Cell nucleus instance segmentation method based on multi-scale feature fusion
Through the combination of multi-scale feature fusion and segmentation network, the problem of complex shape and noise influence of cell nuclei in pathological images is solved, and high-accurate nucleus instance segmentation is achieved, enhancing the network's characterization ability.
Patent Information
- Application Number
- CN202411282433.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-13
- Publication Date
- 2024-12-03
- Estimated Expiration
- 2044-09-13
AI Technical Summary
The complex shape, size and overlap of cell nuclei in pathological images, as well as noise and artifacts, have made it more difficult to segment cell nucleus instances, and it is difficult for the prior art to achieve accurate segmentation.
Using the cell nucleus instance segmentation method based on multi-scale feature fusion, the feature extraction module, mask optimization algorithm module and post-processing algorithm module are combined with the U-Net model and other algorithms, and the point, line and surface features are used to achieve the accuracy of the cell nucleus. Positioning and shape descriptions to generate the final instance segmentation diagram.
It improves the accuracy of cell nucleus instance segmentation, can effectively process complex shapes and overlapping cell nuclei, reduces the impact of noise, generates clear segmented images, and enhances the network's characterization ability.
Smart Images

Figure CN119067994A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of medical image processing, and in particular to a cell nucleus instance segmentation method based on multi-scale feature fusion. Background Art
[0002] In pathology, the morphology and distribution characteristics of cell nuclei are of great significance for the diagnosis and treatment of diseases. By accurately segmenting the cell nuclei, doctors can perform more precise pathological analysis and identify abnormal morphology and structure of cells, thereby helping to determine the type, grade, and prognosis of the disease. In addition, cell nucleus instance segmentation can also be used to evaluate the effects of drugs on cells and tissues, as well as to monitor the effectiveness of treatment. Due to the complexity and diversity of cell nuclei in pathological images, cell nucleus instance segmentation is a challenging task. Cell nuclei usually have irregular shapes, different sizes and colors, and may overlap and touch. In addition, pathological images are often accompanied by noise, artifacts, and uneven illumination, which further increases the difficulty of the segmentation task. Summary of the invention
[0003] In order to overcome the deficiencies of the above technologies, the present invention provides a cell nucleus instance segmentation method based on multi-scale feature fusion, which improves the accuracy of cell nucleus instance segmentation.
[0004] The technical solution adopted by the present invention to overcome the technical problems is:
[0005] A cell nucleus instance segmentation method based on multi-scale feature fusion, comprising:
[0006] a) Obtain n cell nucleus images and their corresponding labels to obtain a cell nucleus image set Y and a label set G, where Y = {Y1, Y2, ..., Y i ,...,Y n},Y i is the i-th cell nucleus image, i∈{1,...,n}, G={G1,G2,...,G i ,...,G n}, G i is the i-th cell nucleus image Y i The corresponding label;
[0007] b) Divide the cell nucleus image set Y into a training set P and a test set T, where P = {P1, P2, ..., P i ,...,P o}, P i is the i-th cell nucleus image in the training set P, i∈{1,...,o}, o is the total number of cell nucleus images in the training set P, T={T1,T2,…,T i ,…,T q}, T iis the i-th cell nucleus image in the test set T, i∈{1,...,q}, q is the total number of cell nucleus images in the test set T;
[0008] c) Take the i-th cell nucleus image P in the training set P i The corresponding label G i Convert to nucleus edge outline label and heatmap labels All the cell nucleus edge contour labels in the training set P constitute the cell nucleus edge contour label set All heat map labels in the training set P constitute the heat map label set
[0009] d) Establishing a segmentation network, which consists of a feature extraction module, a mask optimization algorithm module, and a post-processing algorithm module;
[0010] e) Take the i-th cell nucleus image P in the training set P i Input into the feature extraction module of the segmentation network and output the feature map Feature Map Feature Map
[0011] f) Using feature maps Get the key point map of the cell nucleus
[0012] g) The feature map Feature Map Key points of the cell nucleus Input to the mask optimization algorithm module of the segmentation network and output the feature map
[0013] h) The feature map With feature map Input into the post-processing algorithm module of the segmentation network, and output the final cell nucleus instance segmentation map i) Use the SGD optimizer to take advantage of the total loss function L total Train the segmentation network to obtain the optimized segmentation network; j) Take the i-th cell nucleus image P in the test set T i Input into the optimized segmentation network, and output the cell nucleus instance segmentation map Furthermore, in step a), n cell nucleus images and their corresponding labels are obtained from the MoNuSeg dataset. Furthermore, in step b), the cell nucleus image set Y is divided into a training set P and a test set T in a ratio of 7:3.
