Identification of regions of interest and extraction of time value curves in imaging procedures

• US 8,699,770 B2
• Filed: 10/19/2012
• Issued: 04/15/2014
• Est. Priority Date: 06/24/2008
• Status: Active Grant

• Current Assignee: Bayer Healthcare LLC (Bayer AG)
• Original Assignee: Bayer Medical Care Inc. (Bayer AG)
• Inventors: Kemper, Corey A., Kalafut, John F.
• Primary Examiner(s): LaBalle, Clayton E
• Assistant Examiner(s): Rhodes, Jr., Leon W
• Application Number:
  US13/655,525
• Publication Number:
  US 20130044926A1
• Time in Patent Office: 543 Days
• Field of Search: 382/128, 382/130
• US Class Current: 382/128
• CPC Class Codes: G06F 18/23213 with fixed number of clusters, e.g. K-means clustering
• CPC Class Codes: G06T 2207/10081 Computed x-ray tomography [CT]
• CPC Class Codes: G06T 2207/30061 Lung
• CPC Class Codes: G06T 7/0012 Biomedical image inspection
• CPC Class Codes: G06T 7/0016 involving temporal comparison
• CPC Class Codes: G06T 7/11 Region-based segmentation
• CPC Class Codes: G06T 7/143 involving probabilistic approaches, e.g. Markov random field [MRF] modelling
• CPC Class Codes: G06V 10/763 Non-hierarchical techniques, e.g. based on statistics of modelling distributions
• CPC Class Codes: G06V 2201/031 of internal organs

Abstract

A method of extracting at least one time-value curve to determine a protocol in an imaging procedure using an imaging system, includes: determining a first N-dimensional data set of pixel values of a portion of a body of the patient at a first time using the imaging system, wherein N is an integer; determining at least a second N-dimensional data set of pixel values of the portion at a second time using the imaging system; computing a predetermined number of correlated segments of the imaged portion corresponding to a predetermined number of regions of interest of the patient by computing a similarity metric of a time series of pixel values; computing the at least one time-value curve for at least one of the regions of interest; and determining a protocol for a diagnostic scan using the image system based at least in part upon data from the time value curve.

Claims

1. A method of extracting time-value curves for use in determining a protocol for a diagnostic imaging procedure to be performed on a patient, comprising:

(a) determining a series of 0 through T M-dimensional data sets of pixel values from a series of images of a cardiovascular system of the patient acquired using an imaging system, wherein M is an integer, T is an integer, the 0 data set corresponds to a data set at a time 1=0, the T data set corresponds to a data set at a time t=T, and wherein a pixel at a location represented by a variable x has an enhancement profile vector defined as;

y_x=[y_x(0)y_x(1) . . . y_x(T−

1)y_x(T)]

and an enhancement level y of the pixel at the location x is defined over time as;

y_x(t)=y_x(0)+s_x(t)+η

_x(t)

wherein y_x(0) is a baseline enhancement level for the pixel, s is a change in signal due to flow of a fluid in the patient and η

is a noise term;

(b) computing a predetermined number of correlated segments of the cardiovascular system corresponding to two regions of interest thereof by computing a similarity metric of a time series of pixel values in the images of the series;

(c) computing one of the time-value curves for each of the two regions of interest; and

(d) determining the protocol for the diagnostic imaging procedure based at least in part upon data extracted from the time value curves computed for the two regions of interest of the cardiovascular system of the patient.

2. The method of claim 1 wherein M=2 thereby rendering the M-dimensional data sets as two-dimensional data sets.

3. The method of claim 1 wherein M=3 thereby rendering the M-dimensional data sets as three-dimensional data sets.

4. The method of claim 1 wherein the predetermined number of correlated segments are computed using K-means clustering.

5. The method of claim 4 wherein K clusters are computed by minimizing a sum of squares of distances between data points and a K cluster centroid corresponding thereto, wherein K is an integer corresponding to the predetermined number of correlated segments.

6. The method of claim 5 wherein the pixel values are randomly assigned to the K clusters and each of the K cluster centroids is a component-wise mean of the pixel values in the K cluster corresponding thereto after centering and normalizing the pixel values to a zero mean and unit standard deviation as follows:

7. The method of claim 6 wherein each of the K cluster centroids is determined using the following formula:

8. The method of claim 7 wherein the distance from the centered and normalized pixel values to each of the K cluster centroids y_c is determined using the formula:

9. The method of claim 8 wherein each of the centered and normalized pixel values is assigned to the K cluster to which it exhibits the minimum computed distance to the K cluster centroid thereof.

