CN114693646B - Corneal endothelial cell active factor analysis method based on deep learning - Google Patents
Corneal endothelial cell active factor analysis method based on deep learning Download PDFInfo
- Publication number
- CN114693646B CN114693646B CN202210335690.0A CN202210335690A CN114693646B CN 114693646 B CN114693646 B CN 114693646B CN 202210335690 A CN202210335690 A CN 202210335690A CN 114693646 B CN114693646 B CN 114693646B
- Authority
- CN
- China
- Prior art keywords
- cell
- image
- cells
- corneal endothelial
- area
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0012—Biomedical image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/60—Analysis of geometric attributes
- G06T7/62—Analysis of geometric attributes of area, perimeter, diameter or volume
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
- G06T2207/30041—Eye; Retina; Ophthalmic
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Medical Informatics (AREA)
- Quality & Reliability (AREA)
- Radiology & Medical Imaging (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Geometry (AREA)
- Eye Examination Apparatus (AREA)
- Measuring Or Testing Involving Enzymes Or Micro-Organisms (AREA)
Abstract
The invention provides a corneal endothelial cell activity factor analysis method based on deep learning, which comprises the steps of segmenting a cell image to obtain a segmented image, obtaining a cell area image in the segmented image, counting the number and the distribution density of cells in the cell area image, calculating a general radial distribution function of the cell area image based on the number and the distribution density of the cells, obtaining a growth potential energy curve of the cells based on the general radial distribution function, obtaining a corneal endothelial cell activity factor based on the change rate of the growth potential energy curve, and obtaining the physiological state of the cells in the cell image based on the corneal endothelial cell activity factor. The analysis method disclosed by the disclosure can be used for obtaining the physiological state of the corneal endothelial cells by obtaining the potential energy conversion constant of the cell image, and in this case, the detection efficiency of the physiological state of the corneal endothelial cells can be improved.
Description
Technical Field
The disclosure relates to a biological information analysis and processing system, in particular to a corneal endothelial cell active factor analysis method based on deep learning.
Background
In a healthy physiological state, corneal endothelial cells are uniformly hexagonal and tightly embedded together, and the higher the proportion of hexagonal cells on an endothelial cell layer, the better the physiological state of corneal endothelial cells. When cells are damaged, corneal endothelial cells often appear in a non-hexagonal state or are unevenly distributed in an endothelial cell layer. And since corneal endothelial cells are non-regenerative, the number of hexagonal cells gradually decreases with age. Therefore, it is necessary to indicate the physiological or functional state of the corneal cells by some parameter indexes such as area cells and cell density.
In the prior art, parameter indexes of corneal endothelial cells are often analyzed during cell image research, and then the functional states of the corneal endothelial cells are obtained by analyzing the parameter indexes. However, it is not very effective to obtain the functional state of corneal endothelial cells by analyzing the parameter index, and thus a method capable of sensitively and effectively expressing the functional state of corneal endothelial cells is lacking.
Disclosure of Invention
The present disclosure has been made in view of the above circumstances, and an object thereof is to provide a method for analyzing a corneal endothelial cell activity factor based on deep learning, which obtains a corneal endothelial cell activity factor of a cell image and obtains a physiological state of a corneal endothelial cell from the corneal endothelial cell activity factor.
The cell image is segmented to obtain a segmented image, a cell area image is obtained in the segmented image, the cell area image is provided with a central cell positioned in the center of the image, the number and the distribution density of the cells in the cell area image are counted, a general radial distribution function of the cell area image is calculated based on the number and the distribution density of the cells, a growth potential energy curve of the cells is obtained based on the general radial distribution function, a corneal endothelial cell activity factor is obtained based on the change rate of the growth potential energy curve, and the physiological state of the cells in the cell image is obtained based on the corneal endothelial cell activity factor.
In the disclosure, compared with the method of directly analyzing the cell image to obtain the reference performance index of the cell and further obtain the physiological state of the corneal endothelial cell, the detection efficiency of the physiological state of the cell can be higher through the fitting of the potential energy distribution function.