[0014] Further, step c) comprises the following steps:
[0015] c-1) Use cv2.boundingRect() function to obtain the i-th cell nucleus image P in the training set P i The corresponding label G i The boundary rectangle information includes the center point coordinates, the height of the boundary rectangle, and the width of the boundary rectangle;
[0016] c-2) Input the height and width of the bounding rectangle into the gaussian_radius() function to calculate the radius of the Gaussian kernel;
[0017] c-3) The diameter Dia of the Gaussian kernel is calculated by Dia=2*(radius+1);
[0018] c-4) Input the diameter Dia into the gussian2D() function of the astropy library to obtain the Gaussian kernel Gk; c-5) Place the Gaussian kernel Gk at the center point coordinates of the bounding rectangle to obtain the heat map label c-6) Take the i-th cell nucleus image P in the training set P i The corresponding label G i Convert to numpy array in numpy format through np.array() function;
[0019] c-7) Use cv2.Canny() to perform edge detection on the numpy array and convert it into an edge detection image of uint8 type through the np.uint8() function;
[0020] c-8) Set the values of non-zero pixels in the edge detection image to 255 and the values of the remaining pixels to 0 to obtain the cell nucleus edge contour label Further, step e) comprises the following steps:
[0021] e-1) The feature extraction module of the segmentation network consists of a U-Net model, a first two-dimensional deconvolution layer, a second two-dimensional deconvolution layer, and a third two-dimensional deconvolution layer. The i-th cell nucleus image P in the training set P is i Input into the U-Net model of the feature extraction module, and output the feature map M i ;
[0022] e-2) The feature map M i Input to the first 2D deconvolution layer of the feature extraction module, and output the feature map
[0023] e-3) The feature map M i Input to the second 2D deconvolution layer of the feature extraction module, and output the feature map
[0024] e-4) The feature map M i Input to the third 2D deconvolution layer of the feature extraction module, and output the feature map
[0025] Furthermore, in step f), the peak_local_max() function in the skimage package is used to find the feature map Local peak value, get the key point map of the cell nucleus Further, step g) comprises the following steps:
[0026] g-1) The feature map Input into the mask optimization algorithm module of the segmentation network, and use the Image.new() function to initialize a feature map Blank feature maps of the same size g-2) The key point map of the cell nucleus Input into the mask optimization algorithm module of the segmentation network, and obtain the key point map of the cell nucleus through the np.column_stack(np.where()) function The key points of the n nuclei are obtained, and the key point set K is obtained, K = {k1, k2, ..., k i ,…,k n}, k i is the i-th cell nucleus key point, i∈{1,…,n};
[0027] g-3) The feature map Input into the mask optimization algorithm module of the segmentation network, and obtain the feature map through the cv2.findContours() function The m cell nucleus boundaries in the , we get the cell nucleus boundary set E, E = {e1, e2, …, e i ,...,e n},e i is the boundary of the ith cell nucleus;
[0028] g-4) Use sklearn.neighbors.NearestNeighbors() function to find the i-th cell nucleus key point k i Perform the nearest neighbor search, and when the corresponding boundary i-th cell nucleus boundary e is found i Stop searching when the blank feature map The corresponding position in is marked as 1, and the blank feature map After all corresponding positions are marked, the feature map is obtained Further, step h) comprises the following steps:
[0029] h-1) The feature map With feature map Input to the post-processing algorithm module of the segmentation network for subtraction operation to obtain the difference map h-2) Use cv2.connectedComponentsWithStats() function to compare the difference graph Perform connected component analysis and output the set C of each connected region, C = {c1, c2, ..., c i ,...,c n},c i Difference In the jth small area, if the jth small area c i If the jth small area c is smaller than 5 pixels, it will be deleted. i If it is greater than or equal to 5 pixels, it will be retained to obtain the feature map h-3) Use cv2.dilate() function to dilate the feature map Perform the expansion operation to obtain the final cell nucleus instance segmentation map The convolution kernel during the expansion operation is kernel, Furthermore, in step i), by formula L total =L p +L edge +L m Calculate the total loss function L total , where L p is the mean square error loss function, the heat map label is the mean square error loss function L p Label map, feature map is the mean square error loss function L p Original image, L edge is the binary cross entropy loss function, feature map is the binary cross entropy loss function L edge Original image, cell nucleus edge contour label is the binary cross entropy loss function L edge Label map of L m is the cross entropy loss function, feature map is the cross entropy loss function L m The original image, the i-th cell nucleus image Y i The corresponding label G i is the cross entropy loss function L m Label map of .
