JP4801974B2 - Cell image processing device, corneal endothelial cell image processing device, cell image processing program, and corneal endothelial cell image processing program - Google Patents

Description

  The present invention relates to a cell image processing device, a corneal endothelial cell image processing device, a cell image processing program, and a corneal endothelial cell image processing program for analyzing a cell image captured by an imaging means.

  Conventionally, an image input unit for capturing an image of a subject's corneal endothelial cells and an image signal obtained by the image input unit are selected from any of the captured corneal endothelial cells. An outer shape identifying means for identifying the outer shape of the corneal endothelial cell, a corneal endothelial cell area calculating means for calculating the area of the corneal endothelial cell based on the outer shape data identified by the outer shape identifying means, and the cornea A corneal endothelial cell measuring device is known that includes cell density calculation means for calculating the cell density of corneal endothelial cells based on the area data of a plurality of corneal endothelial cells calculated by the endothelial cell area calculation means (see Patent Document 1). ).

  The outer shape identification means inputs and displays a designated point at substantially the center of the imaged arbitrary corneal endothelial cell (cell image), and a plurality of radial lines and corneal endothelial cells (cell image) extending radially from the designated point. The outer shape of the corneal endothelial cell is identified by detecting the intersection point with the edge which is the cell wall.

In addition, a monitor that displays a photographed corneal endothelial cell image, an input device that marks the center point of each cell in the cell image displayed on the monitor, and the coordinates of the marked center point are specified and marked There is known a corneal endothelial cell measuring device that includes a calculation processing device that detects a cell wall between center points and calculates the area of each cell based on the detection result (see, for example, Patent Document 2).
JP-A-9-31442 JP 2001-314374 A

  However, in the former corneal endothelial cell measuring apparatus, the processing is continued even if not all the radial lines have detected edges, and therefore the average value is substituted when it cannot be detected. For this reason, a noisy cell image does not necessarily indicate a cell wall.

  It is also conceivable that the detected cell image overlaps with the adjacent cell image, or the adjacent cell images are not in contact with each other.

  Furthermore, in the latter corneal endothelial cell measuring apparatus, since the cell wall between the marked center points is detected, it is difficult to always obtain an accurate cell density.

  Accordingly, the present invention provides a cell image processing device, a corneal endothelial cell image processing device, a cell image processing program, and a cell image processing device capable of obtaining a more accurate cell density by performing image processing as preprocessing before cell density detection. An object of the present invention is to provide a corneal endothelial cell image processing program.

In order to achieve this object, the invention of claim 1 includes an edge detection means for detecting an edge which is a cell wall of each cell of a cell image, and a cell comprising a plurality of cell walls based on the edge detected by the edge detection means. And assuming a normal extending from the plurality of cell walls of each cell thus determined toward the inside of each cell, and determining the position of the average length of the normal from the cell wall into the cell. The reference point calculation means for automatically calculating the reference point that is the center of each cell from the position of the average length of the normal line from the cell wall into the cell, and the edge detected by the edge detection means a cell feature calculating means for determining the area and shape of each cell as a cell characterized by, the obtaining the number of cells from the number of the reference point, the operation control for obtaining the cell density from the number and the cell characteristics of the obtained cell Characterized in that it comprises a road.

According to a second aspect of the present invention, there is provided an edge detecting means for detecting an edge which is a cell wall of each cell of a corneal endothelial cell image, and a reference for obtaining a reference point in each cell based on the edge detected by the edge detecting means. Point calculation means; cell feature calculation means for determining the area / shape of each cell as a cell feature from the edges detected by the edge detection means;
Is the arithmetic control circuit and a large cell area and shape of the cell is determined to be divided larger than the cell of the average size for determining the flattening of the respective cells from the reference point and the cell feature a cell dividing means for dividing the operation control the large cells is controlled actuated by the circuit when, if the area and shape of the cells is small cells that have been determined to be integrated smaller than the cell of average size A cell integration unit that is controlled by the arithmetic control circuit to integrate the small cells, and the arithmetic control circuit performs an operation control of a cell density calculation unit after dividing and integrating the cells, The cell density is obtained from the number of.

  Furthermore, in the invention according to claim 3, the arithmetic control circuit determines whether or not the cell wall can be detected, and if it can be detected, the cell division / integration processing is automatically performed, and then the cell density is determined. The corneal endothelial cell image processing device is characterized in that the cell density is obtained from the number of cells by controlling the operation of the calculating means.

According to a fourth aspect of the present invention, there is provided an edge detecting means for detecting an edge which is a cell wall of each cell of a corneal endothelial cell image, and a reference for obtaining a reference point in each cell based on the edge detected by the edge detecting means. Point calculating means; shape information forming means for forming predetermined peripheral shape information for imposing the reference point for image processing; peripheral shape expanding means for expanding the peripheral shape information from the reference point; and the peripheral shape A corneal endothelial cell image processing apparatus comprising: the peripheral shape information expanded by an expansion means; and a cell wall detection means for detecting the shape of the cell wall information as an edge from the cell wall information, wherein the peripheral shape expansion is adjacent to each other When the plurality of peripheral shape information expanded by the means contact each other without contacting the cell wall information in the adjacent direction, the plurality of peripheral shapes that are in contact with each other Cell division in which the one large area cell is divided into a plurality of pieces so that the plurality of peripheral shape information in contact with each other is included in each cell, judging that the state information exists in one large area cell to be divided Means are provided .

According to a fifth aspect of the present invention, an edge detection step for detecting an edge which is a cell wall of each cell of a cell image, and a cell comprising a plurality of cell walls based on the edge detected by the edge detection step are obtained. Assuming a normal extending from the plurality of cell walls of each cell determined toward the inside of each cell, the normal is obtained by determining the position of the average length of the normal from the cell wall into the cell. A reference point calculation step for automatically calculating a reference point that is the center of each cell from the position of the average length from the cell wall to the inside of the cell, and from the edge detected in the edge detection step, a cell feature calculation step of calculating the area and shape as the cell characteristics, the determining the number of cells from the number of the reference points, cells densely obtaining cell densities several said cell characteristic of the obtained cell Characterized in that it comprises a calculation step.