Further, in the analysis method related to the present disclosure, optionally, the cell image is a cell image of a corneal endothelial cell, and the cell image is an 8-bit binary image. This enables edge information and edge features of the cell to be clearly identified.
In addition, in the analysis method according to the present disclosure, optionally, a growth potential distribution map of the cells is obtained based on the overall radial distribution function, and a fourth-order polynomial is used to perform curve fitting on the growth potential distribution map to obtain a growth potential curve. In this case, a fitted curve is obtained, and the change rate of the growth potential energy curve can be analyzed conveniently.
Further, in the analysis method related to the present disclosure, optionally, the cell region image includes a central cell and peripheral cells adjacent to the central cell.
In addition, in the analysis method according to the present disclosure, the cell image may be optionally subjected to edge extension processing, incomplete cells among the cells whose edges are identified in the cell image may be subjected to pixel replication, and the incomplete cells may be subjected to pixel replication. This can improve the accuracy of counting the number of cells, particularly the number of hexagonal cells.
In addition, in the analysis method according to the present disclosure, optionally, the segmented image is obtained by a segmentation model that segments the cell image by a U-net image segmentation technique to obtain a segmented image, and the segmentation model is trained using Soft-close as a loss function. Thus, a segmented image having a degree of identification can be obtained.
In addition, in the analysis method according to the present disclosure, optionally, a minimum value of the growth potential energy curve is obtained, and the corneal endothelial cell activity factor is obtained based on a second derivative of the growth potential energy curve at the minimum value.
In the analysis method according to the present disclosure, the physiological state of the cell in the cell image is optionally determined by a region in which the corneal endothelial cell activity factor is located, and when the corneal endothelial cell activity factor is located in the first region, the cell is considered to be in an abnormal state, and when the corneal endothelial cell activity factor is not located in the first region, the cell is considered to be in a normal state. In this case, it can be judged whether or not the cells are in a normal state by judging whether or not the corneal endothelial cell activity factor is in the first region.
In addition, in the analysis method related to the present disclosure, optionally, the physiological state of the cells in the cell image is obtained based on an evaluation parameter index of the corneal endothelial cells, the evaluation parameter index including at least one of cell density, hexagonal cell proportion, cell number, average cell area, cell area variation coefficient, maximum cell area, cell area standard deviation, and minimum cell area, and the first region is obtained based on the evaluation parameter index of the corneal endothelial cells and the corneal endothelial cell activity factor. Therefore, the physiological state of the corneal endothelial cells can be judged through the corneal endothelial cell active factor, and the judgment accuracy is improved.
In the analysis method according to the present disclosure, the central cell may be subjected to a swelling treatment, and whether or not the central cell is a hexagonal cell may be determined based on an overlapping region of the swelled region and another cell region; calculating the number of pixel points of the cell region image, obtaining the area of the cell region image according to the number of the pixel points and the size of the pixel points, and obtaining the average cell area based on the area of the cell region image and the number of cells in the cell region image; and evaluating the position of the central point of the cell through the geometrical central moment of the cell area image.
By the analysis method of the corneal endothelial cell activity factor based on deep learning, the corneal cell activity is converted into the corneal endothelial cell activity factor, and finally, the physiological state of the cornea is judged through the corneal endothelial cell activity factor, so that the detection efficiency of the physiological state of the corneal endothelial cells can be improved.
Drawings
The disclosure will now be explained in further detail by way of example only with reference to the accompanying drawings, in which:
fig. 1 is a schematic flow chart illustrating an analysis method of corneal endothelial cell activity factor based on deep learning according to an embodiment of the present disclosure.
Fig. 2 is a schematic diagram showing a cell image according to an embodiment of the present disclosure.
Fig. 3 is a schematic diagram illustrating a segmented image according to an embodiment of the present disclosure.
Fig. 4 is a schematic diagram showing a cell region image according to an embodiment of the present disclosure.