10. The method of claim 9 wherein the actions of claims 7 through 9 are repeated until convergence, resulting in a segmented image having K segments corresponding to the determined K clusters.

11. The method of claim 10 further comprising filtering the segmented image to eliminate pixels that are not well correlated with neighboring pixels.

12. The method of claim 11 wherein each pixel of the segmented image is compared with all eight of its neighboring pixels in filtering.

13. The method of claim 11 further comprising morphologically opening the segmented image.

14. The method of claim 13 further comprising overlaying the segmented image upon a contrast enhanced bolus data set and computing enhancement profiles for each of the K segments.

15. The method of claim 14 further comprising semantically labeling at least one of the enhancement profiles with a semantic label corresponding to an appropriate one of the two regions of interest of the patient based upon at least one characteristic of the labeled enhancement profile as compared to at least one other computed enhancement profile.

16. The method of claim 10 further comprising overlaying the segmented image upon a contrast enhanced bolus data set and computing enhancement profiles for each of the K segments.

17. The method of claim 16 further comprising semantically labeling at least one of the enhancement profiles with a semantic label corresponding to an appropriate one of the two regions of interest of the patient based upon at least one characteristic of the labeled enhancement profile as compared to at least one other computed enhancement profile.

18. The method of claim 1 wherein a first of two regions of interest is a portion of a pulmonary artery of the patient and a second of the two regions of interest is a portion of an aorta of the patient.

19. The method of claim 1 wherein a first of two regions of interest is at least one of vessels and structure of a right side of a heart of the patient and a second of the two regions of interest is at least one of vessels and structure of a left side of the heart of the patient.

20. The method of claim 1 wherein the fluid includes a contrast medium.

21. The method of claim 1 wherein the method is embodied in software embodied in at least one of (i) the imaging system, (ii) an injector system and (iii) a system separate and standalone from the imaging system and the injector system.

22. A system for extracting time-value curves for use in determining a protocol for a diagnostic imaging procedure to be performed on a patient, the system comprising:

(a) an input system for input of data output from at least one imaging system, the data comprising a series of 0 through T M-dimensional data sets of pixel values from a series of images of a cardiovascular system of the patient acquired using the at least one imaging system, wherein M is an integer, T is an integer, the 0 data set corresponds to a data set at a time t=0, the T data set corresponds to a data set at a time t=T, and wherein a pixel at a location represented by a variable x has an enhancement profile vector defined as;

y_x=[y_x(0)y_x(1) . . . y_x(T−

1)y_x(T)]

and an enhancement level y of the pixel at the location x is defined over time as;

y_x(t)=y_x(0)+s_x(t)+η

_x(0

wherein y_x(0) is a baseline enhancement level for the pixel, s is a change in signal due to flow of a fluid in the patient and η

is a noise term;

(b) at least one processor in communicative connection with the input system and adapted to compute (i) a predetermined number of correlated segments of the cardiovascular system corresponding to two regions of interest thereof by computing a similarity metric of a time series of pixel values in the images of the series and (ii) one of the time value curves for each of the two regions of interest; and

(c) at least one parameter generator system to determine the protocol for the diagnostic imaging procedure based at least in part upon data extracted from the time value curves computed for the two regions of interest of the cardiovascular system of the patient.

23. The system of claim 22 further comprising a memory system in communicative connection with the at least one processor, the memory system having stored therein an algorithm executable by the at least one processor to compute the predetermined number of correlated segments of the cardiovascular system corresponding to the two regions of interest of the patient.

24. The system of claim 22 wherein the system is in communicative connection with at least one of (i) the at least one imaging system (ii) at least one injector system.

25. The system of claim 22 wherein a first of two regions of interest is a portion of a pulmonary artery of the patient and a second of the two regions of interest is a portion of an aorta of the patient.

26. The system of claim 22 wherein a first of two regions of interest is at least one of vessels and structure of a right side of a heart of the patient and a second of the two regions of interest is at least one of vessels and structure of a left side of the heart of the patient.

27. The system of claim 22 wherein the fluid includes a contrast medium.