Automated method for digital image quantitation

• US 5,797,396 A
• Filed: 12/19/1996
• Issued: 08/25/1998
• Est. Priority Date: 06/07/1995
• Status: Expired due to Fees

• Current Assignee: University of Florida Research Foundation, Incorporated (State University System of Florida)
• Original Assignee: University of Florida Research Foundation, Incorporated (State University System of Florida)
• Inventors: Wilson, David C., Geiser, Edward A.
• Primary Examiner(s): JAWORSKI, FRANCIS J
• Application Number:
  US08/770,912
• Time in Patent Office: 614 Days
• Field of Search: 382/132, 382/128, 382/199, 382/266, 128/653.1, 128/654, 128/660.01, 128/660.07, 128/662.02, 128/695, 128/696
• US Class Current: 600/407
• CPC Class Codes: G06T 2207/30048 Heart; Cardiac
• CPC Class Codes: G06T 7/62 of area, perimeter, diameter or volume
• CPC Class Codes: G06T 7/66 of image moments or centre of gravity

Abstract

A method for quantitatively analyzing digital images of approximately elliptical body organs, and in particular, echocardiographic images. In a first preferred embodiment, the invention automatically determines the center-point of an elliptical organ and a search region for border discrimination by finding the maximum value produced by a series of circular arc filters applied to the image data. In a second preferred embodiment, a pair of elliptical arc filters are used to make a first approximation of the positions of the right and left epicardium so that matched elliptical filters can be used to estimate the anterior epicardial position. The estimated position of the anterior epicardial, together with the posterior epicardial position, determines the center-point of the left ventricle. The positions of the organ borders are then approximated by generating amplitude distributions along each of a number of circular segments based on the center-point. The approximated borders are then compared on a best-fit basis to a set of paired elliptical arcs used to model the borders. The image may then be displayed with the modeled border superimposed on the image. A method of automated video densitometry is also disclosed whereby a sequence of end diastolic images are obtained before and after injection of a contrast agent.

Claims

1. A method for quantitatively analyzing a diagnostic image of a left ventricle including the steps of:

a. generating a diagnostic image frame in a digital format containing rows and columns of pixels of a left ventricle having a posterior epicardial border, and an anterior epicardium;

b. determining a first tentative center-point for the left ventricle and the position of the posterior epicardial border of the left ventricle depicted in the diagnostic image by frame repeatedly filtering the image frame with a set of circular arc filters until a maximum value for the set of filters is obtained;

c. determining a second tentative center-point for the left ventricle and the position of the anterior epicardium of the left ventricle depicted in the diagnostic image frame by repeatedly filtering the image frame with a set of matched elliptical arc filters until a maximum value for the set of matched filters is obtained;

d. determining a final center-point for the left ventricle depicted in the diagnostic image frame as midway between the anterior epicardial border and the posterior epicardial border along a line through the first and second tentative center-points; and

e. indicating the position of the final center-point in connection with the diagnostic image frame of the left ventricle.

2. The method of claim 1, wherein the step of determining the position of the posterior epicardial border of the left ventricle further includes filtering the image frame in a lower half region of a sector scan at pixel locations having intensities which exceed the maximum intensity value of the image by more than 60%.

3. The method of claim 1, wherein the step of determining the position of the posterior epicardial border of the left ventricle further includes filtering the image frame in multiple passes, wherein a first pass produces a best column corresponding to the column which yields the most consistent maximum filter output values across the frame.

4. The method of claim 3, wherein a second pass searches in a small band of rows which are proximate the best column, wherein the second pass produces a best row corresponding to the row which yields the most consistent maximum filter output values across the image frames.

5. The method of claim 4, wherein a third pass searches in a small region of columns and rows which are approximately the best column and best row, and thereby determines the position of the posterior epicardial border of the left ventricle depicted in the image frame.

6. The method of claim 5, wherein the step of determining the position of the anterior epicardium of the left ventricle depicted in the image frame comprises the steps of:

a. approximating first tentative positions of the right and left epicardium by repeatedly filtering the image frame with a set of elliptical arc matched filters until a maximum value for the set of filters is obtained;

b. establishing a search region for the anterior epicardium by;

(1) calculating a cumulative intensity level distribution in the anterior direction by summing across a plurality of elliptical arcs determined by the distance from the second tentative center-point and the first tentative positions of the right and left epicardium,

(2) establishing a threshold for the intensity level distribution,

(3) determining the rows in which the anterior epicardium may exist by determining the first and last rows which contain intensity values which consistently exceed the threshold across the image frames;

c. searching the search region by;

(1) performing a second derivative operation upon the pixels that lie within the search region and generating a plurality of anterior epicardium pixel locations which exceed a predetermined threshold; and

(2) determining the anterior epicardium pixel locations which have the smallest variation in position across the image frame.