Fig. 5 is a diagram illustrating an overall radial distribution function of an image of a cell region according to an embodiment of the present disclosure.
Fig. 6 is a schematic diagram illustrating a growth potential curve of a cell according to embodiments of the present disclosure.
Fig. 7 is a fitting graph showing a growth potential curve according to an embodiment of the present disclosure.
Fig. 8 is a segmented picture showing corneal endothelial cells in a normal state according to the embodiment of the present disclosure.
Fig. 9A is a divided picture showing corneal endothelial cells in a first abnormal state according to the embodiment of the present disclosure.
Fig. 9B is a divided picture showing corneal endothelial cells in a second abnormal state according to the embodiment of the present disclosure.
Detailed Description
Hereinafter, preferred embodiments of the present disclosure will be described in detail with reference to the accompanying drawings. In the following description, the same components are denoted by the same reference numerals, and redundant description thereof is omitted. The drawings are schematic and the ratio of the dimensions of the components and the shapes of the components may be different from the actual ones.
It should be noted that the terms "first," "second," "third," and "fourth," etc. in the description and claims of the present disclosure and in the above-described drawings are used for distinguishing between different objects and not for describing a particular order. Further, the terms "comprises," "comprising," or any other variation thereof, such that a process, method, system, article, or apparatus that comprises or has a list of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include or have other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.
The disclosure relates to a corneal endothelial cell activity factor analysis method based on deep learning. The deep learning-based corneal endothelial cell activity factor analysis method to which the present disclosure relates may be referred to as an "activity factor calculation method" or an "analysis method".
Corneal endothelial cells, a non-regenerative cell of the human body, exhibit its important ring in visual function. And because the corneal endothelium can not be regrown, when the corneal endothelial cells are damaged, the activity of the cells can be reduced, and particularly, when the corneal endothelial cells are changed abnormally, the activity of the corneal endothelial cells can be obviously changed. The disclosure shows an analysis method of corneal endothelial cell activity factors based on deep learning, which expresses the activity of corneal cells by the activity factors, and finally judges the physiological state of the cornea through the activity factors.
Fig. 1 is a schematic flow chart illustrating an analysis method of corneal endothelial cell activity factor based on deep learning according to an embodiment of the present disclosure.
In some examples, referring to fig. 1, the analysis method may include: acquiring a segmented image of the cell image (step S110); acquiring a cell region image in the segmentation image (step S120); counting the number and distribution density of cells in the cell region image, and calculating an overall radial distribution function of the cell region image based on the number and distribution density of the cells (step S130); obtaining a growth potential curve of the cells based on the global radial distribution function (step S140); a corneal endothelial cell activity factor is obtained based on the rate of change of the growth potential curve (step S150).
The analysis method disclosed by the invention can be used for detecting the corneal endothelial cells, can be used for obtaining the active factors of the corneal endothelial cells in the cell image, and further can be used for finding out a proper interval to judge the physiological state of the corneal endothelial cells by analyzing the active factors of the corneal endothelial cells. This can assist in diagnosis of the physiological state of endothelial cells.
Fig. 2 is a schematic diagram showing a cell image according to an embodiment of the present disclosure.
In some examples, the cell image may be a cell image of corneal endothelial cells. The cell image can be obtained by observing corneal endothelial cells by a scanning electron microscope. In some examples, the cell image may be a full image. In other examples, the cell image may be obtained by cutting out a certain area in the full image.
In some examples, the analysis methods involved in the present disclosure may be implemented by software. Specifically, the software may sequentially perform steps S110 to S150, and in some examples, the software may have a plurality of controls, and the plurality of controls may respectively perform different steps. In some examples, the regions may be autonomously selected by an intercept control of the software. In this case, when the cells in some regions of the whole image are visually observed to have significant abnormalities, the regions can be directly selected for analysis, thereby improving the analysis efficiency.
