Automated methods for pre-selection of voxels and implementation of pharmacokinetic and parametric analysis for dynamic contrast enhanced MRI and CT

A method of filtering out non-conforming pixel data of an image obtained from enhancing tissue, said method comprising:

• identifying a number N of time points at which signal intensity data are obtained for the image;
• setting t1 as a time point at which a peak signal intensity is expected for tumorous tissue;
• calculating, based on N and t1, a madness value for each pixel using each time point at which signal intensity is obtained after a pre-selected time point t1;
• and
• filtering out pixel data that do not satisfy a pre-determined madness threshold.

A method of filtering data of an image obtained from enhancing tissue, the method comprising:

• identifying one or more qualifying pixels that satisfy an enhancement threshold and a madness threshold by at least:
• determining an enhancement value for each pixel of the image according to the formula Et=(It−I0)/I0, where I0 is a pre-contrast signal intensity value of the pixel and It is a signal intensity value of the pixel at time t post-contrast;
• comparing Et for each pixel to a pre-specified enhancement threshold;
• and
• classifying the pixel as a non-qualifying data if the pixel does not satisfy the enhancement threshold;
• determining a number N of time points at which signal intensity data are obtained for the image;
• setting t1 as a time point at which a peak signal intensity value is expected for tumorous tissue;
• calculating, based on N and t1, a madness value for each pixel using each time point at which signal intensity is obtained after a pre-selected time point t1;
• and
• classifying as non-qualifying data of the image one or more pixels below the enhancement threshold and data that do not satisfy the predetermined madness threshold;
• and
• determining a color density for each pixel by at least:
• for each pixel of the image, determining, as a color density ratio, a ratio of qualifying pixels to one or more pixels of the image for a neighborhood of pixels of a pre-specified size;
• and
• selecting, as high color density pixels, pixels with a color density ratio above a pre-determined percentile of color density ratios of the qualifying pixels.

A program storage device readable by machine, tangibly embodying a program of instructions executable by the machine to perform a method of filtering noise in image data, comprising:

• determining a signal intensity value, Ipeak, corresponding to a pixel count of a greatest number of pixels, Npeak, in image data;
• setting a noise threshold at a signal intensity value, Inoise, corresponding to a pixel count, NI, such that NI is determined as a function of Npeak;
• and
• filtering from the image data one or more pixels with signal intensity values below the noise threshold.

A program storage device readable by machine, tangibly embodying a program of instructions executable by the machine to perform a method of filtering noise in image data, comprising:

• identifying a number N of time points at which signal intensity data are obtained for image data;
• setting t1 as a time point at which a peak signal intensity is expected for tumorous tissue;
• calculating, based on N and t1, a madness value for each pixel using each time point at which signal intensity is obtained after a pre-selected time point t1;
• and
• filtering out pixel data that do not satisfy a pre-determined madness threshold, wherein the madness value is calculated based on the formula:
• Mad
• j
• ⁢
• =
• min
• ∈
• j
• ⁢
• 1
• -
• 
• I
• j
• -
• 1
• +
• I
• j
• +
• 1
• -
• 2
• ⁢
• I
• j
• 
• I
• j
• -
• 1
• +
• I
• j
• +
• 1
• ;
• ⁢
• j
• =
• t
• 1
• +
• 1
• ÷
• N
• -
• 1

A program storage device readable by machine, tangibly embodying a program of instructions executable by the machine to perform a method of filtering noise in image data, comprising:

• identifying qualifying pixels that satisfy an enhancement threshold, a madness, and a noise threshold by:
• determining an enhancement value for each pixel of image data according to the formula Et=(It−I0)/I0, where I0 is a pre-contrast signal intensity value of the pixel and It is a signal intensity value of the pixel at time t post-contrast;
• comparing Et for each pixel to a pre-specified enhancement threshold;
• classifying the pixel as a non-qualifying data when the pixel does not satisfy
• the enhancement threshold;
• determining a number N of time points at which signal intensity data are obtained for the image data;
• setting t1 as a time point at which a peak signal intensity value is expected for tumorous tissue;
• calculating, based on N and t1, a madness value for each pixel using each time point at which signal intensity is obtained after a pre-selected time point t1;
• classifying as non-qualifying data of the image data, one or more pixels below the enhancement threshold and data that do not satisfy the pre-determined madness threshold;
• receiving pixel count for each signal intensity value of the image data;
• determining a signal intensity value, Ipeak, corresponding to a pixel count of a greatest number of pixels, Npeak;
• setting the noise threshold at a signal intensity value, Inoise, with a pixel count NI, such that NI is determined based on Npeak;
• classifying as non-qualifying data all pixels below the noise threshold;
• and determining a color density for each pixel by:
• for each pixel of the image data, determining, as a color density ratio, a ratio of qualifying pixels to all pixels of the image data for a neighborhood of pixels of a pre-specified size;
• and
• selecting, as high color density pixels, pixels with a color density ratio above a pre-determined percentile of color density ratios of the qualifying pixels.

A method of filtering out pixels corresponding to tissue of an area of non-interest in image data, said method comprising:

• identifying as cluster centers one or more pixels according to a color density determined for qualifying pixels of image data, the color density determined based on signal intensity time curves of the pixels of the image data;
• growing a cluster unit for the cluster center, by adding to the cluster unit one or more adjacent pixels that are determined to have signal intensity time curves similar to the signal intensity time curve of the cluster center;
• merging the cluster unit with an adjacent cluster unit, when a correlation of the signal intensity time curves of the cluster centers satisfies a predetermined correlation threshold;
• and
• filtering out merged adjacent cluster units, when the merged cluster units exceeds a pre-determined dimension.