JPH0418347B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a threshold determination method for separating cell nuclei, cytoplasm, and other (eg, background) regions in an apparatus for automatically classifying cells.

The conventional threshold value determination method, as already described in Japanese Patent Application No. 87286/1986 by the same applicant,
A method was used to determine threshold values for separating the cell nucleus and cytoplasm, and the cytoplasm and others, respectively, based on a fixed ratio from the minimum and maximum values of the density of the cell image.
However, although this method is effective when there are small fluctuations in the cytoplasmic concentration, there is a problem in that it is not always possible to obtain an accurate threshold when the cytoplasmic concentration varies widely depending on the cell type. Ta.

An object of the present invention is to provide a method for stably and accurately determining threshold values for separating the cell nucleus and cytoplasm, and the cytoplasm and others, from a concentration histogram for cells in which the concentration and size of the cytoplasm vary greatly.

The stained cells have different concentrations of the cell nucleus, cytoplasm, etc., and the concentrations decrease in the above order. Therefore, the density histogram of the cell image created by photoelectrically converting the cell image into an electric signal and quantizing the electric signal has a distribution corresponding to the cell nucleus, cytoplasm, and the like. Therefore, in the present invention, the concentration histogram is defined as (A) cell nucleus;
(B) When there are three peaks corresponding to the cytoplasm and other distributions, that is, the cytoplasm has an appropriate concentration and size; (B) When the cytoplasm distribution is buried in the distribution of the cell nucleus, that is, the cytoplasm concentration is close to the concentration of the cell nucleus; (C) the distribution of the cytoplasm is buried in other distributions, that is, the concentration of the cytoplasm is close to other concentrations; and (D) the distribution of the cytoplasm is small and the distribution is unclear. In each case, we will determine how to determine the threshold that separates the cytoplasm from the cell nucleus, and the cytoplasm from other parts.

Furthermore, if the peak of the cytoplasmic concentration distribution in the above concentration histogram is of a peak type among the above (A) to (D), judging from the distribution between the peaks, there is no cytoplasmic peak in the above (B) or ( In the case of the two-peak type (C) or (D), the threshold value is determined based on the relationship between the cell nucleus or other distribution where the cytoplasm is buried and the distribution that does not include the cytoplasm, and the threshold is determined based on the relationship between the distribution of the cytoplasm and the distribution that does not include the cytoplasm. This provides a possible method.

Hereinafter, the present invention will be explained in detail with reference to Examples. FIG. 1 shows a configuration for implementing the present invention. First, a quantized cell image is stored in the image memory 1, read out, a density histogram is created in the histogram creation circuit 2, and this is stored in the computer 3.
Load into. The computer 3 processes this concentration histogram, analyzes the number of peaks in the concentration histogram and the shape of the distribution, and determines a threshold T _N that separates the cell nucleus from the cytoplasm, and a threshold T _C that separates the cytoplasm from the rest. Next, a threshold value T _N is set in the threshold value circuit 4, and an image of the cell nucleus is created whose region is a portion of the cell image stored in the image memory 1 that is equal to or greater than T _N. Furthermore, a threshold value T _C is set in the threshold value circuit 4, and the image memory 1
This is to create an image of the cell nucleus and cytoplasm with the area above T _C among the cell images in .

Below, the processing procedure performed by the computer 3 will be explained in the second section.
This will be explained using FIGS. 7 to 7. Note that FIG. 2 is a flowchart showing the processing procedure of the computer 3.

(1) In FIG. 1, the computer 3 processes the concentration histogram obtained from the histogram creation circuit 2, detects the peaks of the cell nucleus, cytoplasm, and other concentration distributions, and divides the concentration distribution into 3-peak type, 2-peak type, and 2-peak type depending on the number of peaks. It is classified into three types: peak type and one peak type.