According to a sixth aspect of the present invention, there is provided an edge detection step for detecting an edge which is a cell wall of each cell of a corneal endothelial cell image, and a reference for obtaining a reference point in each cell based on the edge detected by the edge detection step. A point calculation step; a cell feature calculation step for obtaining the area / shape of each cell as a cell feature from the edge detected in the edge detection step; and determination of division / integration of each cell from the cell feature and the reference point A cell dividing step for dividing the large cell when there is a large cell for which it is determined that the area / shape of the cell should be larger than a cell having an average size, and an area of the cell a cell integration step of integrating the small cell when the shape is small cells that have been determined to be integrated smaller than the cell of average size, - splitting of the cell After the coupling process, characterized in that it has a cell density calculation step of obtaining a cell density from the number of the cells.

Further, the invention of claim 7 determines whether or not the cell wall can be detected by the arithmetic control circuit, and if the cell wall can be detected, the cell division / integration processing is automatically performed, and then the cell The cell density is obtained from the number.

The invention according to claim 8 is an edge detection step for detecting an edge which is a cell wall of each cell of a corneal endothelial cell image, and a reference for obtaining a reference point in each cell based on the edge detected by the edge detection step. A point calculating step, a shape information forming step for forming predetermined peripheral shape information for imposing the reference point for image processing, a peripheral shape expanding step for expanding the peripheral shape information from the reference point, and the peripheral shape A corneal endothelial cell image processing program comprising: the peripheral shape information expanded by an expansion means; and a cell wall detection step for detecting the shape of the cell wall information as an edge from the cell wall information, wherein the peripheral shape expansion is adjacent to each other A plurality of the circumferential shape information expanded by the step contacted each other without contacting the cell wall information in the adjacent direction. In some cases, the plurality of circumferential shape information in contact with each other is determined to be present in one large area cell to be divided, and the one large area is included so that each of the plurality of circumferential shape information in contact with each other is included. A cell dividing step for dividing the cell into a plurality of cells is provided .

  According to such an image processing apparatus and image processing program, it is possible to obtain a more accurate cell density by performing image processing as a pre-process before cell density detection.

Hereinafter, embodiments of the present invention will be described with reference to the drawings. FIG. 1 shows a corneal endothelial cell imaging apparatus.
[Constitution]
As shown in FIG. 1, this corneal endothelial cell imaging apparatus captures corneal endothelial cells from an imaging means 1 such as a CCD camera that images corneal endothelial cells of a subject's eye and a corneal endothelial cell image captured by the imaging means 1. Corneal endothelial cell measuring device 2 that performs measurement and image processing, display means (display device) 3 such as a liquid crystal display (display) that displays images and data processed by corneal endothelial cell measuring device 2, and display means 3 An input means 4 such as a mouse for performing an operation on the image or data displayed on the screen is provided.

Although a CCD camera is employed as the imaging means 1, any imaging means 1 can be employed as long as it can image the corneal endothelium. Further, although the display unit 3 is used as the output unit, an output unit such as a printer can be added in addition to the display unit 3. Furthermore, although the mouse is used as the input means 4, other input means such as a keyboard and a light pen can be used.
<Corneal endothelial cell measuring apparatus 2>
The corneal endothelial cell measuring apparatus 2 includes an arithmetic processing means (arithmetic processing means) 5 and an image memory (storage means) 6.

  The arithmetic processing means 5 includes an arithmetic control circuit (arithmetic control means) 7 having a CPU, ROM, RAM, input / output interface, operation control circuit (operation control means), etc. (not shown). The arithmetic control circuit 7 reads and stores the program stored in the ROM at the time of activation into the RAM, thereby executing the calculation and control of the entire corneal endothelial cell photographing apparatus according to the program read into the RAM. . The image memory 6 includes an original image memory unit (original image storage unit) 6a that stores an original image captured by the imaging unit 1, and a processed image memory unit (processed image storage unit) that stores an image processed image. ) 6b and memory 6c.

  When the corneal endothelial cell image of the eye cornea as shown in FIGS. 5A and 6A is photographed by the imaging means 1, the corneal endothelium imaged by the imaging means 1 is stored in the original image memory unit 6a. A cell image is stored via the arithmetic control circuit 7.

  The arithmetic control circuit 7 displays the corneal endothelial cell image stored in the image memory 6 on the display means 3 and processes the image displayed on the display means 3 by operating the input means 4. Is supposed to do. This corneal endothelial cell image has a large number (plurality) of cells 10 surrounded by a cell wall (edge) 9.

  Further, the arithmetic processing means 5 has an image processing means 8 controlled by the arithmetic control circuit 7 as shown in FIG.

  As shown in FIG. 2, the image processing means 8 includes a zoom means 8a that enlarges / reduces the corneal endothelial cell image and enlarges the designated range of the corneal endothelial cell image and displays the designated area on the display means 3, and each cell. Edge detection means 8b for detecting 10 cell walls 9 as edges, and a nuclear reference point (nucleus) for obtaining the center of gravity or vertex or center point of each cell 10 based on the cell wall detected by the edge detection means 8b A reference point calculation means 8c for calculating the shape reference position) 13a.

  Further, the image processing means 8 calculates the cell feature to obtain features such as the area, perimeter, shape (circularity), maximum length, width, aspect ratio, etc. of the cell 10 surrounded by the cell wall 9 detected by the edge detection means 8b. Means 8f are provided.

  Further, the image processing means 8 has a nucleus removing means 8g for removing the nuclear reference point 13a of the cell 10 from the image, and a thinning means 8h for thinning the cell wall 9 portion of each cell 10. The thinning means 8h removes the beard extending from the cell wall 9 of the thinned cell 10 or the edge floating from the cell wall 9 with a limited length or the point.

  Further, the image processing means 8 has labeling means 8i for attaching a label such as coloring to the cell image whose cell wall has been thinned by the thinning means 8h. The labeling means 8i deletes an edge when the same label is provided on the top, bottom, left and right.