[0030] The beneficial effects of the present invention are as follows: point features can accurately locate the position of the cell nucleus, line features can describe the shape and outline of the cell nucleus, and surface features provide an overall description of the cell nucleus area. The features of points, lines and surfaces are comprehensively utilized to achieve a comprehensive description of the cell nucleus and increase the representation ability of the network. First, a mask optimization algorithm module is proposed to effectively merge point features with boundary features. Secondly, we introduce a post-processing method that combines target information and cell nucleus mask images to generate an instance segmentation image. The effectiveness of each component in the proposed cell nucleus instance segmentation method based on multi-scale feature fusion. BRIEF DESCRIPTION OF THE DRAWINGS
[0031] Figure 1 is a structural diagram of a segmentation network of the present invention;
[0032] Figure 2 The segmentation results of the MoNuSeg dataset, Kumar dataset, and CoNSep dataset are shown below. DETAILED DESCRIPTION
[0033] The following is combined with Figure 1 , Attachment Figure 2 The present invention is further described.
[0034] A cell nucleus instance segmentation method based on multi-scale feature fusion, comprising:
[0035] a) Obtain n cell nucleus images and their corresponding labels to obtain a cell nucleus image set Y and a label set G, where Y = {Y1, Y2, ..., Y i ,...,Y n},Y i is the i-th cell nucleus image, i∈{1,...,n}, G={G1,G2,...,G i ,...,G n}, G i is the i-th cell nucleus image Y i The corresponding label.
[0036] b) Divide the cell nucleus image set Y into a training set P and a test set T, where P = {P1, P2, ..., P i ,...,P o}, P i is the i-th cell nucleus image in the training set P, i∈{1,...,o}, o is the total number of cell nucleus images in the training set P, T={T1,T2,...,T i ,...,T q}, T i is the i-th cell nucleus image in the test set T, i∈{1,…,q}, q is the total number of cell nucleus images in the test set T.
[0037] c) Take the i-th cell nucleus image P in the training set P i The corresponding label G i Convert to nucleus edge outline label and heatmap labels All the cell nucleus edge contour labels in the training set P constitute the cell nucleus edge contour label set All heat map labels in the training set P constitute the heat map label set
[0038] d) Establish a segmentation network, which consists of a feature extraction module, a mask optimization algorithm module, and a post-processing algorithm module.
[0039] e) Take the i-th cell nucleus image P in the training set P i Input into the feature extraction module of the segmentation network and output the feature map Feature Map Feature Map
[0040] f) Using feature maps Get the key point map of the cell nucleus
[0041] g) The feature map Feature Map Key points of the cell nucleus Input to the mask optimization algorithm module of the segmentation network and output the feature map
[0042] h) The feature map With feature map Input into the post-processing algorithm module of the segmentation network, and output the final cell nucleus instance segmentation map i) Use the SGD optimizer to take advantage of the total loss function L total Train the segmentation network to obtain the optimized segmentation network. j) Take the i-th cell nucleus image P in the test set T i Input into the optimized segmentation network, and output the cell nucleus instance segmentation map In one embodiment of the present invention, in step a), n cell nucleus images and their corresponding labels are obtained from the MoNuSeg dataset.
[0043] In one embodiment of the present invention, in step b), the cell nucleus image set Y is divided into a training set P and a test set T in a ratio of 7:3.
[0044] In one embodiment of the present invention, step c) comprises the following steps:
[0045] c-1) Use cv2.boundingRect() function to obtain the i-th cell nucleus image P in the training set P i The corresponding label G i The boundary rectangle information includes the center point coordinates, the height of the boundary rectangle, and the width of the boundary rectangle.
[0046] c-2) Input the height and width of the bounding rectangle into the gaussian_radius() function to calculate the radius of the Gaussian kernel.
[0047] c-3) The diameter Dia of the Gaussian kernel is calculated by Dia=2*(radius+1).
[0048] c-4) Input the diameter Dia into the gussian2D() function of the astropy library to obtain the Gaussian kernel Gk. c-5) Place the Gaussian kernel Gk at the center point coordinates of the bounding rectangle to obtain the heat map label c-6) Take the i-th cell nucleus image P in the training set P i The corresponding label G i Convert the array into numpy format using the np.array() function.