In some examples, the cell image may be a binary image. In some examples, if a non-binary image is employed, too much unnecessary information may be introduced, which may in turn reduce the efficiency of image processing. In some examples, the image may be 8 bits. If the number of image bits is too high, the image accuracy will be too high, and the image processing efficiency will be reduced, whereas if the number of image bits is too low, the statistical accuracy of the relevant data will be reduced in the case of analyzing the image, for example, obtaining the number of cells and the distribution density.
In some examples, when an image of a cell is acquired, incomplete cells may be formed when the edge of the cell is at the edge of the image. In this case, the edge of the cell can be expanded to fill the cell to form a complete cell, so that no large error is introduced when counting the number or density of the complete cell. This can improve the accuracy of detecting the cell density. In some examples, the edges may be filled in such a way that pixel replication is performed at the edges of the cells.
Fig. 3 is a schematic diagram illustrating a segmented image according to an embodiment of the present disclosure.
In some examples, fig. 3 is a segmented image after the cell image (see fig. 2) has been segmented. In some examples, the image may be divided into several specific regions with unique properties by segmentation and the object of interest extracted, e.g., the cell edge of the corneal endothelial cells segmented. In some examples, the segmented image may be obtained by processing the cell image through a segmentation model, and segmenting the cell image using a U-net image segmentation technique to obtain the segmented image. In this case, an image of the corneal endothelial cell edge or cell contour can be effectively extracted. This can improve the efficiency of image analysis.
In some examples, the segmentation model is also often trained with a loss function after the segmentation process. In some examples, multiple data sets may be set to train the segmentation model.
In some examples, the segmentation model may be trained using Soft-clDice as a loss function. In some examples, the segmentation model trained by the Soft-clDice loss function has accurate connection information, higher graph similarity, better euler characteristics, better Dice coefficient, and higher accuracy. Thereby enabling accurate cell segmentation images to be obtained.
In some examples, the function of Soft-clDice is defined as follows:
In some examples, the segmented image may be obtained by simultaneously segmenting the cellular image through a U-net image segmentation technique and training the segmentation model using the Soft-clDice loss function.
Fig. 4 is a schematic diagram showing a cell region image according to an embodiment of the present disclosure.
In some examples, a cell region image as shown in fig. 4 can be obtained after a certain region of the cell image is cut and segmented by the cutting control.
In some examples, the cell region image may include a central cell and surrounding cells adjacent to the central cell. In some examples, the central cell may have a cell center point. In some examples, the location of the center point of the cell may be estimated by calculating the geometric moments of the cell region.
In some examples, after the capture control captures a region of the cell image, parameters of the image processing such as pixel size and pixel increment may be set. For example, setting the pixel size to 1 μm and the increment to 4px, two parameters are set to calculate the image, so as to obtain an energy elastic coefficient K (see fig. 4). In some examples, the energy elastic coefficient K also represents an active factor of corneal endothelial cells.
Fig. 5 is a diagram illustrating an overall radial distribution function of an image of a cell region according to an embodiment of the present disclosure.
In some examples, the overall radial distribution function of the cell may represent the activity of the division of the cell. In some examples, the overall radial distribution function of the cells may be obtained by analyzing the number and distribution density of nuclei within the micro-element region on the cell image in fig. 4 (refer to fig. 5). In some examples, the correlation between cell distribution and cell activity may be studied by a radial distribution function of the cells. Selecting a reference cell, and calculating a function formula of the overall radial distribution function of the cell image through regression analysis, wherein the function formula satisfies the following formula:
In the present disclosure, after obtaining the overall radial distribution function of the cells, the growth potential of the cells can be obtained by correlating the overall radial distribution function with the growth environment parameters of the cells. In some examples, the growth environment parameter of the cell is affected by temperature, and the activity of cell growth division is related to boltzmann's constant. Thus, the growth potential of the cell may satisfy the following formula:
V(r)＝-k B Tln[g(r)]wherein k is B Boltzmann constant, T is absolute temperature. In some examples, the absolute temperature may be considered constant. In this case, the potential for corneal cell growth can be determined by ln [ g (r) ]]And (4) mapping expression.