  Further, the image processing means 8 has cell cell wall detection means 8j for detecting a cell wall by expanding a circle from substantially the center of the labeled cell image. When the cell cell wall detecting means 8j is expanded from the reference point 13 (center of the nuclear reference point 13a) until a circumferential shape information image (annular image) such as a circle described later comes into contact with the cell wall 9 of the cell 10, A position where the circumferential shape information image (annular image) contacts is detected as a cell wall 9.

  Further, the image processing means 8 removes wrinkles and nuclei into one when the wrinkles and nuclei remain in the cell image processed by the labeling means 8i and one cell image is originally divided. Cell integrating means 8k to be integrated, and cell dividing means 8L for dividing the cell wall into cells when the cell images processed by the labeling means 8i originally have to be divided into a plurality of cell walls, and finally processed Cell density calculation means 8m is provided for analyzing a corneal endothelial cell processing image composed of a plurality of cell images to obtain a cell density.

Further, the image processing means 8 includes a shape information forming means 8n for forming a predetermined circumferential shape information image (annular image) including a reference point 13 for image processing, and a circumferential shape information image (annular image) from the reference point 13. ) Is expanded. The shape information forming means 8n forms a circumferential shape information image (annular image) which is other circumferential shape information such as a circle, an ellipse, or a polygon.
[Action]
Hereinafter, the control action by the arithmetic control circuit 7 of the corneal endothelial cell imaging apparatus having such a configuration will be described with reference to the flowcharts of FIGS.
(1) Outline of Flowchart for Cell Density Analysis FIG. 3 schematically shows an overall flow chart for analyzing cell density, and FIGS. 3 shows a corneal endothelial cell image that is captured and stored in the original image memory unit 6a and displayed on the display means 3. FIG. FIG. 6A is a partially enlarged view in which a part of FIG. 5A is enlarged and displayed by the zoom means 8a. 7 and 8 are corneal endothelial cell images different from those in FIGS. In FIGS. 7 and 8, the cell wall portion between the cells 10 and 10 is indicated by a white line for the sake of convenience.

  In FIG. 5, FIG. 6, FIG. 7, and FIG. 8, the corneal endothelial cell image is composed of a large number of cells (cell images, that is, corneal endothelial cell images) 10, and black portions between adjacent cells 10 are cells 10, 10 respectively. It becomes the cell wall between.

  In step S1, corneal endothelial cells of the eye to be examined are photographed by the imaging means 1, and the photographed corneal endothelial cell image is stored in the original image memory unit 6a via the arithmetic control circuit 7, and FIG. When the corneal endothelial cell image is displayed on the display means 3 as shown in FIG. 6A, the process proceeds to step S2.

  Next, in step S2, the arithmetic control circuit 7 executes a “cell wall detection” process.

  This “cell wall detection” process includes an “edge detection” process in step S21, a “whisker removal” process in step S22, and an “endpoint extension” process in step S23.

  In step S21, the image processing means 8 is controlled to detect the cell wall 9 between the cells 10 as shown in FIGS. 5B and 6B as an edge.

  Thereafter, the arithmetic control circuit 7 uses the reference point calculation means 8c shown in FIG. 2 to center the centroid or vertex of each cell 10 from the cell wall (edge) 9 shown in FIGS. 6 (a), 7 and 8. A nuclear reference point (nuclear reference position) 13a such as a center point is calculated. This calculation assumes a line or normal 14 extending from the plurality of points (positions) of each cell wall (edge) 9 toward the center or the center side of each cell 10 as shown in FIG. By voting at the average length in the direction of this line or normal 14 (determining the position of the average length), a nuclear reference point (nuclear reference position) 13a as shown in FIG. 9B is obtained. Is calculated.

  Thereafter, the arithmetic control circuit 7 calculates the address of the central position of the core reference point 13 a and stores the calculated address in the memory 6 c as the reference point (center of gravity) 13. Such storage is executed in order for each cell 10, and the process proceeds to step S22.

  In step S22, the arithmetic control circuit 7 thins the cell wall (edge) 9 as shown in FIG. 10 (a) by the thinning means 8h in FIG. 2, and then lengthens the beard 15 and the floating edge 15a. And the point 16 is removed, and the process proceeds to step S23.

  In step S23, the arithmetic control circuit 7 uses the thinning means 8h to extend and connect two end points that are the shortest distance within + -90 ° of each cell wall 9 to form one edge of the cell wall 9. For example, the end points a1 and a2 of the cell 10a in FIG. 10A, the end points a3 and a4 of the cell 10b, the a5 and a6, and the end points a7 and a8 of the cell 10c are connected as shown in FIG. 10B. Cell wall 9 is obtained. In addition, the cells which are not corrected in FIG. 10B are 10d1 to 10d5 and 10e1 to 10e4 in FIG. 10C.

  Then, the arithmetic control circuit 7 performs such “cell wall detection” processing in steps S21 to S23, and then proceeds to step S3.

  In this step S3, the arithmetic control circuit 7 executes the following “shape recognition” processing. This “shape recognition” processing includes “labeling” processing in step S31, “feature amount calculation” processing in step S32, “label division” processing in step S33, and “label integration” processing in step S34.

  In step S31, the arithmetic and control circuit 7 performs labeling processing for attaching labels L1, L2, L3... Ln to a plurality of adjacent cells 10 as shown in FIG. Transition.

In step S32, the arithmetic control circuit 7 causes the cell feature calculation means 8f to show the area, perimeter, shape (circularity) of the cell 10 surrounded by the cell wall 9 shown in FIG. 9A, as shown in FIG. Then, the aspect ratio obtained from the maximum length and width, and the ratio of the distance between the maximum length (Maxrad) and the minimum length (Minrad) from the reference point (center of gravity) to the cell wall 9 are obtained, and the process proceeds to step S33. . Here, the circularity is
Circularity = 4π × area / perimeter ²
As required. The aspect ratio is
Aspect ratio is calculated as maximum length / width.

In step S33, first of all, in the image subjected to the labeling process in step S31, a cell 10f that originally has a cell wall at the constricted portion indicated by arrow A1 in FIG. It calculates | requires from the cell characteristic of each calculated cell. That is, when the average size of each cell 10 is obtained, as is apparent from FIG. 9A , the cell is larger than the average size cell and there are a plurality of assumed reference points 13. Is detected as, for example, a cell 10f that originally has a cell wall.