7. The method of claim 6 wherein said region is limited to a range of approximately 25% of the major epicardial diameter of the ventricle.

8. The method of claim 5 further comprising the steps of:

a. computing several equiangular regions wherein said regions are grouped into grouped regions and contain cumulative gray level distribution of pixels at selected radial distances from the center-point;

b. computing said first, second, and final center points again along the elliptical arcs parallel to the estimated epicardial order;

c. summing the amplitude distributions along the elliptical arcs parallel to the epicardial border; and

d. calculating the figures of merit of the elliptical arc filters to better define the epicardial position estimates.

9. The method of claim 8 wherein said equiangular regions comprise 64 such regions.

10. The method of claim 9 wherein said equiangular regions are grouped into 8 regions.

11. The method of claim 8 further comprising the step of averaging said level distributions at each said radial length.

12. The method of claim 11 further comprising the step of repeating the averaging step along the elliptical arcs parallel to the estimate epicardial border.

13. The method of claim 8 further comprising the step of calculating a figure of merit by defining a small region within the chamber cavity using the estimated position of the epicardium.

14. The method of claim 13 wherein said small region is further defined by a lower limit equal to the row of the center-point, an upper limit equal to 1/2 of the anterior radius, and a width equal to the columns at ±

1/4 the radius to the left and right of the column center-point.

15. The method of claim 13 wherein the average intensity value or "chamber" gray level of said small region is calculated and compared to the intensity values of the elliptical arcs defining the epicardium and wherein if the epicardium intensity data is less than or equal to the chamber gray level within its region, then the epicardial data is considered to be poor, and wherein if the gray level of the epicardium within a region is greater than the chamber mean but less than the mean gray level for the region plus 1/2 of its standard deviation, the epicardial data is questionable, and wherein if the epicardial gray level for the region is greater than the mean of its region plus of its standard deviation, then the estimated position is likely good.

16. The method of claim 5 further comprising the steps of:

a. computing several equiangular regions wherein the regions contain cumulative gray level distribution of pixels at selected radial distances from the center-point and said equiangular regions are used to adjust the epicardial border;

b. averaging the cumulative distributions at each radial length for each region;

c. computing the averages again along the elliptical arcs parallel to the estimated epicardial order;

d. summing the distributions along the elliptical arcs parallel to the epicardial border; and

e. computing the overall average intensity level along the arc corresponding to the epicardial curve for each region.

17. The method of claim 1 further comprising the steps of:

a. repeating said search to further define the anterior epicardial pixel locations which have the smallest variation in position across the image frame; and

b. searching the further defined epicardial border regions using elliptical filters which are dynamically determined by the distance to the anterior and posterior walls, the semi-major axes, and by the distance from said second tentative center-point to the right and left border regions.

25. The method claim 1 further comprising the steps of:

a. searching the left and right epicardial border regions a second time using elliptical filters which are dynamically determined by the radius to the anterior and posterior walls, the semi-major axes, and by the distance from the center-point to the right and left border regions; and

b. calculating the radial separation between non-zero weights of the filter as proportions of the vertical diameter of the left ventricle to optimize the separations, produce the best gradient information, and ensure that the search will not produce a gradient which reaches across from inside the endocardium to outside of the epicardium.

18. A method for quantitatively analyzing a diagnostic digital image of the left ventricle, including the steps of:

a. determining a first tentative center-point for and a region of search for the posterior epicardial border of the left ventricle depicted in the diagnostic image by repeatedly filtering the diagnostic image with a set of circular arc filters until a maximum value for the set of filters is obtained;

b. determining a second tentative center-point for and a region of search for the anterior endocardial and epicardial borders of the left ventricle depicted in the diagnostic image by repeatedly filtering the diagnostic image with a set of coupled elliptical arc filters until a maximum value for the set of coupled filters is obtained;

c. determining a final center-point for the left ventricle depicted in the diagnostic image as midway between the anterior epicardial border and the posterior epicardial border along a line through the first and second tentative center-points; and

d. indicating the position of the final center-point in connection with the diagnostic image.