[0049] c-7) Use cv2.Canny() to perform edge detection on the numpy array and convert it into an edge detected image of uint8 type using the np.uint8() function.
[0050] c-8) Set the values of non-zero pixels in the edge detection image to 255 and the values of the remaining pixels to 0 to obtain the cell nucleus edge contour label In one embodiment of the present invention, step e) comprises the following steps:
[0051] e-1) The feature extraction module of the segmentation network consists of a U-Net model, a first two-dimensional deconvolution layer, a second two-dimensional deconvolution layer, and a third two-dimensional deconvolution layer. The i-th cell nucleus image P in the training set P is i Input into the U-Net model of the feature extraction module, and output the feature map M i .
[0052] e-2) The feature map M i Input to the first 2D deconvolution layer of the feature extraction module, and output the feature map
[0053] e-3) The feature map M i Input to the second 2D deconvolution layer of the feature extraction module, and output the feature map
[0054] e-4) The feature map M i Input to the third 2D deconvolution layer of the feature extraction module, and output the feature map
[0055] In one embodiment of the present invention, in step f), the peak_local_max() function in the skimage package is used to find the feature map. Local peak value, get the key point map of the cell nucleus In one embodiment of the present invention, step g) comprises the following steps:
[0056] g-1) The feature map Input into the mask optimization algorithm module of the segmentation network, and use the Image.new() function to initialize a feature map Blank feature maps of the same size g-2) The key point map of the cell nucleus Input into the mask optimization algorithm module of the segmentation network, and obtain the key point map of the cell nucleus through the np.column_stack(np.where()) function The key points of the n nuclei are obtained, and the key point set K is obtained, K = {k1, k2, ..., k i ,…,k n}, k i is the i-th cell nucleus key point, i∈{1,...,n}.
[0057] g-3) The feature map Input into the mask optimization algorithm module of the segmentation network, and obtain the feature map through the cv2.findContours() function The m cell nucleus boundaries in the , we get the cell nucleus boundary set E, E = {e1, e2, …, e i ,...,e n},e i is the boundary of the ith cell nucleus.
[0058] g-4) Use sklearn.neighbors.NearestNeighbors() function to find the i-th cell nucleus key point k i Perform the nearest neighbor search, and when the corresponding boundary i-th cell nucleus boundary e is found i Stop searching when the blank feature map The corresponding position in is marked as 1, and the blank feature map After all corresponding positions are marked, the feature map is obtained In one embodiment of the present invention, step h) comprises the following steps:
[0059] h-1) The feature map With feature map Input to the post-processing algorithm module of the segmentation network for subtraction operation to obtain the difference map h-2) Use cv2.connectedComponentsWithStats() function to compare the difference graph Perform connected component analysis and output the set C of each connected region, C = {c1, c2, ..., c i ,...,c n},c i Difference In the jth small area, if the jth small area c i If the jth small area c is smaller than 5 pixels, it will be deleted. i If it is greater than or equal to 5 pixels, it will be retained to obtain the feature map h-3) Use cv2.dilate() function to dilate the feature map Perform the expansion operation to obtain the final cell nucleus instance segmentation map The convolution kernel during the expansion operation is kernel, In one embodiment of the present invention, in step i), the formula L total =L p +L edge +L m Calculate the total loss function L total , where L p is the mean square error loss function, the heat map label is the mean square error loss function L p Label map, feature map is the mean square error loss function L p Original image, L edge is the binary cross entropy loss function, feature map is the binary cross entropy loss function L edge Original image, cell nucleus edge contour label is the binary cross entropy loss function L edge Label map of L m is the cross entropy loss function, feature map is the cross entropy loss function L m The original image, the i-th cell nucleus image Y i The corresponding label G i is the cross entropy loss function L m Label map of .
[0060] We implemented these models using PyTorch 3.7 on an NVIDIA Tesla V100 system. During training, we kept the original dimensions of all datasets and applied data augmentation techniques such as random horizontal and vertical flips and random rotations. We used stochastic gradient descent (SGD) for optimization with a learning rate of 0.00001 and an adaptive learning rate decay strategy.