  In this case, in step S33, the cell dividing means 8L is controlled to enter the dividing line 10La as the cell wall in the constricted portion A1 at the center of the cell 10f in FIG. 9 (a) as shown in FIG. 9 (d). Then, the cell 10f in FIG. 9A is divided into two cells 10f1 and 10f2 as shown in FIG. 9D, and the process proceeds to step S34.

  In this division processing, the detection of the dividing line of the cell 10f is executed by the arithmetic control circuit 7 controlling the shape information forming means 8n and the peripheral shape expansion means 8o of FIG. 2 as follows.

  First, as shown in FIG. 9 (c), the shape information forming means 8n has circles (circumferential shape information) R1, R2, centered on the central position (reference point 13) of the core reference point 13a of each cell 10. Alternatively, as shown in FIG. 13B, for example, circles (circumferential shape information) R1 and R2 centering on the reference points 13a1 and 13a2 are formed.

  Thereafter, the circumferential shape expanding means 8o expands the circles (circumferential shape information) R1 and R2 at the same rate (gradually increasing the diameter and gradually increasing the area of the circle).

  At this time, the arithmetic control circuit 7 detects the cell if the cell wall detection means 8j detects the cell wall in the vicinity of the portion where the circles R1, R2 are in contact, but detects the cell wall in the vicinity of the portion where the circles R1, R2 are in contact. If not, it is detected that two cells are one cell. If the cell wall is not detected, the dividing line 10La is inserted as a cell wall at the portion where the circles R1 and R2 are in contact as shown in FIG. 9D.

  Next, in step S34, the arithmetic and control circuit 7 determines that the cell wall should not originally exist in the portion indicated by the ellipse 19 in FIG. 13A in the image subjected to the labeling process in step S31. When there is an edge 20 such as a cell wall between 10b1 and 10b2, the cell integration means 8k is controlled to remove the edge 20, and the cells 10b1 and 10b2 are integrated, as shown in FIG. One cell 10b is selected, and the process proceeds to step S4.

  In this case, the presence of the edge 20 makes it possible to obtain from the cell feature of each cell calculated by the cell feature calculation means 8f that the cells on both sides of the edge 20 are much smaller than the average cell area. . In addition, if the area of the cell when the edge 20 is removed is within the same range or a predetermined range as the average cell area, it can be determined that the edge 20 is an unnecessary line. Therefore, in such a case, the original cell shape can be obtained by removing the edge 20.

  In this way, when the division / integration processing is completed, for example, a corneal endothelial cell image as shown in FIG. 19 is obtained.

In step S4, the arithmetic control circuit 7 controls the operation of the cell density calculating means 8m, and after the corneal endothelial cell image of FIG. 19 is obtained by the division and integration processing as shown in FIG. The inner area (total area of the cells) of the outer (outer) line 26 of the outermost cell 10 provided with. Then, the arithmetic control circuit 7 analyzes the cell density from the obtained total area of the cells, the outermost peripheral cell 10 and the number of the inner cells 10, displays the analysis result on the display means 3 and ends. To do. The image processed in each step is appropriately stored in the processed image memory unit 6b. The number of cells can be obtained from the number of reference points (not shown).
(2) Another Example of Cell Density Analysis In step S1, corneal endothelial cells of the eye to be examined are photographed by the imaging means 1, and the photographed corneal endothelial cell images are stored in the original image memory unit 6a via the arithmetic control circuit 7. When the corneal endothelial cell image as shown in FIGS. 5A and 6A is displayed on the display means 3 while being stored, the process proceeds to step S2.

  In step S2, “cell wall detection” processing is executed. That is, in step S21, the arithmetic control circuit 7 controls the operation of the image processing means 8 to detect the cell wall 9 between the cells 10 as shown in FIGS. 5B and 6B as an edge. Thereafter, the reference point 13 (not actually displayed) is calculated in the same manner as in the above-described embodiment.

  In step S21, the arithmetic control circuit 7 thins the cell wall (edge) 9 as shown in FIG. 10A by the thinning means 8h of FIG. 2, and proceeds to step S22.

  In this step S22, the arithmetic control circuit 7 controls the operation of the image processing means 8 to remove the whiskers 15 and points 16 etc. that are different from the cell wall 9 of the cell 10 shown in FIG. The “end point extension” process is executed, and the process proceeds to step S3.

In step S3, the processes of steps S31 to S34 are executed.
That is, in step S31, the arithmetic control circuit 7 changes the colors of the adjacent cells 10 to different labels as shown in FIG. 11B (for example, labels L1, L2, L3,. Ln, etc.) is applied, and the process proceeds to step S32. In step S32, as described above, the “feature amount calculation” process is executed, and the process proceeds to step S33.

  In step S33, the arithmetic control circuit 7 uses the cell feature calculation means 8f to add the cell 10a, which should originally have a cell wall, to the portion indicated by the ellipse 17 in FIG. 13B in the labeled image. When it is detected, the cell dividing means 8L is controlled so that the dividing line 18 is inserted as a cell wall as shown in FIG. 13 (c) in the central constricted portion of the cell 10a in FIG. 13 (b). Cell 10a is divided into two cells 10a1 and 10a2 as shown in FIG. 13C, and the process proceeds to step S34.

  In this case, the detection of the cell 10a is performed by placing the reference points 13a1 and 13a2 (not actually displayed) into the approximate center of the left and right portions of the cell 10a by operating the input means 4 described above, and by the shape information forming means 8n. For example, circles (circumferential shape information) R1 and R2 centering on the reference points 13a1 and 13a2 are formed, and the circles (circumferential shape information) R1 and R2 are expanded at the same rate (the diameter is gradually increased) by the peripheral shape expansion means 8o. Then, when the area of the circle is gradually increased), the cell cell wall detecting means 8j does not detect the cell wall in the vicinity of the portion where the circles R1 and R2 are in contact. When the cell cell wall detecting means 8j does not detect the cell wall, the dividing line 18 is inserted as a cell wall at the portion where the circles R1 and R2 are in contact.

