plurality of instructions to:
• obtain first CT projection data comprising data for each of a set of three or more measured x-ray energy channels;
• transform the first projection data into second projection data, wherein the second projection data comprises data for a set of two or more synthetic energy channels, each formed as a weighted combination of the measured energy channels, wherein the number of synthetic energy channels is less than the number of measured energy channels;
• wherein weights of said weighted combination forming each synthetic energy channel are determined in an optimization procedure which comprises:
• applying an error estimator model configured to receive as input a candidate set of synthetic energy channels formed according to a candidate set of weights for the weighted combination, and to generate as output a predicted covariance matrix of outputs of a pre-defined material decomposition algorithm;
• applying a pre-defined cost function to the covariance matrix, wherein a cost output from the cost function is correlated with a predicted measure of error for the outputs of the material decomposition algorithm;
• and
• adjusting the candidate weights iteratively to minimize the cost output to thereby arrive at an optimized set of weights.

A non-transitory computer-readable medium for storing executable instructions, which cause a method for data reduction of multi-energy computed tomography (CT) projection data to be performed, the method comprising:

• obtaining first CT projection data comprising data for each of a set of three or more measured x-ray energy channels;
• transforming the first projection data into second projection data, wherein the second projection data comprises data for a set of two or more synthetic energy channels, each formed as a weighted combination of the measured energy channels, wherein the number of synthetic energy channels is less than the number of measured energy channels;
• wherein weights of said weighted combination forming each synthetic energy channel are determined in an optimization procedure which comprises:
• applying an error estimator model configured to receive as input a candidate set of synthetic energy channels formed according to a candidate set of weights for the weighted combination, and to generate as output a predicted covariance matrix of outputs of a pre-defined material decomposition algorithm;
• applying a pre-defined cost function to the covariance matrix, wherein a cost output from the cost function is correlated with a predicted measure of error for the outputs of the material decomposition algorithm;
• and
• adjusting the candidate weights iteratively to minimize the cost output to thereby arrive at an optimized set of weights.