In some examples, a growth potential profile of the cell may be obtained by the overall radial distribution function. In some examples, the cell growth potential energy profile may be used to indicate the activity of cell growth. Thus, the activity of the cells can be roughly estimated by analyzing the profile of the growth potential.
Fig. 6 is a schematic diagram illustrating a growth potential curve of a cell according to embodiments of the present disclosure.
In some examples, a minimum of the growth potential curve may be obtained, and the corneal endothelial cell activity factor may be obtained based on a second derivative of the growth potential curve at the minimum. In some examples, growth potential energy conversion efficiency is highest when the cell potential energy is at the lowest point of potential energy conversion efficiency, and corneal endothelial cell activity factor is also highest at that time.
In some examples, the growth potential distribution curve may be derived to obtain a first derivative, obtain a minimum of minimum values when the first derivative is 0, solve the second derivative, and calculate a constant of the conversion efficiency of the growth potential at the minimum when the growth potential is at the lowest point of the corneal endothelial activity factor and the conversion efficiency of the growth potential at the lowest point is the highest (see fig. 6). In some examples, the high conversion efficiency of the growth potential energy may also indicate that the growth state of the corneal cells is in an excellent state. When the corneal endothelial cell activity factor is at the highest value, it can indicate that the cornea is in a healthy physiological state.
In some examples, corneal endothelial cell activity factors may also be used to indicate the energy conversion efficiency or the division and growth efficiency of corneal cells. In some examples, when the corneal endothelial cell activity factor is located in different regions, it may indicate that the cells in the cell image are in different physiological states.
In some examples, a cell may be considered abnormal when the corneal endothelial cell activity factor is in the first region and normal when the corneal endothelial cell activity factor is in the second region. In some examples, there may be different second regions under different conditions. For example, the second region used to indicate the normal state of the cell may be different at different ages. In this case, the physiological state of the corneal endothelial cells can be judged by analyzing the image to determine whether or not the corneal endothelial cell activity factor is in a normal range under the same conditions.
In some examples, the first region may correspond to a different physiological abnormal region. In some examples, when the corneal endothelial cell activity factor is in different regions, it may indicate that the cornea is in a different condition.
In some examples, the corneal endothelial cells may also have different corneal endothelial cell activity factors when the patient is at different stages of the same condition.
Fig. 7 is a fitting graph showing a growth potential curve according to an embodiment of the present disclosure. In some examples, the potential energy distribution on the profile of the growth potential energy is a dispersion curve, in which case fitting the dispersion curve can facilitate observing the trend of the curve of the profile. In some examples, a discrete curve of the growth potential distribution map may be fitted by a fourth order polynomial, and then a potential distribution graph is obtained. In fig. 5, r between 12 and 24 is selected, and a local fitting graph of the potential energy distribution curve shown in fig. 7 can be obtained through software analysis and fitting.
In some examples, the physiological state of the cell image may be obtained by evaluating a parameter. In some examples, the evaluation parameter may be at least one selected from the group consisting of cell density, hexagonal cell proportion, cell number, average cell area, cell area coefficient of variation, maximum cell area, cell area standard deviation, and minimum cell area. In some examples, the first region may be obtained based on an evaluation parameter index of corneal endothelial cells and a corneal endothelial cell activity factor.
In some examples, the proportion of hexagonal cells may be obtained by counting the number of hexagonal cells. In some examples, the proportion of hexagonal cells may be the proportion of hexagonal cells in the total cell number. In some examples, under ideal physiological conditions, the proportion of hexagonal cells is greater than 50%.
In some examples, the central cell may be dilated, and the dilated region may coincide with the surrounding cell region. In this case, whether or not the central cell is a hexagonal cell can be determined by the number of cells around the central cell in the overlapping region. In some examples, a region of each cell may be subjected to a swelling treatment of 8*8. After the expansion process, the boundary of the obtained cell region image is relatively smooth. This can improve the noise immunity of the image processing.