  Further, as shown in FIG. 14 (a), there should be three cells 10 (labels L1 to L3) as shown in FIG. 14 (c), but between cells L1 and L3 in FIG. 14 (c). If the cell wall 9 is not as shown in FIG. 14A and is a cell with one large label L0, three reference points 13 (estimated points that are not actually displayed) are displayed as shown in FIG. 14B. The center circles R1, R2, and R3 are expanded (the diameter is gradually increased in the same manner), and a line passing through a portion where the circles R1, R2, and R3 are in contact is drawn as a cell wall. The cell with label L0 is divided into three cells 10 with labels L1 to L3 as shown in FIG.

  Next, in step S34, the arithmetic control circuit 7 should have no cell wall in the portion indicated by the ellipse 19 in FIG. 13A in the image subjected to the labeling process in step S23 in FIG. In addition, when an edge 20 such as a cell wall is detected between the cells 10b1 and 10b2 by the cell feature calculating unit 8f, the cell integrating unit 8k is controlled to remove the edge 20 and integrate the cells 10b1 and 10b2. As shown in FIG. 13C, one cell 10b is selected, and the process proceeds to step S4. The integration determination at this time is made based on whether or not the reference point 13 is present in the labeled cell.

  In this way, when the division / integration processing is completed, for example, a corneal endothelial cell image as shown in FIG. 19 is obtained.

In step S46, the arithmetic control circuit 7 controls the operation of the cell density calculating means 8m, and after the corneal endothelial cell image of FIG. 19 is obtained by the division and integration processing as shown in FIG. The area (total area) inside the line 26 connecting the outer periphery of the cell (cell 10) selected by each reference point 13 (not actually displayed) of the cell 10 is obtained, and the obtained area and the outermost cell 10 are obtained. The cell density is analyzed from the number of cells 10 inside and the inside thereof, and the analysis result is displayed on the display means 3 and the process is terminated. The image processed in each step is appropriately stored in the processed image memory unit 6b. In this case, it can also be determined from the shape.
(3) More specific example of cell integration processing When performing cell integration processing of a corneal endothelial cell image whose edges have been thinned in step S2 of FIG. 3, the arithmetic control circuit 7 first controls the operation of the image processing means 8. Then, the beard 15 and the point 16 that are different from the cell wall 9 of the cell 10 shown in FIG. 11A are removed, and the process proceeds to step S3.

In step S3, the processes of steps S31 to S34 are executed.
That is, in step S3, the arithmetic control circuit 7 changes the colors of the adjacent cells 10 to different labels as shown in FIG. 11B (for example, labels L1, L2, L3,. Ln etc.) is performed, and the process proceeds to step S32. In step S32, the above-described “feature amount calculation” process is executed, and the process proceeds to step S33. In step S33, the above-described “label division” process is executed, and the process proceeds to step S34. In this step S34, the “integration processing” in steps S341 to S344 of the flowchart shown in FIG. 4 is executed.

In step S341, the labeled image after the “label division” process in step S33 is displayed on the display unit 3, and the process proceeds to step S342.
(Small area integration)
FIG. 15A shows a labeled image, and the small area portions (small cells) 10a and 10b of the labels L1a and L2a are contained in the cells 10 and 10 having the average size of the labels L1 and L2 of the image. .

  In this case, in step S342, the arithmetic control circuit 7 controls the cell integration unit 8k in FIG. 2 to integrate the small area portion 10a of the label L1a into the cell 10 of the label L1, and the small area portion 10b of the label L2a. The small area portions 10a and 10b are not integrated as shown in FIG. 15B by integrating with the cell 10 of the label L2, and the process proceeds to step S343.

This process is a process of integrating a label having a predetermined area or less (for example, an area of 10 pixels or less) with an adjacent label. In this case, the label with the largest contact surface is integrated from the adjacent labels.
(Integration with 1 contact)
In step S343, the arithmetic control circuit 7 determines that the small area portion (small cell) 10a of the label L1 including the reference point 13 (not actually displayed) is in the cell 10 (L) as shown in FIG. 2, the cell integration means 8k of FIG. 2 is controlled to integrate the small area portion 10a of the label L1 into the cell 10 of the label L, so that the small area portion 10a is not as shown in FIG. 16B, step S344. Migrate to

In step S343, the arithmetic control circuit 7 determines that the small area portion 10b of the label L2 that does not include the reference point 13 (not actually displayed) is in the cell 10 (L) as shown in FIG. 2 is controlled to integrate the small area portion 10b of the label L2 into the cell 10 of the label L so that the small area portion 10b is not as shown in FIG. 17B, and the process proceeds to step S344. To do. The processing in step S343 is integrated with the label having the reference point 13 when there is only one label in contact with the label having the reference point 13.
(Integrated with circularity)
Further, when there are small area portions (small cells) 10a to 10c with labels L1 to L3 as shown in FIG. 18 (a), the reference point 13 is in the small area portion 10a of the label L1, but is labeled L2, L2. There are cases where the small area portions 10b and 10c of L3 are not present.

  In this case, in step S344, the arithmetic control circuit 7 removes (deletes) the edges e1 and e2 between the small area portion 10a and the small area portions 10b and 10c and the edge e3 between the small area portions 10b and 10c. Thus, it is determined from the presence or absence of the reference point 13 (not actually displayed) that the small area portions 10b and 10c of the labels L2 and L3 should be integrated into the small area portion 10a of the label L1.

In addition, the arithmetic control circuit 7 controls the cell integration means 8k of FIG. 2 based on this determination, integrates the small area portions 10a to 10c into one, and the small area portions 10a to 10c become the same as FIG. After the state of one cell 10 that is not as in (), the process proceeds to step S343 and loops. Then, in all the labeled cells 10, when the integration process in step S344 is completed, the integration process ends.
(Modification)
FIG. 20 schematically shows another example of the entire flowchart for analyzing the cell density.