19. A method for quantitatively analyzing a diagnostic digital image of the heart, said images containing rows and columns of pixels, said method including the steps of:

a. identifying a maximum pixel value in a portion of said image where an interface of myocardium and lung is expected;

b. selecting two threshold values t1 and t2 which are fractions of said maximum pixel value;

c. creating a first image having pixel values by steps comprising;

i. assigning a value of zero to locations in the image where the pixel values are less than t1;

ii. assigning a value of 1 to locations in the image where the pixel values are between t1 and t2; and

iii. assigning a value of 2 to locations in the image where the pixel values are greater than t2;

d. determining an expected vertical diameter of said myocardium;

e. summing the gray level values of pixel locations along a 180 circular arc having a radius r which varies from 1/4 and 3/4 of said expected diameter;

f. calculating a totalsum=sum(r)+2sum(r+step)-sum(r-step)-sum(r-2step)-(r-3step)-sum2(r-step), where sum(r) is the sum of the pixel values along the circular arc, sum2(r) is the sum of the pixel values along a 90 subarc symmetric with respect to the vertical centerline of the left ventricle;

g. calculating t=totalsum*min(r, expectedradius); and

h. comparing the t values for each row and selecting the row with the highest t value as the anterior epicardial border.

20. The method of claim 19, which further comprises the step of determining the position of the posterior epicardial border of the left ventricle by filtering the image frame in a lower region of the image at pixel locations having intensities which exceed the maximum intensity value of the image by more than 60%.

21. The method of claim 19, which further comprises the step of determining the position of the posterior epicardial border of the left ventricle by filtering each image frame in multiple passes, wherein a first pass produces a best column corresponding to the column which yields the most consistent maximum filter output values across the image frame.

22. The method of claim 21, wherein a second pass searches in a small band of rows which are proximate the best column, and wherein the second pass produces a best row corresponding to the row which yields the most consistent maximum filter output values across all image frames.

23. The method of claim 22, wherein a third pass searches in a small region of columns and rows which are approximately the best column and best row, and wherein the third pass determines the position of the posterior epicardial border of the left ventricle depicted in the image frame.

24. The method of claim 19 wherein said threshold values t₁ and t₂ are equal to 1/2 and 3/4, respectively.

26. A method of calculating a figure of merit for a left ventricle comprising the steps of:

a. defining a small region within a chamber cavity which estimates the position of the epicardium, said region defined having a lower limit equal to the row of the center-point of the left ventricle, an upper limit equal to 1/2 of the anterior radius, and having a width equal to the columns at ±

1/4 the radius to the left and to the right of the column of the center-point;

b. calculating the average intensity value of the region; and

c. comparing said average to the intensity values of the elliptical arcs defining the epicardium wherein if the epicardium intensity data is less than or equal to the chamber gray level within its octant, then the epicardial data is considered to be poor, and wherein if the gray level of the epicardium within an octant is greater than the chamber mean but less than the mean gray level for the octant plus 1/2 of its standard deviation, then epicardial data is questionable, and wherein if the epicardial gray level for the octant is greater than the mean of its octant plus of its standard deviation, then the estimated position is likely good.

27. The method of claim 26 wherein said equiangular regions comprises 64 such regions.

28. The method of claim 26 further comprising the step of enhancing the image using the estimated region of interest and figures of merit.

29. The method of claim 26 further comprising the step of facilitating the automatic regional amplification using the figures of merit.

30. A method of determining a wall motion using a set of time varying echocardiographic image frame comprising the steps of:

a. determining the center-point of the left ventricle;

b. forming eight one-dimensional amplitude distributions for each of eight 45°

sectors emanating from the center-point; and

c. obtaining the cross-correlation for each said amplitude distributions between the frame at end-diastole and the frame at end-systole to detect the wall motion in each of the different regions wherein the amount of radial shift producing the maximum cross-correlation gives an estimate of the amount of wall motion in each sector between the two time frames studied.

31. A method of detecting dropout across the interventricular septum between the left and right ventricles comprising the steps of:

a. comparing the minimum average grey level of a small image region in the cavity of a left ventricle around the center-point with the maximum average grey level of a set of similarly sized regions chosen along a line from the center-point of the ventricle in the direction of the septum; and

b. wherein if the maximum ratio of the peak level to the background level for the series of comparisons is less than two, signal dropout is considered to have been detected.

32. Method of automated video densitometry comprising the steps of:

a. obtaining control frames from a baseline period;

b. identifying the sides and last row of a sector scan;

c. detecting the approximate position of the endocardial and epicardial borders by using the endocardial borders derived during the control period as the region of interest for the left ventricular chamber;

d. injecting a contrast agent;

e. obtaining post-contrast frames;

f. estimating the position of the posterior epicardium on each frame by computing a posterior circular arc filter;

g. translating the regions of interest based upon the difference in position of the posterior circular arc filter and the position of the posterior circular arc filter in the first control frame;

h. calculating the mean pixel intensity of the left ventricular chamber for the pixels enclosed by the approximated endocardial border, and similarly calculating the mean pixel intensity of the right ventricular outflow tract;

i. comparing the mean pixel intensity for the sequence of end diastolic frames with the background pixel intensity means from the time-mean brightness curve for each injection;

j. segmenting the regions of the myocardium into regions approximating distributions of coronary arteries; and

k. estimating whether perfusion is present in a region by comparing a pre-contrast mean pixel brightness to that calculated after injection of the contrast agents.