In some examples, when the number of surrounding cells present in the overlapping region is 6, then the central cell may be determined to be a hexagonal cell. Since normal corneal endothelial cells are hexagonal cells, when the number of cells existing in the overlapping region is other than 6, it can be determined that the central cell is an abnormal cell. From this, the number of hexagonal cells can be calculated. In other words, the cell ratio of hexagonal cells can be obtained.
In some examples, the number of pixels of the cell region image may be calculated, the area of the cell region image may be obtained by the number of pixels and the pixel size, and the average cell area may be obtained based on the area of the cell region image and the number of cells in the cell region image. Specifically, the number of pixels in each cell region may be calculated, the area of the cell may be obtained by multiplying the number of pixels by the square of the actual size of each pixel, and the average cell area may be obtained by dividing the accumulated cell area of each cell region by the number of cells.
In some examples, the maximum value of the cell area and the minimum value of the cell area may be obtained by the cell area of each cell region. In some examples, the average cell area may also be referred to as an average of cell areas. In this case, the variance of the cell area and thus the standard deviation of the cell area can be obtained based on the maximum value, the minimum value, and the average value. In some examples, the standard deviation of the cell area is greater than 140 under ideal physiological conditions.
In some examples, the variance of the cell area and the cell density may be multiplied to obtain the area coefficient of variation of the cell. In some examples, the coefficient of cell area variation is greater than 30 under ideal physiological conditions.
In the following, the corneal endothelial cell activity factor is described by a lesion image of corneal endothelial cells.
Fig. 8 is a segmented picture showing corneal endothelial cells in a normal state according to the embodiment of the present disclosure. Fig. 9A is a divided picture showing corneal endothelial cells in a first abnormal state according to the embodiment of the present disclosure. Fig. 9B is a divided picture showing a corneal endothelial cell according to the embodiment of the present disclosure in a second abnormal state.
In the present embodiment, the corneal endothelial cell activity factor represented in fig. 8 is 0.2575 as calculated by the analysis method, and the corneal endothelial cell is in a normal physiological state at this time.
In some examples, the pictures shown in fig. 9A and 9B may be obtained by segmenting some cell images.
As shown in fig. 9A, the corneal endothelial cells showed different sizes in the image, and the active factor of the corneal endothelial cells calculated by the above analysis method was 0.06478. The value of the factor is lower than that of the active factor of normal corneal endothelial cells, so that the inquiry can indicate that the corneal endothelium is in an early pathological stage or the corneal endothelium suffers from viral corneal endopdermatitis. In this case, the pathological state of corneal endothelial cells can be effectively judged by analyzing the active factors, and thus the cornea can be specifically treated according to the disease.
As shown in FIG. 9B, the area of individual corneal endothelial cells was larger, and the number of corneal endothelial cells was smaller overall. Selecting the picture and inserting the picture into analysis software, intercepting the whole picture, setting the pixel size and the pixel increment, and finally obtaining the active factor 0.08165 of the corneal endothelial cells through analysis. By comparing with the normal data under the same conditions (for example, age, sex, etc.), it can be found that the numerical value is also in a lower state. The query analysis of the corneal endothelial cells can indicate that the corneal endothelial cells are in operation, such as after cataract operation and glaucoma-related eye disease operation.
The corneal endothelial cell activity factor analysis method based on deep learning can effectively obtain the physiological state of the corneal endothelial cells, and further can improve the detection efficiency of the physiological state of the corneal endothelial cells.
While the present disclosure has been described in detail above with reference to the drawings and examples, it should be understood that the above description is not intended to limit the disclosure in any way. Those skilled in the art can make modifications and variations to the present disclosure as needed without departing from the true spirit and scope of the disclosure, which fall within the scope of the disclosure.