In step S1a of FIG. 20, the cornea of the eye to be examined is imaged by the imaging means 1.
The corneal endothelial cell image picked up by the image pickup means 1 is stored in the original image memory 6 a via the arithmetic control circuit 7 and displayed on the display means 3. If there is noise in the corneal endothelial cell image displayed on the display means 3, the image is taken again. If the corneal endothelial cell image has little noise as shown in FIGS. 21 and 22, it is not always necessary to re-photograph if the cell wall between the cells (corneal endothelial cells) is clear.

  That is, the arithmetic control circuit 7 scans the entire corneal endothelial cell image in the specified range, and if the corneal endothelial cell image is clear from the detection ratio of the cell wall (edge) with respect to the area of the specified range, the step is performed. The process proceeds to S2a.

  In addition, since illumination light is irradiated to the cornea from diagonally and it is made to image | photograph from the opposite diagonal, the brightness of a corneal endothelial cell image differs between right and left eyes, and a difference arises in the detection accuracy of a cell wall. Therefore, in step S2a, the arithmetic control circuit 7 first corrects the shading of the corneal endothelial cell images of the left and right eyes, that is, if there is a difference in brightness between the corneal endothelial cell images of the left and right eyes, Correction is made so that the brightness of the cell image is the same, so that there is no difference in the detection accuracy of the cell wall. In step S2a, the arithmetic control circuit 7 performs filtering by the edge detection means 8b to remove the noise 22a as shown in FIGS. 5B and 6B.

  Next, in step S2a, the arithmetic control circuit 7 controls the operation of the image processing means 8, detects the cell wall 9 between the cells 10 as shown in FIGS. 5 and 6, and proceeds to step S3a.

  In this step S3a, the arithmetic control circuit 7 controls the operation of the image processing means 8, and performs a labeling process for attaching labels L1 to Ln to a plurality of adjacent cells 10 as shown in FIG. The process proceeds to S4a.

In step S4a, the arithmetic and control circuit 7 should divide the cells 10 labeled L1 to Ln from the cell characteristics such as the area and shape of the cells 10 labeled L1 to Ln subjected to the labeling process in step S3a. If it is determined whether or not to divide, division processing is performed, and the process proceeds to step S5a. For example, if the cell wall must be in the part indicated by the ellipse 13a1 in FIG. 13B, the cell 10a has a large area and the part indicated by the ellipse 13a1 is constricted. Therefore, the cell wall 18 is provided at the constricted portion as shown in FIG. 13C, and the cell 10a is replaced with the cell 10a1,1.
Divide into 0a2.

  Next, in step S5a, the arithmetic control circuit 7 does not have a cell wall in the portion indicated by the ellipse 19 in FIG. 13 (a) in the image subjected to the labeling process in step S3a. When there is an edge 20 such as a cell wall between the cells 10b1 and 10b2, it is determined that the cells 10b1 and 10b2 are small cells to be integrated from the cell characteristics such as the area and shape of the cells 10b1 and 10b2, and the cell integration means 8k. Is controlled to remove the edge 20 and the cells 10b1 and 10b2 are integrated. In addition, it is determined that the small area portion (small cell) as shown in FIGS. In this case, similarly, the cells to be integrated are obtained from the area and shape, and the integration process is performed.

  In this way, when the division / integration processing is completed, for example, a corneal endothelial cell image as shown in FIG. 19 is obtained.

  In this step S6a, the arithmetic control circuit 7 controls the operation of the cell density calculating means 8m, and after the corneal endothelial cell image of FIG. 19 is obtained by the division and integration processing as shown in FIG. The area inside the line 26 connecting the cell walls not adjacent to other selected cells of each cell (cell) selected by each reference point 13 (estimated point not actually displayed) of the cell 10 is obtained. That is, the total area inside the line 26 connecting the outermost cell walls of the outermost peripheral cell 10 among the plurality of cells 10 designated by the reference point 13 is obtained. Then, the arithmetic control circuit 7 analyzes the cell density from the obtained total area, the outermost peripheral cell 10 and the number of the inner cells 10, displays the analysis result on the display means 3, and ends. The image processed in each step is appropriately stored in the processed image memory unit 6b. Also in this case, the number of cells 10 can be obtained from the number of reference points 13 of the cells in which edges are detected.

  As described above, the cell image processing apparatus according to the embodiment of the present invention is based on the edge detection means 8b for detecting the edge that is the cell wall of each cell of the cell image, and the edge detected by the edge detection means 8b. Reference point calculation means 8c for obtaining a reference point in each cell, cell feature calculation means 8f for obtaining the area / shape of each cell as a cell feature from the edge detected by the edge detection means 8b, and the cell feature And an arithmetic control circuit 7 for obtaining a cell density from the number of each cell from the reference point.

  According to this cell image processing apparatus, it is possible to obtain a more accurate cell density by performing image processing as pre-processing before cell density detection.

  In addition, the corneal endothelial cell image processing device according to the embodiment of the present invention is based on edge detection means 8b for detecting an edge that is a cell wall of each cell of a corneal endothelial cell image, and an edge detected by the edge detection means 8b. A reference point calculation means 8c for obtaining a reference point in each cell, a cell feature calculation means 8f for obtaining the area / shape of each cell 10 as a cell feature from the edge detected by the edge detection means 8b, and the cell feature When there is a calculation control circuit 7 that determines division / integration of each cell from the reference point, and a large cell that is determined to be divided from the area / shape, the calculation control circuit 7 controls the operation of the cell. When there is a cell dividing means 8L for dividing a large cell and a small cell determined to be integrated from the area and shape, the operation control circuit 7 controls the operation of the small cell. And a cell integration means 8k to integrate. Moreover, when the arithmetic control circuit 7 detects that the cell wall is clear, it automatically divides and integrates the cells, and then controls the cell density calculation means 8m to control the number of the cells. From this, the cell density is obtained.

  In the corneal endothelial cell image processing apparatus according to the embodiment of the present invention, the arithmetic control circuit determines whether or not the cell wall can be detected. If the cell wall can be detected, the cell division / After the integration process, the cell density calculating means is operated to control the cell density from the number of cells.

  Further, the corneal endothelial cell image processing apparatus according to the embodiment of the present invention is based on edge detection means 8b for detecting an edge that is a cell wall of each cell of a corneal endothelial cell image, and the edge detected by the edge detection means 8b. A reference point calculating means 8c for obtaining a reference point in each cell, shape information forming means 8n for forming predetermined peripheral shape information (circles R1 to R3) for including the reference point 13 for image processing, From the circumferential shape expansion means 8o for expanding the circumferential shape information (circles R1 to R3) from the reference point 13, the circumferential shape information (circles R1 to R3) expanded by the circumferential shape expansion means 8o and the cell wall information Cell wall detecting means (8j) for detecting the shape of the cell wall information as an edge.

  Furthermore, in the corneal endothelial cell image processing apparatus according to the embodiment of the present invention, the plurality of pieces of peripheral shape information (circles R1 to R3) adjacent to each other and expanded by the peripheral shape expansion means 8o are adjacent to each other. When contacted with each other without contacting the cell wall information, it is determined that the plurality of circumferential shape information (circles R1 to R3) that are in contact with each other exist in one large area cell to be divided, Cell division means 8L is provided for dividing the one large area cell into a plurality of pieces so as to include the circumferential shape information (circles R1 to R3).

  According to such a corneal endothelial cell image processing apparatus, it is possible to obtain a more accurate cell density by performing image processing as pre-processing before cell density detection.

  The cell image processing program according to the embodiment of the present invention includes an edge detection step for detecting an edge that is a cell wall of each cell of a cell image, and a reference in each cell based on the edge detected by the edge detection step. A reference point calculation step for obtaining a point; a cell feature calculation step for obtaining the area / shape of each cell as a cell feature from the edge detected in the edge detection step; and the number of each cell from the cell feature and the reference point And determining a cell density.

  According to this cell image processing program, it is possible to obtain a more accurate cell density by performing image processing as preprocessing before cell density detection.

  Further, the corneal endothelial cell image processing program according to the embodiment of the present invention includes an edge detection step for detecting an edge that is a cell wall of each cell of the corneal endothelial cell image, and each edge based on the edge detected by the edge detection step. A reference point calculating step for obtaining a reference point in the cell, a cell feature calculating step for obtaining an area / shape of each cell as a cell feature from the edge detected in the edge detecting step, and the cell feature and the reference point A determination step for determining division / integration of each cell, a cell division step for dividing the large cell when there is a large cell determined to be divided from the area / shape, and integration from the area / shape A cell integration step of integrating the small cells when there is a small cell determined to be, and the number of cells after the cell division / integration processing Having a cell density calculation step of obtaining a Luo cell density.

  Further, in the corneal endothelial cell image processing program according to the embodiment of the present invention, the arithmetic control circuit determines whether or not the cell wall can be detected. After the integration process, the cell density is obtained from the number of cells.

  Further, the corneal endothelial cell image processing program according to the embodiment of the present invention includes an edge detection step for detecting an edge that is a cell wall of each cell of the corneal endothelial cell image, and each edge based on the edge detected by the edge detection step. A reference point calculating step for obtaining a reference point in the cell; a shape information forming step for forming predetermined peripheral shape information for including the reference point for image processing; and a periphery for expanding the peripheral shape information from the reference point. A shape expansion step; and a cell wall detection step of detecting the shape of the cell wall information as an edge from the peripheral shape information expanded by the peripheral shape expansion means and the cell wall information.

  In addition, the corneal endothelial cell image processing program according to the embodiment of the present invention is not in contact with the cell wall information in a direction in which a plurality of the peripheral shape information adjacent to each other and expanded by the peripheral shape expansion step are adjacent to each other. The plurality of pieces of peripheral shape information in contact with each other are determined to exist in one large area cell to be divided, and the plurality of pieces of peripheral shape information in contact with each other are included. A cell dividing step for dividing the area cell into a plurality of cells is provided.

  According to such a corneal endothelial cell image processing program, a more accurate cell density can be obtained by performing image processing as a pre-process before cell density detection.

1 is a block diagram of a corneal endothelial cell image processing apparatus according to the present invention. It is explanatory drawing of the image processing means shown in FIG. 2 is a schematic overall flowchart of an arithmetic control circuit shown in FIG. 3 is a flowchart for explaining integration processing by an arithmetic control circuit shown in FIG. 1. (A) is explanatory drawing which shows the example in case there exists a dotted noise in the whole corneal endothelial cell image imaged by the imaging means of FIG. 1, (b) is explanatory drawing of the corneal endothelial cell image of (a). is there. (A) is explanatory drawing which expanded and showed the corneal-endothelial-cell image partially, when there exists a point-like noise in the whole corneal-endothelial-cell image imaged with the imaging means of FIG. Moreover, (b) is explanatory drawing of the corneal endothelial cell image of (a). It is explanatory drawing of the corneal endothelial cell image imaged by the imaging means of FIG. It is explanatory drawing which shows an example of the image obtained by edge detection. (A)-(d) is explanatory drawing for a division | segmentation process. (A)-(c) is explanatory drawing of processes, such as the beard removal of the thinned cell. It is explanatory drawing for removing an unnecessary part from the image obtained by edge detection, and labeling. (A) is explanatory drawing which shows the relationship between the surrounding length and area of a cell, (b) is explanatory drawing which shows the relationship between the maximum length and width of a cell, (c) is explanatory drawing for an aspect ratio. It is explanatory drawing of the division | segmentation and integration of the cell of the labeled image. It is explanatory drawing which shows the other example of the division | segmentation of the cell of the labeled image. It is explanatory drawing of integration of a small area part. It is explanatory drawing of integration of the small area part of the number of contacts 1. It is explanatory drawing which shows the other example of integration of the small area part of the contact number of 1. It is explanatory drawing of integration of the small area part by circularity. It is explanatory drawing for calculating | requiring a cell density. It is a flowchart which shows the modification of the image processing of this invention. (A) is explanatory drawing which shows the corneal endothelial cell image whose cell wall is entirely clear, (b) is explanatory drawing of the corneal endothelial cell image of (a). (A) is explanatory drawing which expanded and showed the corneal endothelial cell image where the cell wall was entirely clear, (b) is explanatory drawing of the corneal endothelial cell image of (a).