[0061] For the MoNuSeg dataset, the input size is 1000×1000 and the batch size is 12; for the Kumar dataset, the input size is 1000×1000 and the batch size is 8; for the CoNSep dataset, the input size is 512×512 and the batch size is 12. We save a checkpoint every 20 epochs and train for a total of 200 epochs. In addition, we use a specific method to solve the discontinuity problem in the nucleus segmentation in the post-processing stage.
[0062] The proposed method is qualitatively analyzed as shown in the attached Figure 2 As shown in the figure, from top to bottom, they are cropped from the test images of the MoNuSeg, Kumar, and CoNSep datasets. The original images and the corresponding segmentation kernels are listed in rows. Figure 2 It can be seen that the proposed model can successfully segment overlapping cell nuclei, which has obvious advantages over the other three models.
[0063] Table 1 Comparison results between MoNuSeg dataset and SOTAs model
[0064] Table 1 Comparison results between MoNuSeg dataset and SOTAs model
[0065]
[0066] For quantitative analysis, Table 1 shows that our model achieves good performance in the task of cell nucleus segmentation. Specifically, it achieves 0.83711Dice and 0.7213AJI on the MoNuSeg dataset, which is significantly better than Hover-Net and CIA-Net.
[0067] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or replace some of the technical features therein by equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.
Claims
1. A cell nucleus instance segmentation method based on multi-scale feature fusion, characterized in that: include: a) Obtain n cell nucleus images and their corresponding labels to obtain a cell nucleus image set Y and a label set G, where Y = {Y1, Y2, ..., Y i ,...,Y n },Y i is the i-th cell nucleus image, i∈{1,...,n}, G={G1,G2,...,G i ,...,G n }, G i is the i-th cell nucleus image Y i The corresponding label; b) Divide the cell nucleus image set Y into a training set P and a test set T, where P = {P1, P2, ..., P i ,...,P o }, P i is the i-th cell nucleus image in the training set P, i∈{1,...,o}, o is the total number of cell nucleus images in the training set P, T={T1,T2,...,T i ,...,T q }, T i is the i-th cell nucleus image in the test set T, i∈{1,…,q}, q is the total number of cell nucleus images in the test set T; c) Take the i-th cell nucleus image P in the training set P i The corresponding label G i Convert to nucleus edge outline label and heatmap labels All the cell nucleus edge contour labels in the training set P constitute the cell nucleus edge contour label set All heat map labels in the training set P constitute the heat map label set d) Establishing a segmentation network, which consists of a feature extraction module, a mask optimization algorithm module, and a post-processing algorithm module; e) Take the i-th cell nucleus image P in the training set P i Input into the feature extraction module of the segmentation network and output the feature map Feature Map Feature Map f) Using feature maps Get the key point map of the cell nucleus g) The feature map Feature Map Key points of the cell nucleus Input to the mask optimization algorithm module of the segmentation network and output the feature map h) The feature map With feature map Input into the post-processing algorithm module of the segmentation network, and output the final cell nucleus instance segmentation map i) Use the SGD optimizer to take advantage of the total loss function L total Train the segmentation network to obtain the optimized segmentation network; j) Take the i-th cell nucleus image P in the test set T i Input into the optimized segmentation network, and output the cell nucleus instance segmentation map 2. The method for cell nucleus instance segmentation based on multi-scale feature fusion according to claim 1, characterized in that: In step a), n cell nucleus images and their corresponding labels are obtained from the MoNuSeg dataset.
3. The method for cell nucleus instance segmentation based on multi-scale feature fusion according to claim 1, characterized in that: In step b), the cell nucleus image set Y is divided into a training set P and a test set T in a ratio of 7:
3.
