Compensation of computed tomography data for detector afterglow

Claim 2

A medical imaging system comprising: a source of radiation; a detector for sensing radiation from said source and producing an output signal representing the sensed radiation, said detector having an exponential impulse response that is characterized by a plurality of N components having different time constants; means for periodically sampling the output signal from said detector to acquire a set of radiation attenuation values; means for recursive filtering the set of attenuation values to remove effects due to the exponential impulse response to produce a set of filtered values, said filtering means employs a filter function having terms for the plurality of impulse response components; and means for reconstructing an image from the set of filtered values.

Claim 3

The medical imaging system as recited in claim 1 wherein said means for recursive filtering produces a given filtered value x k by applying the following expression to each attenuation value y(kΔt): ##EQU8## where k designates the value in the sets, Δt is the periodic sampling interval, β n =α n (1-e - Δt/τ.sbsp.n) in which α n represents the strength of an impulse response component having time constant τ n , and S nk =x k-1 +e - Δt/τ.sbsp.n S n (k-1).

Claim 6

The method for producing an image as recited in claim 3 which further comprises storing the attenuation values in a memory; and said recursive filtering step reads the attenuation values from the memory.

Claim 9

The method as recited in claim 5 further comprising: characterizing the response of the detector to derive values of α n and τ n for each component used in producing the second set of data values.

Claim 10

The method as recited in claim 6 in which the step of characterizing the radiation response of the radiation detector comprises: generating a beam of radiation which strikes the radiation detection from time t 0 to time t 1 ; periodically sampling a signal produced by the radiation detector, such sampling occurs while the beam of radiation is being generated and continues until time t 2 which is later than time t 1 , to produce a plurality of signal samples; determining a time constant τ n for each component of the impulse response; calculating an X-ray flux intensity value Ψ from a sample acquired between time t 0 and t 1 ; and deriving a value for α n for each component of the impulse response from signal samples acquired between times t 1 and t 2 according to the expression: ##EQU10##

Claim 11

The method as recited n claim 6 in which the step of characterizing the response of the detector comprises: generating a beam of radiation from time t 0 to time t 1 ; acquiring a first set of test radiation attenuation values by periodically sampling the signal produced by the radiation detector, such sampling occurs while the beam of radiation is being generated and continues until time t 2 which is later than time t 1 ; determining a time constant τ n for each component of the impulse response; calculating an X-ray flux intensity value from a sample taken between times t 0 and t 1 ; using the calculated X-ray flux intensity value to normalize signal samples acquired between times t 1 and t 2 and then taking the logarithm of the normalized signal samples; deriving α n for each component of the impulse response from said normalized signal samples, log[y(T)], according to the expression log[y(T)]=log α n -(T/τ n ).

Claim 12

The method as recited in claim 8 in which the step of deriving α n comprises: (a) selecting a normalized signal sample that was acquired at a time T at which all the components of the impulse response can be neglected except the component which has the longest time constant τ n ; (b) solving the expression log[y(T)]=log α n -(T/τ n ) for α n ; (c) removing the effects of the component of the impulse response which has the longest time constant τ n from the normalized signal samples; (d) successively repeating steps (a) through (c) for each remaining components of the impulse response in descending order of time constant length, each time step (c) is executed the effects of the remaining component with the longest time constant τ n are removed from the normalized signal samples.