Explanation of symbols

7 ... arithmetic control circuit 8 ... image processing means 8b ... edge detection means 8c ... reference point calculation means 8f ... cell feature calculation means 8j ... cell cell wall detection means 8k ... cell integration means 8L ... cell division means 8m ... cell density calculation means 8n ... shape information forming means 8o ... circumferential shape expanding means 10 ... cell (corneal endothelial cell image)
10a to 10d ... Small area part (small cell)
13 ... Reference point 14 ... Cell wall (edge)
R1-R3 ... Circle (circumferential shape information)

Claims (8)

Edge detection means for detecting an edge that is a cell wall of each cell of the cell image;
This method is based on the assumption of normals extending from the plurality of cell walls of each of the obtained cells to the inside of each cell based on the edges detected by the edge detection means. By determining the position of the average length of the line from the cell wall into the cell, the reference point that is the center of each cell is automatically determined from the position of the average length of the normal line from the cell wall into the cell. A reference point calculating means for calculating
Cell feature calculation means for obtaining the area / shape of each cell as a cell feature from the edge detected by the edge detection means;
A cell image processing apparatus comprising: an arithmetic control circuit that obtains the number of cells from the number of the reference points and obtains a cell density from the obtained number of cells and the cell feature . Edge detection means for detecting an edge that is a cell wall of each cell of the corneal endothelial cell image;
Reference point calculation means for obtaining a reference point in each cell based on the edge detected by the edge detection means;
Cell feature calculation means for obtaining the area / shape of each cell as a cell feature from the edge detected by the edge detection means;
A calculation control circuit for determining division / integration of each cell from the cell feature and the reference point;
A cell dividing means for dividing the large cells is controlled operated by the arithmetic control circuit when the area and shape of the cell is large cells that are determined to be divided larger than the cell of average size,
Cell integration means for integrating the small cells that are controlled by the arithmetic and control circuit when there is a small cell that is determined to be integrated smaller than an average size cell of the area and shape of the cell ;
The arithmetic control circuit, after dividing and integrating the cells, controls the operation of a cell density calculation means so as to obtain a cell density from the number of cells. apparatus.   The arithmetic control circuit determines whether or not the cell wall can be detected. If the cell wall can be detected, the cell division / integration processing is automatically performed, and then cell density calculation means is operated to control the cell. The corneal endothelial cell image processing apparatus according to claim 2, wherein the cell density is obtained from the number of the corneal endothelial cells. Edge detection means for detecting an edge that is a cell wall of each cell of the corneal endothelial cell image;
Reference point calculation means for obtaining a reference point in each cell based on the edge detected by the edge detection means;
Shape information forming means for forming predetermined peripheral shape information that includes the reference point for image processing;
Circumferential shape expansion means for expanding the circumferential shape information from the reference point;
A corneal endothelial cell image processing apparatus comprising: cell shape detection means for detecting the shape of the cell wall information as an edge from the circumference shape information expanded by the circumference shape expansion means and the cell wall information ,
When the plurality of circumferential shape information that are adjacent to each other and expanded by the circumferential shape inflating means do not contact the cell wall information in the adjacent direction and contact each other, the plurality of circumferential shape information that are in contact with each other should be divided Cell division means for dividing the one large area cell into a plurality of pieces so as to include each of the plurality of peripheral shape information in contact with each other, determined to exist in one large area cell, is provided. corneal endothelial cell image processing apparatus characterized by. An edge detection step for detecting an edge that is a cell wall of each cell of the cell image;
This method is based on the assumption of normals extending from the plurality of cell walls of the respective cells thus obtained toward the inside of each cell based on the edges detected by the edge detection step. By determining the position of the average length of the line from the cell wall into the cell, the reference point that is the center of each cell is automatically determined from the position of the average length of the normal line from the cell wall into the cell. A reference point calculating step to calculate
A cell feature calculation step for determining the area / shape of each cell as a cell feature from the edge detected in the edge detection step;
A cell image processing program comprising: a cell density calculation step of obtaining a cell density from the number of reference points and obtaining a cell density from the obtained number of cells and the cell feature . An edge detection step for detecting an edge that is a cell wall of each cell of the corneal endothelial cell image;
A reference point calculation step for obtaining a reference point in each cell based on the edge detected by the edge detection step;
A cell feature calculation step for determining the area / shape of each cell as a cell feature from the edge detected in the edge detection step;
A determination step of determining division / integration of each cell from the cell feature and the reference point;
A cell dividing step of dividing the large cell when there is a large cell whose area / shape is determined to be larger than an average size cell; and
A cell integration step of integrating the small cells when there is a small cell that is determined to be integrated with an area / shape of the cell smaller than an average size cell;
A corneal endothelial cell image processing program comprising a cell density calculation step of obtaining a cell density from the number of cells after the cell division / integration processing . It is determined whether or not the cell wall can be detected by an arithmetic control circuit, and if it can be detected, the cell division / integration processing is automatically performed, and then the cell density is obtained from the number of cells. The corneal endothelial cell image processing program according to claim 6 . An edge detection step for detecting an edge that is a cell wall of each cell of the corneal endothelial cell image;
A reference point calculation step for obtaining a reference point in each cell based on the edge detected by the edge detection step;
A shape information forming step for forming predetermined peripheral shape information that includes the reference point for image processing;
A circumferential shape expansion step for expanding the circumferential shape information from the reference point;
A cell wall detecting step of detecting the shape of the cell wall information as an edge from the peripheral shape information expanded by the peripheral shape expansion means and the cell wall information;
A corneal endothelial cell image processing program comprising:
When the plurality of circumferential shape information adjacent to each other and expanded by the circumferential shape expansion step contact each other without contacting the cell wall information in the adjacent direction, the plurality of circumferential shape information in contact with each other should be divided A cell dividing step is provided for dividing the one large area cell into a plurality of pieces so that the plurality of peripheral shape information in contact with each other are included, as determined to exist in one large area cell. corneal endothelial cell image processing program characterized by.