4. The method for cell nucleus instance segmentation based on multi-scale feature fusion according to claim 1, characterized in that: Step c) comprises the following steps: c-1) Use cv2.boundingRect() function to obtain the i-th cell nucleus image P in the training set P i The corresponding label G i The boundary rectangle information includes the center point coordinates, the height of the boundary rectangle, and the width of the boundary rectangle; c-2) Input the height and width of the bounding rectangle into the gaussian_radius() function to calculate the radius of the Gaussian kernel; c-3) The diameter Dia of the Gaussian kernel is calculated by Dia=2*(radius+1); c-4) Input the diameter Dia into the gussian2D() function of the astropy library to obtain the Gaussian kernel Gk; c-5) Place the Gaussian kernel Gk at the center coordinate of the bounding rectangle to obtain the heat map label c-6) Take the i-th cell nucleus image P in the training set P i The corresponding label G i Convert to numpy array in numpy format through np.array() function; c-7) Use cv2.Canny() to perform edge detection on the numpy array and convert it into an edge detection image of uint8 type through the np.uint8() function; c-8) Set the values of non-zero pixels in the edge detection image to 255 and the values of the remaining pixels to 0 to obtain the cell nucleus edge contour label 5. The method for cell nucleus instance segmentation based on multi-scale feature fusion according to claim 1, characterized in that: Step e) comprises the following steps: e-1) The feature extraction module of the segmentation network consists of a U-Net model, a first two-dimensional deconvolution layer, a second two-dimensional deconvolution layer, and a third two-dimensional deconvolution layer. The i-th cell nucleus image P in the training set P is i Input into the U-Net model of the feature extraction module, and output the feature map M i ; e-2) The feature map M i Input to the first 2D deconvolution layer of the feature extraction module, and output the feature map e-3) The feature map M i Input to the second 2D deconvolution layer of the feature extraction module, and output the feature map e-4) The feature map M i Input to the third 2D deconvolution layer of the feature extraction module, and output the feature map 6. The method for cell nucleus instance segmentation based on multi-scale feature fusion according to claim 1, characterized in that: In step f), the peak_local_max() function in the skimage package is used to find the feature map Local peak value, get the key point map of the cell nucleus 7. The method for cell nucleus instance segmentation based on multi-scale feature fusion according to claim 1, characterized in that: Step g) comprises the following steps: g-1) The feature map Input into the mask optimization algorithm module of the segmentation network, and use the Image.new() function to initialize a feature map Blank feature maps of the same size g-2) The key point map of the cell nucleus Input into the mask optimization algorithm module of the segmentation network, and obtain the key point map of the cell nucleus through the np.column_stack(np.where()) function The key points of the n nuclei are obtained, and the key point set K is obtained, K = {k1, k2, ..., k i ,...,k n }, k i is the i-th cell nucleus key point, i∈{1,...,n}; g-3) The feature map Input into the mask optimization algorithm module of the segmentation network, and obtain the feature map through the cv2.findContours() function The m cell nucleus boundaries in the , we get the cell nucleus boundary set E, E = {e1, e2, ..., e i ,...,e n },e i is the boundary of the ith cell nucleus; g-4) Use sklearn.neighbors.NearestNeighbors() function to find the i-th cell nucleus key point k i Perform the nearest neighbor search, and when the corresponding boundary i-th cell nucleus boundary e is found i Stop searching when the blank feature map The corresponding position in is marked as 1, and the blank feature map After all corresponding positions are marked, the feature map is obtained 8. The method for cell nucleus instance segmentation based on multi-scale feature fusion according to claim 1, characterized in that: Step h) comprises the following steps: h-1) The feature map With feature map Input to the post-processing algorithm module of the segmentation network for subtraction operation to obtain the difference map h-2) Use cv2.connectedComponentsWithStats() function to compare the difference graph Perform connected component analysis and output the set C of each connected region, C = {c1, c2, ..., c i ,...,c n },c i Difference In the jth small area, if the jth small area c i If the jth small area c is smaller than 5 pixels, it will be deleted. i If it is greater than or equal to 5 pixels, it will be retained to obtain the feature map h-3) Use cv2.dilate() function to dilate the feature map Perform the expansion operation to obtain the final cell nucleus instance segmentation map The convolution kernel during the expansion operation is kernel, 9. The method for cell nucleus instance segmentation based on multi-scale feature fusion according to claim 1, characterized in that: In step i), the formula L total =L p +L edge +L m Calculate the total loss function L total , where L p is the mean square error loss function, the heat map label is the mean square error loss function L p Label map, feature map is the mean square error loss function L p Original image, L edge is the binary cross entropy loss function, feature map is the binary cross entropy loss function L edge Original image, cell nucleus edge outline label is the binary cross entropy loss function L edge Label map of L m is the cross entropy loss function, feature map is the cross entropy loss function L m The original image, the i-th cell nucleus image Y i The corresponding label G i is the cross entropy loss function L m Label map of .
Citation Information
Patent Citations
Cell medical image segmentation method and system
CN114119525A
Cell nucleus segmentation method based on region enhancement
CN114612481A
U-Net-based cell nucleus image segmentation optimization method
CN116309255A
Interactive cell nucleus segmentation method in combination with Voronoi contour and weighted convex difference
CN116664482A
Method for improving cell nucleus segmentation accuracy in cervical cancer liquid-based cell screening analysis
CN117876333A