(2) Next, if the classification result in (1) above is a two-peak type, it is further subdivided into three types depending on the shape of the cytoplasmic distribution. In other words, in Fig. 3, the peak P _p of other distributions and the peak P _o of the distribution of cell nuclei
Find the density value D _nio that has the minimum count value A between . Next, the two density values D _{l on the lower density side than D nio and the density value D u on the} higher density side than D _nio are calculated by adding a predetermined offset value V _pff to D _nio (D _nio + V _pff ). After obtaining the values, they are classified according to the conditions of the following formula. That is, D _l −P _p ≧C ₀ ...(1) P _o −D _u ≧C ₁ ...(2) (P _o −D _u )≧(D _l −P _p ) ...(3) However, C ₀ , C ₁ ...constants Here, if either of the above conditions (1) or (2) is satisfied, and also the condition of (3) is satisfied, the distribution of the cytoplasm is buried in the distribution of the cell nucleus. determine the type.

If either of the above conditions (1) and (2) are satisfied, but the condition (3) is not satisfied, it is determined that the cytoplasmic distribution is buried in other distributions.

If both the above conditions (1) and (2) are not satisfied, the cell is judged to have a small cytoplasm and an unclear type.

(3) Once the above classification is completed, next find a threshold value T _N that separates the cell nucleus from the cytoplasm, and a threshold value T _C that separates the cytoplasm from the rest.

When the distribution type obtained by the concentration histogram classification process is a three-peak type, the threshold value T _N is set between the peak P _o of the distribution of the cell nucleus and the peak P _C of the distribution of the cytoplasm, as shown in Fig. 4. , 2 with the value obtained by adding the offset value V _pff to the minimum count value B
The two concentration values D _l and D _u are determined, and the threshold value T _N is calculated by T _N (D _u −D _l )·W3U+D _l (4). In addition, the threshold T _C is the peak P _p of other distributions and the peak of cytoplasmic distribution.
P _C , two density values D l ′ and D _{u ′} having the value obtained by adding the offset value V _pff to the minimum count value C are found in the same way as when finding the threshold value T _N above, and T The threshold T _C is calculated by _c = (D _u ′−D _l ′)・W3L+D _l ′ (5). Here, W3U and W3L used in equations (4) and (5) are division parameters, and each is set to a value of about 0.3 to 0.6.

For example, if W3L and W3U are both 0.35,
In the density histogram shown in Figure 6, D _u
= 206, D _l = 146, D _u ′ = 88, D _l ′ = 76,
The threshold values T _N and T _c are determined by equations (4) and (5), and T _N =
167, T _C =80.

In addition, the threshold values T _N and T _C are determined from the ratio of each concentration difference, T _N = (P _o − D _u )・(D _u − D _l )/D _l −P _C + D _l ……(4)′ T _C = (P _C −D _u ′)・(D _u ′−D _l ′)/D _l ′−P _p +D _l ′…
…(5)′ can also be obtained.

(4) If the result obtained by the concentration histogram classification process is a two-peak type, and furthermore, the two-peak type classification process shows that the distribution of the cytoplasm is buried in the distribution of the cell nucleus, the threshold T _N As shown in Fig. 3, and T _C are two values that have the value obtained by adding the offset value V pff to the minimum count value A between the peak P _o of the distribution of cell nuclei and the peak _{P p} of other distributions. It is calculated from the density values D _l and _Du by the following formula.

T _C = (D _u − D _l ) · W2L + D _l } T _N = (P _o − T _C ) · W20 + T _c ... (6) If the cytoplasmic distribution is buried in other distributions, the threshold T _N and T _C is calculated by determining the concentration values D _l and D _u in the same manner as above, and using the following formula.

T _N = (D _u − D _l ) · W2L + D _l T _C = (T _N − P _p ) · W21 − P _p … (7) In the case of a type with small and unclear cytoplasm, the concentration value is calculated in the same manner as above and Determine D _l and D _u and calculate using the following formula.

T=(D _u −D _l )・W2L+D _l T _C =(T−P _C )・W22L+P _C T _N =(P _o −T)・W22U+T …(8) Here, equation (6), (7) Equation, W2L used in equation (8),
W20, W21, W22L, and W22U are division parameters, each of which is set to a value of about 0.2 to 0.7.