33. A method for quantitatively analyzing a diagnostic image of a left ventricle including the steps of:

a. generating a diagnostic image frame of a left ventricle having a posterior epicardial border and an anterior epicardial border;

b. determining a first tentative center-point (x₀, y₀) for said left ventricle and the position of said posterior epicardial border by repeatedly filtering the image frame with a set of circular arc filters until a maximum value for the set of filters is obtained;

c. determining a second tentative center-point (x₁, y₁) for said left ventricle and the position of said anterior epicardium by repeatedly filtering the image frame with a set of matched elliptical arc filters until a maximum value for the set of matched filters is obtained; and

d. determining a final center-point for said left ventricle by using said first and second tentative center-points and weighting functions.

34. The method of claim 33 wherein the step of determining a final center is further comprised of the steps:

a. selecting a narrow band about said y₀ value;

b. forming a one-dimensional distribution D(y) by summing the pixel values in bands along circular arcs centered at (x₀, y₀);

c. assigning a value to an epicardial weight function epiW(y) at the location y, wherein said epiW(y) is 1 if D(y)>

2/3 of the maximum for the entire distribution, otherwise said epiW(y) is 0; and

d. assigning a value to an endocardial weight function endoW(y) at the location y, wherein said endoW(y) is 1 if D(y)>

1/2 of the maximum for the entire distribution, otherwise said endoW(y) is 0.

35. The method of claim 34 further comprising the step of testing for a cavity by:

a. forming a second one-dimensional distribution E(y) defined by the average (D(y-10)+D(y-9) . . . +D(y+9)+D(y+10))/21;

b. determining maximum E_max and minimum E_min values of E(y);

c. increasing said endoW(y) to 2, if endoW(y)=1 and E(y-10)<

(2*E_min +E_max)/3; and

d. increasing said epiW(y) to 2, if epiW(y)=1 and E(y-10)<

(2*E_min +E_max)/ 3.

36. The method of claim 35 further comprising the steps of:

a. locating a pixel Q(x,y) near a column containing said point (x₀,y₀);

b. calculating a filter value at (x,y) defined by V(x,y,k,k+c)= (s_k -s_k+c)*epiW(y₁) *(s_k+c -s_k)*endoW(y₂)!(k-(k+c)), wherein s_k represents the sum of pixel values on a selected set of points distributed along an elliptical arc of radius at a distance k from the center, y₁ <

y₂, y₁ =y-k, y₂ =y-(k+c), and c is an incremental constant;

c. locating the second tentative center-point for the left ventricle (x₁, y₁) as the maximum resulting pixel location; and

d. determining the center-point of the left ventricle (x₀, y₂), wherein y₂ =(y₀ +r₀ +y₁ -r₁)/2, r₀ is the distance from the first tentative center to the epicardial border, and r₁ is the distance from the second tentative center to the epicardial border.

37. A method for determining the optical density of a digital image of a left ventricle said method comprising the steps of:

a. obtaining a control frame from a baseline period prior to injecting any contrast agent;

b. eliminate any patient data or irrelevant information from said image;

c. approximating the position of the endocardial and epicardial borders on said control frame;

d. establishing a region of interest for said right ventricular outflow tract;

e. injecting contrast agent and obtaining an ED frame;

f. calculating the mean pixel intensity of the left ventricular chamber for the pixels enclosed by the approximate endocardial contour, and likewise calculating the mean pixel intensity of the right ventricular chamber to obtain right and left ventricular time-brightness curves;

g. determining the maximum right ventricular mean pixel brightness and the area under the right ventricular time-brightness curve above background;

h. determining the minimum left ventricular mean pixel brightness and the area under the left ventricular time-brightness curve above background;

i. obtaining a time-mean pixel brightness curve by comparing the mean pixel brightness for said ED frame following injection of the contrast agent to the background frame;

j. approximating the epicardial and endocardial contours of the peristomal short axis view; and

k. calculating the mean pixel brightness of the myocardium of the left ventricle as the mean pixel brightness of the pixel between the epicardial and endocardial borders.

38. The method of claim 37 further comprising the step of determining whether perfusion is present in a region by comparing a pre-contrast mean pixel brightness to that calculated after peripheral injection.