Claims (8)
1. A corneal endothelial cell activity factor analysis method based on deep learning is characterized in that,
segmenting a cell image to obtain a segmented image, obtaining a cell region image in the segmented image, wherein the cell region image is provided with a central cell positioned in the center of the image, counting the number and the distribution density of the cells in the cell region image, calculating a general radial distribution function of the cell region image based on the number and the distribution density of the cells, obtaining a growth potential energy distribution diagram of the cells based on the general radial distribution function, carrying out curve fitting on the growth potential energy distribution diagram by adopting a fourth-order polynomial to obtain a growth potential energy curve of the cells, obtaining a minimum value of the growth potential energy curve based on the change rate of the growth potential energy curve, obtaining a corneal endothelial cell activity factor based on a second derivative of the growth potential energy curve at the minimum value, and obtaining a physiological state of the cells in the cell image based on the corneal endothelial cell activity factor;
wherein the step of calculating an overall radial distribution function of the cell region image based on the cell count and the distribution density comprises: selecting a reference cell, and calculating a functional expression of a total radial distribution function of the cell region image through regression analysis, wherein the functional expression of the total radial distribution function satisfies the formula:
wherein g (r) represents the overall radial distribution function, r represents the distance between the edge of the selected region after the reference cell is given and the center of the reference cell, n i (r) represents the number of cells contained within different r, ρ represents the average cell density over the image of the cell region, and dr represents the change in distance between the center of the reference cell and the edge of the selected region given the reference cell.
2. The analytical method of claim 1,
the cell image is a cell image of corneal endothelial cells, and the cell image is an 8-bit binary image.
3. The analytical method of claim 1,
the cell region image includes a central cell and surrounding cells adjoining the central cell.
4. The analytical method of claim 1,
and performing edge expansion processing on the cell image, identifying incomplete cells in cells at the edge of the cell image, and performing pixel replication on the incomplete cells.
5. The analytical method of claim 1,
and obtaining a segmentation image through a segmentation model, segmenting the cell image through a U-net image segmentation technology to obtain a segmentation image, and training the segmentation model by using Soft-clDice as a loss function.
6. The analytical method of claim 2,
and judging the physiological state of the cells in the cell image through the area where the corneal endothelial cell activity factor is located, and when the corneal endothelial cell activity factor is located in the first area, considering the cells to be in an abnormal state, and when the corneal endothelial cell activity factor is not located in the first area, considering the cells to be in a normal state.
7. The analytical method of claim 6,
obtaining a physiological state of cells in the cell image based on an evaluation parameter index of the corneal endothelial cells, the evaluation parameter index including at least one of cell density, hexagonal cell proportion, cell number, average cell area, cell area variation coefficient, maximum cell area, cell area standard deviation, and minimum cell area, the first region being obtained based on the evaluation parameter index of the corneal endothelial cells and the corneal endothelial cell activity factor.
8. The assay of claim 7,
performing expansion treatment on the central cell, and judging whether the central cell is a hexagonal cell or not based on the overlapped area of the expanded area and other cell areas; calculating the number of pixel points of the cell region image, obtaining the area of the cell region image according to the number of the pixel points and the size of the pixel points, and obtaining the average cell area based on the area of the cell region image and the number of cells in the cell region image; and evaluating the position of the central point of the cell through the geometrical central moment of the cell area image.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210335690.0A CN114693646B (en) | 2022-03-31 | 2022-03-31 | Corneal endothelial cell active factor analysis method based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210335690.0A CN114693646B (en) | 2022-03-31 | 2022-03-31 | Corneal endothelial cell active factor analysis method based on deep learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114693646A CN114693646A (en) | 2022-07-01 |