For example, if the concentration histogram shown in Figure 7 shows a type in which the cytoplasmic distribution is buried in other distributions, W2L = 0.2, W21 = 0.24, and D _u =
202, D _l = 114, P _p = 58, the thresholds T _N and T _C are
According to equation (7), T _N =132 and T _C =76.

Also, if the concentration histogram is a two-peak type,
It can also be determined by the following method. That is, as shown in FIG. 5, the half width of the distribution is obtained from the peaks P _o and P _p of the cell nucleus and other peaks, and the threshold values T _N and T _C are determined by the following equations.

T _N = P _o −w・W T _C = P _p +w′・W′ ...(9) Here, W is the half-width on the low concentration side of the distribution of cell nuclei, and W′ is the high concentration side of the other distributions. is the half-width of In addition, w and w' are partitioning parameters specific to each case: when the distribution of the cytoplasm is buried in the distribution of the cell nucleus, when the distribution of the cytoplasm is buried under another distribution, and when the cytoplasm is small and unclear, respectively. Set a value of 2 or more.

(5) If the classification result of the concentration histogram is one-peak type, set the threshold values T _N and T _C to predetermined values, respectively.

(6) When the above processing is completed, the computer 3 sets the threshold value T N for dividing the cell nucleus and cytoplasm into the threshold circuit 4 to obtain the area of the cell nucleus, and then sets the threshold value T _N for dividing the cell nucleus and other parts. T _C is set in the threshold circuit 4, and the areas of the cell nucleus and cytoplasm are determined.

As described in detail above, according to the present invention, by quantitatively processing the shape of the density histogram, it is possible to select a threshold determination method that flexibly responds to various changes, thereby stably and accurately determining the threshold value. A threshold value can be determined.

[Brief explanation of drawings]

FIG. 1 is a block diagram for determining a threshold value according to the present invention, FIG. 2 is a flowchart showing the processing procedure of the computer 3 in FIG. 1, and FIGS.
This figure is a characteristic diagram showing a density histogram for explaining the process of FIG. 2. 1... Image memory, 2... Histogram creation circuit, 3... Computer, 4... Threshold circuit.

Claims (1)

[Claims] 1. Determine a first threshold value that separates the cell nucleus from the cytoplasm, and a second threshold value that separates the cytoplasm from the background, from the density histogram of the cell image created by quantizing the electrical signal obtained by photoelectrically converting the cell image. In the method, the concentration histogram is classified into 3-peak type and 2-peak type according to the number of peaks,
In the peak type, the first and second threshold values are determined near the concentrations in the two valleys between the peaks, and in the two-peak type, a predetermined offset value is further added to the minimum count value between the two peaks. The first concentration difference between the concentration value on the high concentration side corresponding to the count value and the concentration of the first peak on the higher concentration side, and the low count value corresponding to the above minimum count value plus a predetermined offset value. A second concentration difference between the concentration value on the concentration side and a second peak concentration on the lower concentration side is calculated, and the distribution of the cytoplasm is buried in the distribution of the cell nucleus due to the first and second concentration difference. Is it the first type or the second type where the cytoplasmic distribution is buried in the background distribution?
It is determined whether the cytoplasm is small and its distribution is unclear, and if it is the first type, a second threshold value is determined near the concentration of the minimum count value between the peaks. The first threshold is determined from the second threshold and the concentration of the first peak, and if it is the second type, the first threshold is determined near the concentration of the minimum count value between the peaks, and the first threshold is determined in the vicinity of the concentration of the minimum count value between the peaks. The second threshold value is determined from the first threshold value and the concentration of the second peak, and if it is the third type, the concentration near the concentration of the minimum count value between the peaks and the concentration of the first peak are determined. A threshold value determination method characterized in that the first threshold value is determined from the above, and the second threshold value is determined from the concentration near the concentration of the minimum count value between the peaks and the concentration of the second peak.