| CN114693646B true CN114693646B (en) | 2023-04-11 |
Family
ID=82141111
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210335690.0A Active CN114693646B (en) | 2022-03-31 | 2022-03-31 | Corneal endothelial cell active factor analysis method based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114693646B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116091451A (en) * | 2023-01-09 | 2023-05-09 | 中国科学院苏州生物医学工程技术研究所 | Retinal pigment epithelial cell image segmentation method and system |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1860363A (en) * | 2003-08-28 | 2006-11-08 | 赛路拉公司 | Methods and apparatus for sorting cells using an optical switch in a microfluidic channel network |
| CN104603291A (en) * | 2012-06-22 | 2015-05-06 | Htg分子诊断有限公司 | Molecular malignancy in melanocytic lesions |
Family Cites Families (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103955941B (en) * | 2014-05-21 | 2016-08-17 | 重庆理工大学 | A kind of specularmicroscope and the method for analysis |
| JP6345332B1 (en) * | 2017-11-21 | 2018-06-20 | 国立研究開発法人理化学研究所 | Classification device, classification method, program, and information recording medium |
| JP2022506154A (en) * | 2018-11-01 | 2022-01-17 | バークレー ライツ，インコーポレイテッド | Methods for Assaying Living Cells in a Microfluidic Environment |
| CN112215790A (en) * | 2019-06-24 | 2021-01-12 | 杭州迪英加科技有限公司 | KI67 index analysis method based on deep learning |
| US20210278655A1 (en) * | 2020-03-05 | 2021-09-09 | Case Western Reserve University | Automated segmentation and guided correction of endothelial cell images |
| CN113592783B (en) * | 2021-07-08 | 2024-01-19 | 北京大学第三医院（北京大学第三临床医学院） | Accurate quantification method and device for cell basic index in cornea confocal image |
-
2022
- 2022-03-31 CN CN202210335690.0A patent/CN114693646B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1860363A (en) * | 2003-08-28 | 2006-11-08 | 赛路拉公司 | Methods and apparatus for sorting cells using an optical switch in a microfluidic channel network |
| CN104603291A (en) * | 2012-06-22 | 2015-05-06 | Htg分子诊断有限公司 | Molecular malignancy in melanocytic lesions |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114693646A (en) | 2022-07-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109325942B (en) | Fundus image structure segmentation method based on full convolution neural network | |
| CN111985536B (en) | Based on weak supervised learning gastroscopic pathology image Classification method | |
| US9684959B2 (en) | Methods and systems for automatic location of optic structures in an image of an eye, and for automatic retina cup-to-disc ratio computation | |
| Kande et al. | Segmentation of exudates and optic disk in retinal images | |
| Harini et al. | Feature extraction and classification of retinal images for automated detection of Diabetic Retinopathy | |
| CN110555845A (en) | Fundus OCT image identification method and equipment | |
| CN109961848B (en) | Macular image classification method and device | |
| CN110751172B (en) | Method and system for deducing category of pathological full-film image for weak supervision learning | |
| CN105069803A (en) | Classifier for micro-angioma of diabetes lesion based on colored image | |
| CN114693646B (en) | Corneal endothelial cell active factor analysis method based on deep learning | |
| Aminah et al. | Diabetes prediction system based on iridology using machine learning | |
| CN117409004B (en) | Lung rehabilitation intelligent auxiliary system based on medical image analysis | |
| Hijazi et al. | Retinal image classification for the screening of age-related macular degeneration | |
| CN110660070A (en) | Rice vein image extraction method and device | |
| CN114240874A (en) | Bone age assessment method and device based on deep convolutional neural network and feature fusion and computer readable storage medium | |
| CN115731203A (en) | Cataract image identification method and device, computer equipment and readable storage medium | |
| CN114972272A (en) | Grad-CAM-based segmentation method for new coronary pneumonia lesions | |
| CN111028230A (en) | Fundus image optic disc and macula lutea positioning detection algorithm based on YOLO-V3 | |
| Devasia et al. | Automatic Optic Disc Localization and Segmentation using Swarm Intelligence. | |
| KR101779642B1 (en) | Method of comparing images of irises by intelligent selection of textured zones | |
| Habashy | Identification of diabetic retinopathy stages using fuzzy C-means classifier | |
| Balakrishnan et al. | A hybrid PSO-DEFS based feature selection for the identification of diabetic retinopathy | |
| Alves et al. | The challenges of applying deep learning for hemangioma lesion segmentation | |
| CN115170492A (en) | Intelligent prediction and evaluation system for postoperative vision of cataract patient based on AI (artificial intelligence) technology | |
| Santos et al. | A semiautomatic superpixel based approach to cup-to-disc ratio measurement |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |