CN111951221A - Glomerular cell image identification method based on deep neural network - Google Patents
Glomerular cell image identification method based on deep neural network Download PDFInfo
- Publication number
- CN111951221A CN111951221A CN202010670136.9A CN202010670136A CN111951221A CN 111951221 A CN111951221 A CN 111951221A CN 202010670136 A CN202010670136 A CN 202010670136A CN 111951221 A CN111951221 A CN 111951221A
- Authority
- CN
- China
- Prior art keywords
- image
- neural network
- glomerular
- slice
- segmentation
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 95
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 67
- 210000003904 glomerular cell Anatomy 0.000 title claims abstract description 21
- 230000001575 pathological effect Effects 0.000 claims abstract description 60
- 210000004027 cell Anatomy 0.000 claims abstract description 49
- 230000011218 segmentation Effects 0.000 claims abstract description 48
- 230000001434 glomerular Effects 0.000 claims abstract description 43
- 238000003062 neural network model Methods 0.000 claims abstract description 33
- 238000007781 pre-processing Methods 0.000 claims abstract description 6
- 238000012549 training Methods 0.000 claims description 74
- 210000003855 cell nucleus Anatomy 0.000 claims description 28
- 238000002372 labelling Methods 0.000 claims description 14
- 238000005520 cutting process Methods 0.000 claims description 13
- 230000008569 process Effects 0.000 claims description 13
- 238000010586 diagram Methods 0.000 claims description 12
- 238000012805 post-processing Methods 0.000 claims description 10
- 238000005070 sampling Methods 0.000 claims description 10
- 230000008602 contraction Effects 0.000 claims description 8
- 230000002159 abnormal effect Effects 0.000 claims description 6
- 239000011159 matrix material Substances 0.000 claims description 6
- 210000004940 nucleus Anatomy 0.000 claims description 5
- 230000004913 activation Effects 0.000 claims description 4
- 238000002790 cross-validation Methods 0.000 claims description 3
- 239000000284 extract Substances 0.000 claims description 3
- 238000012216 screening Methods 0.000 claims description 3
- 238000012360 testing method Methods 0.000 claims description 3
- 238000012545 processing Methods 0.000 abstract description 10
- 238000005516 engineering process Methods 0.000 abstract description 6
- 238000013473 artificial intelligence Methods 0.000 abstract description 5
- 238000013135 deep learning Methods 0.000 abstract description 5
- 230000004927 fusion Effects 0.000 abstract 1
- 210000003734 kidney Anatomy 0.000 abstract 1
- 238000003709 image segmentation Methods 0.000 description 29
- 230000000694 effects Effects 0.000 description 10
- 238000000605 extraction Methods 0.000 description 10
- 238000003672 processing method Methods 0.000 description 7
- 210000001519 tissue Anatomy 0.000 description 6
- 238000003745 diagnosis Methods 0.000 description 5
- 230000003044 adaptive effect Effects 0.000 description 4
- 238000001514 detection method Methods 0.000 description 4
- 230000000007 visual effect Effects 0.000 description 4
- 238000001574 biopsy Methods 0.000 description 3
- 230000006870 function Effects 0.000 description 3
- 210000005084 renal tissue Anatomy 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 230000008901 benefit Effects 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 239000003086 colorant Substances 0.000 description 2
- 238000013527 convolutional neural network Methods 0.000 description 2
- 238000012937 correction Methods 0.000 description 2
- 230000007797 corrosion Effects 0.000 description 2
- 238000005260 corrosion Methods 0.000 description 2
- 238000013136 deep learning model Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000018109 developmental process Effects 0.000 description 2
- 238000002059 diagnostic imaging Methods 0.000 description 2
- 238000003384 imaging method Methods 0.000 description 2
- 230000001788 irregular Effects 0.000 description 2
- 238000011862 kidney biopsy Methods 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 230000007170 pathology Effects 0.000 description 2
- 238000002679 ablation Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000002902 bimodal effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 201000010099 disease Diseases 0.000 description 1
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 239000003814 drug Substances 0.000 description 1
- 230000003834 intracellular effect Effects 0.000 description 1
- 208000017169 kidney disease Diseases 0.000 description 1
- 230000003902 lesion Effects 0.000 description 1
- 239000003550 marker Substances 0.000 description 1
- 230000004660 morphological change Effects 0.000 description 1
- 238000011369 optimal treatment Methods 0.000 description 1
- 238000010827 pathological analysis Methods 0.000 description 1
- 230000035755 proliferation Effects 0.000 description 1
- 238000004451 qualitative analysis Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000002271 resection Methods 0.000 description 1
- 238000007790 scraping Methods 0.000 description 1
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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- 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
- G06T7/10—Segmentation; Edge detection
- G06T7/136—Segmentation; Edge detection involving thresholding
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- 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
- G06T2207/30024—Cell structures in vitro; Tissue sections in vitro
-
- 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
- G06T2207/30084—Kidney; Renal
-
- 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
- G06T2207/30242—Counting objects in image
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Abstract
The invention discloses a glomerular cell image recognition method based on a deep neural network, which is based on artificial intelligence and a deep learning technology to obtain a pathological image to be detected; preprocessing a pathological image to obtain a plurality of slice images; inputting each slice image into a preset neural network model for recognition and segmentation to obtain a glomerular region map; performing cell counting on the glomerular region map; sub-images of glomeruli inside the kidney in the pathological image can be segmented rapidly and accurately, cells inside the glomeruli are counted by using a traditional method and a deep learning fusion model, and the problems of large workload, low efficiency and high misdiagnosis rate of artificial identification of the glomeruli in the pathological image are solved; the method optimizes the algorithm for segmenting the sub-glomerular images and counting the glomerular cells in the pathological images, uses more data to train the algorithm, and improves the accuracy of segmentation and counting, and relates to the field of biomedical image processing.
Description
Technical Field
The invention relates to the technical field of biomedical image processing and deep learning, in particular to a glomerular cell image identification method based on a deep neural network.
Background
The rapid development of modern medicine has prompted detection methods and display means to be increasingly more accurate, more intuitive, and more sophisticated. The human body medical image contains abundant human body information, and after the image is obtained by means of a medical imaging technology, the image needs to be analyzed, identified, segmented, calibrated, classified and interpreted, and according to clinical application of the medical image and medical problems needing to be solved, parts which need to be enhanced or features which need to be extracted are determined, so that more visual data are provided. The method can be used as a basis for reasonably arranging the examination procedures of the patients, so that the most objective diagnosis purpose can be achieved by the fastest speed and the most economic means, and the selection, the determination and the implementation of the optimal treatment scheme are further facilitated. Therefore, processing the acquired medical image by a high-tech means is a research work which needs to be completed urgently in medical imaging. The method has important significance for reducing misdiagnosis probability caused by human factors, improving the working efficiency of medical workers and providing more objective diagnosis basis. Clinically, the kidney disease examination generally adopts biopsy, and the obtained diagnosis result is the most accurate. Biopsy, known as biopsy, is a pathological examination of a diseased tissue taken from a patient's living body by surgical methods such as partial resection, clamping, aspiration with a puncture needle, scraping, and ablation to confirm a diagnosis. The method has the advantages of fresh tissue and basically maintained original appearance of the lesion. For clinical work, the examination method is helpful for timely and accurately diagnosing diseases and judging the curative effect. The kidney tissue activity detection is generally performed by observing a tissue section image obtained by kidney biopsy by an experienced physician to obtain the pathological morphological change of glomeruli and the proliferation and distribution condition of nucleus in the glomerulus, and a series of conclusions are obtained after the analysis is performed by matching with the clinical experience of the physician, so that a pathological diagnosis report is given. In order to ensure that existing problems can be found, doctors need to observe the obtained kidney biopsy tissue slices as many as possible, about 800 pictures of each examined person need to be observed clinically, and the number shows that the workload of identifying the target object through visual observation is large, the efficiency is low, and the misdiagnosis rate is increased along with the increase of the fatigue degree of the observer. With the rapid development of computer technology, imaging technology and image processing technology, the automatic analysis of the kidney tissue section becomes possible, and after the intelligent processing of the graph by the computer, the traditional visual observation and the supervisor judgment of the image diagnosis can be abandoned. The deep learning algorithm based on pathological image glomerular segmentation and intraglomerular cell counting is provided, glomerular extraction and intraglomerular nucleus counting are completed, and qualitative analysis and quantitative calculation of kidney tissue slice images are realized. At present, the accuracy of glomerular segmentation, cell segmentation and technology thereof needs to be improved, the accuracy cannot reach the practical level, further optimization is needed, the accuracy of glomerular segmentation is improved, and the intraglomerular cell count is realized. It has mainly 3 defects: 1. the traditional method for identifying glomeruli in pathological images through visual observation by doctors has large workload and low efficiency, and the misdiagnosis rate can also rise along with the increase of the fatigue degree of observers. 2. The accuracy of the existing glomerulus segmentation and glomerulus cell counting algorithm is low, continuous optimization is needed, and the segmentation and counting accuracy is improved. 3. Because pathological images are large and cells in glomeruli are dense, the traditional single neural network method is low in recognition rate, and better recognition can be achieved only through a specific post-processing method after deep learning segmentation.
Disclosure of Invention
The present invention is directed to a glomerular cell image recognition method based on a deep neural network, so as to solve one or more technical problems in the prior art and provide at least one useful choice or creation condition.
The invention provides a glomerular cell image identification method based on a deep neural network, which utilizes a corresponding ROI extraction algorithm to extract an interested area of a pathological image, filters a blank area of the pathological image and reduces the workload of subsequent image cutting; the process of observing the slices by a doctor using a microscope is simulated by adopting a multilayer cutting or layer-by-layer focusing picture cutting method; when the slice is generated, multiple layers of slices with different sizes are generated to prepare for subsequent characteristic extraction, and the generalization performance of learning is enhanced; by combining a threshold-based image segmentation method, such as histogram bimodal image segmentation, fixed threshold image segmentation, half threshold segmentation, iterative threshold image segmentation, adaptive threshold image segmentation, optimal threshold image segmentation and the like with a deep neural network cell detection algorithm of FastFCN, SegNet, Unet and various variants, RefineNet and the like, the glomerular cells are counted more accurately.
The invention aims to solve the problems and provides a glomerular cell image recognition method based on a deep neural network, which specifically comprises the following steps:
acquiring a pathological image to be detected;
preprocessing a pathological image to obtain a plurality of slice images;
inputting each slice image into a recognition segmentation neural network model for recognition segmentation to obtain a glomerular region slice image;
and inputting the glomerular region slice image into a preset neural network model to identify cells in the glomerulus and count the cells.
Further, the method for preprocessing the pathological image to obtain a plurality of slice images comprises the following steps:
labeling and extracting an interested region in the pathological image:
the Region of the pathological image needing to be processed is marked in a square frame, circle, ellipse, irregular polygon and the like, and is called a Region of Interest (ROI). For a pathological full-field slice (WSI), an edge blank area generated during slice production and gaps between tissues exist in an Image, and the workload and the difficulty of subsequent processing can be effectively reduced by using a corresponding ROI extraction algorithm.
Obtaining a plurality of slice images by performing multi-slice segmentation in the region of interest in the pathological image:
one big characteristic of WSI pathology images is that the number of image pixels is very large, generally in the scale range of 10^6 × 10^ 6. Due to the limitations of the physical memory and the computing speed of the computer, it is impractical to directly input the entire WSI into the memory of the computer for operation. Therefore, image segmentation, that is, image segmentation processing, is performed on the WSI image according to the characteristics of the WSI image, and after the pathological image is processed into a plurality of slice images, corresponding training and prediction are performed. The multilayer cutting or layer-by-layer focusing picture cutting method simulates the process of observing the section by a doctor using a microscope. In general, with a constant field of view of the lens, as the magnification increases, the physical field of view scales down and the pixel resolution scales up. The concept of layer-by-layer focusing is used here, and when generating slices, multiple layers of slices with different sizes are generated to prepare for subsequent characteristic extraction. The method can ensure that the image characteristics such as textures, colors and the like with different scales can be well observed in the deep convolutional neural network, and can effectively capture massive large focuses and tiny small focuses so as to enhance the generalization performance of learning.
Further, the method for extracting the region of interest in the pathological image comprises any one of histogram doublet method image segmentation, fixed threshold value image segmentation, half threshold value segmentation image segmentation, iterative threshold value image segmentation, adaptive threshold value image segmentation and optimal threshold value image segmentation.
Furthermore, the method for obtaining the glomerular region slice image by inputting each slice image into the recognition segmentation neural network model for recognition segmentation comprises the following steps:
constructing any one neural network of FastFCN, SegNet, U-net, various variants and RefineNet;
dividing each slice image into a training set and a test set by any one of a leave-out method, a cross-validation method, a leave-one-out method and a self-service method;
the preset number of the slice images in the training set can be selected by adjusting the size of the cutting graph according to the learning training requirement of the neural network; the more the predetermined number, the higher the accuracy of the trained identified segmented neural network model.
Inputting the slice images in the training set into a neural network for training to obtain slice images in the training set, inputting the slice images in the training set into the neural network for training to obtain a recognition segmentation neural network model, and inputting each slice image into a preset neural network model for recognition segmentation to obtain a slice image containing a glomerular region.
Preferably, the neural network is a U-Net, the U-Net is a full convolution neural network, and the network is an end-to-end network, namely, the input and the output are images; inputting the pathological image into U-Net, performing downsampling on the pathological image by adopting an activation function after passing through a convolution layer in a contraction path, extracting a feature map of the pathological image, performing upsampling in an expansion path, and adding a corresponding feature map obtained in the contraction path during each upsampling, thereby finally realizing the extraction of a first abnormal probability matrix of each pixel point in the pathological image; the input of the model is a three-channel picture with the size of 256 × 3(256, 256, 3), the output is a single-channel picture with the size of 256 × 1(256, 256, 1), and the input size of the picture can be adjusted according to actual conditions.
Preferably, the U-Net is composed of an encoder and a decoder, the input of the encoder is a slice image, the encoder outputs the feature map of size 963 × 32 output by the fourth layer to the fourth layer of the decoder, and outputs the feature maps output by the first, second and third layers to the first, second and third layers of the decoder, respectively; the encoder is used for extracting cell characteristic information: color, morphology, size, position, texture, etc., consisting of 5 levels, each level containing a number of modified resblocks, and using a downsampling operation to generate input for the next level, the encoder input image in the neural network being an RGB color map of 256 × 3, the encoder first level generating a first level profile of 256 × 67 size with 1 modified ResBlock, generating a 2 nd level input of 128 × 67 size with 1 downsampling, the 2 nd level generating a second level profile of 128 × 195 size with 1 modified ResBlock, generating a 3 rd level input of 64 × 195 size with 1 downsampling, the third level generating a third level profile of 64 × 451 size with 1 downsampling, generating a 4 th level input of 32 × 451 size with 1 modified ResBlock, the fourth level generating a fourth level 963 rd level profile of 64 × 451 size with 1 modified ResBlock, the 5 th layer input with the size of 16 x 963 is obtained by 1 time of down-sampling, the fifth layer generates a feature map with the size of 16 x 1987 as a final feature map of a decoder through 1 improved ResBlock, and therefore the encoder extracts global information of the whole image with low resolution through the operation of convolution and down-sampling level by level; therefore, the context information of the cells in the whole image can be provided, and the context information can be understood as the characteristics of the dependence relationship between the cells and the surrounding environment, and the characteristics are helpful for carrying out classification judgment on the cells.
The decoder consists of 4 levels, and the second layer, the third layer and the fourth layer respectively output the feature map to the previous layer through up-sampling;
in the fourth level, the 16 × 963 size feature map of the 5 th level of the encoder is firstly subjected to upsampling to generate a 32 × 512 size feature map, then the feature maps with the same size as the fourth level of the encoder are spliced through skip connection, and finally the feature map with the size of 32 × 1987 of the fourth level of the decoder is generated through 1 improved ResBlock. Similarly, the third, second and first stages of the decoder perform the same process, and the corresponding feature sizes produced are 64 × 963, 128 × 451, 256 × 195. The skip connection on the same layer level of the encoder and the decoder can fuse the fine position information of the encoder and the rich-expression semantic information of the decoder, so that the features with stronger expression capability are extracted for the segmentation task.
Furthermore, the invention uses the improved U-Net deep neural network to segment the glomeruli and the cells in the glomeruli in the small picture, and merges the characteristic diagrams by taking advantage of the shorcuit connection mode of the ResNet model and the concat mode in the DenseNet model on the basis of the traditional U-Net, namely, the final summation operation of the original ResBlock is replaced by the concat operation, namely, the improved ResBlock, thereby obtaining the basic structure used by the invention, thus being capable of well realizing the multiplexing of the characteristic diagrams and improving the effect of the model. The network can identify pathological features of the glomeruli and accurately segment the glomeruli, and the segmentation accuracy can reach 98.1%.
Training the neural network through a training set to obtain slice images in the training set, inputting the slice images into the neural network, and training to obtain a preset neural network model;
inputting the slice images in the training set into a neural network for training;
because the slice images in the training set are larger, the slice images in the training set are directly input into a neural network for training, and because of the limitation of computer hardware, a computer reports memory overflow errors, so that training fails; therefore, before the pictures are input into the neural network for training, the slice images in the training set are cut into the size of 256 × 3(256, 256, 3), then every 8 pictures of 256 × 3 are input into the neural network as a batch for training, the binary labeled images corresponding to the slice images are processed in the same way, and the size of the cut small pictures and the input batch size can be adjusted according to the corresponding hardware level of the computer.
Further, the training process of the preset neural network model comprises the following steps:
labeling glomeruli in the pathological image, and performing sub-image overlapping cutting on a region of interest (ROI) in the pathological image according to a certain size (such as 2048 × 3 size) to obtain sub-image data;
the sub-image data is input into a neural network for training, and image enhancement operation is carried out, wherein the image enhancement operation comprises random overturning along an x or y axis of an image matrix, rotation according to a random angle and random contrast adjustment, so that the generalization capability of the neural network is improved;
after training of the glomerular prediction model is completed, the prediction effect of the model is further improved by using a specific post-processing method on the output result of the model, namely, the prediction result is optimized to obtain the final predicted glomerular label, and the post-processing method comprises corrosion and expansion of the prediction result and screening of the prediction result; according to the finally predicted glomerular marker, segmenting the glomerulus from the pathological image by using opennlide to obtain a segmented subgraph containing a glomerular image and generate a binary labeled image corresponding to the segmented subgraph, marking the part, which does not contain the glomerulus, in the binary labeled image as a background mark 0 and the part containing the glomerulus as an abnormal value mark 1, and removing cells in the region except the glomerulus by using the binary image of the glomerulus in the cell counting process;
further, inputting the glomerular region slice image into a preset neural network model for identifying the cells in the glomerulus and counting the cells as follows:
carrying out cell nucleus marking on the glomerular region slice, and taking the marked binary image as a label corresponding to the glomerular cell nucleus; after cell nucleuses in the glomeruli are labeled and corresponding binary images are generated, the segmentation subgraphs are used as training data, and the corresponding binary images are used as labels and input into an improved neural network for training.
After the training of the cell prediction model is finished, the cell nucleuses in the segmentation subgraph are predicted by using the trained neural network, the predicted cell nucleuses are the cell nucleuses in the whole segmentation subgraph and also comprise the cell nucleuses in the areas except the glomeruli, at the moment, the binary images of the glomeruli in the corresponding segmentation subgraph are multiplied by the predicted binary images of the cell nucleuses, the areas except the glomeruli are changed into 0, and the numerical values of the areas in the glomeruli are not changed.
Segmenting the glomerulus from the pathological image according to the labeling of the glomerulus to obtain a segmented subgraph containing a glomerulus image and generate a binary image corresponding to the segmented subgraph, marking the part, which does not contain the glomerulus, of the binary image as a background mark 0, and marking the part containing the glomerulus as 1; and inputting the slice image into a neural network for training to obtain a preset neural network model.
And then, the nuclear prediction result is optimized by using post-processing methods such as morphology and the like, so that the problems of noise, adhesion and the like in the nuclear prediction result are solved, and the final prediction result is improved.
Then, the number of nuclei in the predicted binary image is counted to count the cells.
The invention has the beneficial effects that: the invention discloses a glomerular cell image recognition method based on a deep neural network, which effectively combines an artificial intelligence method and a traditional post-processing method together: the identification of glomeruli and cells thereof is a more complex process, the traditional image processing method is difficult to realize, meanwhile, a deep learning model cannot perform specific correction on a certain detail well, and the traditional image processing method can adjust specific details well. Therefore, the algorithm uses an artificial intelligence method to carry out overall prediction and uses a traditional image processing method to carry out detail adjustment, thereby achieving better cell identification effect. The problems that the workload of a traditional doctor for identifying glomeruli in a pathological image by naked eyes is large, the efficiency is low, and the misdiagnosis rate is higher and higher after the doctor observes for a long time are solved; the method optimizes the algorithm for segmenting the sub-glomerular images and counting the glomerular cells in the pathological images, trains the algorithm by using more data and improves the accuracy of segmentation and counting.
Drawings
The above and other features of the present invention will become more apparent by describing in detail embodiments thereof with reference to the attached drawings in which like reference numerals designate the same or similar elements, it being apparent that the drawings in the following description are merely exemplary of the present invention and other drawings can be obtained by those skilled in the art without inventive effort, wherein:
FIG. 1 is a flow chart of a glomerular cell image recognition method based on a deep neural network according to the present invention;
FIG. 2 is a graph showing image contrast before and after a threshold segmentation method;
fig. 3 is a diagram showing a neural network model structure used in the present invention.
Detailed Description
The conception, the specific structure and the technical effects of the present invention will be clearly and completely described in conjunction with the embodiments and the accompanying drawings to fully understand the objects, the schemes and the effects of the present invention. It should be noted that the embodiments and features of the embodiments in the present application may be combined with each other without conflict.
Referring to fig. 1, a flowchart of a glomerular cell image recognition method based on a deep neural network according to the present invention is shown, and the method according to an embodiment of the present invention is described below with reference to fig. 1.
The invention provides a glomerular cell image identification method based on a deep neural network, which specifically comprises the following steps:
acquiring a pathological image to be detected;
preprocessing a pathological image to obtain a plurality of slice images;
inputting each slice image into a recognition segmentation neural network model for recognition segmentation to obtain a glomerular region slice image;
and inputting the glomerular region slice image into a preset neural network model to identify cells in the glomerulus and count the cells.
Further, the method for preprocessing the pathological image to obtain a plurality of slice images comprises the following steps:
labeling and extracting an interested region in the pathological image:
in machine vision, image processing problems, the Region to be processed, called Region of Interest, i.e. Region of Interest (ROI), is delineated from the processed image in the form of a box, circle, ellipse, irregular polygon, etc. For a pathological full-field slice/pathological Image (WSI), an edge blank area generated during slice production and gaps between tissues exist in the Image, and the workload and the difficulty of subsequent processing can be effectively reduced by using a corresponding ROI extraction algorithm.
Obtaining a plurality of slice images by performing multi-slice segmentation in the region of interest in the pathological image:
a large feature of pathological imaging (WSI) is that the number of image pixels is extremely large, generally in the scale range of 10^6 × 10^ 6. Due to the limitations of the physical memory and the computing speed of the computer, it is impractical to directly input the entire WSI into the memory of the computer for operation. Therefore, image segmentation, that is, image segmentation processing, is performed on the WSI image according to the characteristics of the WSI image, and after the pathological image is processed into a plurality of slice images, corresponding training and prediction are performed. The multilayer cutting or layer-by-layer focusing picture cutting method simulates the process of observing the section by a doctor using a microscope. In general, with a constant field of view of the lens, as the magnification increases, the physical field of view scales down and the pixel resolution scales up. The concept of layer-by-layer focusing is used here, and when generating slices, multiple layers of slices with different sizes are generated to prepare for subsequent characteristic extraction. The method can ensure that the image characteristics such as textures, colors and the like with different scales can be well observed in the deep convolutional neural network, and can effectively capture massive large focuses and tiny small focuses so as to enhance the generalization performance of learning.
Further, the method for extracting the region of interest in the pathological image comprises any one of histogram doublet method image segmentation, fixed threshold value image segmentation, half threshold value segmentation image segmentation, iterative threshold value image segmentation, adaptive threshold value image segmentation and optimal threshold value image segmentation.
As shown in fig. 2, the image contrast before and after the threshold segmentation method, and an image segmentation method based on a threshold value, such as histogram doublet method image segmentation, fixed threshold value image segmentation, half threshold value segmentation image segmentation, iterative threshold value image segmentation, adaptive threshold value image segmentation, optimal threshold value image segmentation, etc., is used. The method uses the idea of clustering, and divides the gray scale number of the image into 2 parts according to the gray scale, so that the gray scale value difference between the two parts is maximum. The processing method is applied to the extraction of the ROI in the WSI image so as to reduce the subsequent calculation amount.
Furthermore, the method for obtaining the glomerular region slice image by inputting each slice image into the recognition segmentation neural network model for recognition segmentation comprises the following steps:
constructing any one neural network of FastFCN, SegNet, U-net, various variants and RefineNet;
dividing each slice image into a training set and a test set by any one of a leave-out method, a cross-validation method, a leave-one-out method and a self-service method;
the method for training the neural network through the training set to obtain the preset neural network model by inputting the slice images in the training set into the neural network for training comprises the following steps: and inputting the slice images in the training set into a neural network for training to obtain slice images in the training set, and inputting the slice images in the training set into the neural network for training to obtain a preset neural network model.
The preset number of the slice images in the training set can be selected by changing the size of the slice according to the learning training requirement of the neural network; the more the predetermined number, the higher the accuracy of the trained preset neural network model.
Preferably, the neural network is a U-Net, the U-Net is a full convolution neural network, and the network is an end-to-end network, namely, the input and the output are images; inputting the pathological image into U-Net, performing downsampling on the pathological image by adopting an activation function after passing through a convolution layer in a contraction path, extracting a feature map of the pathological image, performing upsampling in an expansion path, and adding a corresponding feature map obtained in the contraction path during each upsampling, thereby finally realizing the extraction of a first abnormal probability matrix of each pixel point in the pathological image; the input of the model is a three-channel picture with the size of 256 × 3(256, 256, 3), the output is a single-channel picture with the size of 256 × 1(256, 256, 1), and the input size of the picture can be adjusted according to actual conditions.
As shown in fig. 3, the neural network model structure used in the present invention is a neural network model structure, the neural network model structure U-Net is composed of an encoder and a decoder, the input of the encoder is a slice image, the encoder outputs the 963 × 32 feature maps output by the fourth layer to the fourth layer of the decoder, and the feature maps output by the first, second and third layers are output to the first, second and third layers of the decoder, respectively; the encoder is used for extracting cell characteristic information: color, morphology, size, position, texture, etc., consisting of 5 levels, each level containing a number of modified resblocks, and using a downsampling operation to generate input for the next level, the encoder input image in the neural network being an RGB color map of 256 × 3, the encoder first level generating a first level profile of 256 × 67 size with 1 modified ResBlock, generating a 2 nd level input of 128 × 67 size with 1 downsampling, the 2 nd level generating a second level profile of 128 × 195 size with 1 modified ResBlock, generating a 3 rd level input of 64 × 195 size with 1 downsampling, the third level generating a third level profile of 64 × 451 size with 1 downsampling, generating a 4 th level input of 32 × 451 size with 1 modified ResBlock, the fourth level generating a fourth level 963 rd level profile of 64 × 451 size with 1 modified ResBlock, the 5 th layer input with the size of 16 x 963 is obtained by 1 time of down-sampling, the fifth layer generates a feature map with the size of 16 x 1987 as a final feature map of a decoder through 1 improved ResBlock, and therefore the encoder extracts global information of the whole image with low resolution through the operation of convolution and down-sampling level by level; therefore, the context information of the cells in the whole image can be provided, and the context information can be understood as the characteristics of the dependence relationship between the cells and the surrounding environment, and the characteristics are helpful for carrying out classification judgment on the cells.
The decoder consists of 4 levels, and the second layer, the third layer and the fourth layer respectively output the feature map to the previous layer through up-sampling;
in the fourth level, the 16 × 963 size feature map of the 5 th level of the encoder is firstly subjected to upsampling to generate a 32 × 512 size feature map, then the feature maps with the same size as the fourth level of the encoder are spliced through skip connection, and finally the feature map with the size of 32 × 1987 of the fourth level of the decoder is generated through 1 improved ResBlock. Similarly, the third, second and first stages of the decoder perform the same process, and the corresponding feature sizes produced are 64 × 963, 128 × 451, 256 × 195. The skip connection on the same layer level of the encoder and the decoder can fuse the fine position information of the encoder and the rich-expression semantic information of the decoder, so that the features with stronger expression capability are extracted for the segmentation task.
Further, the invention uses an improved U-net neural network to segment the cells in the small picture, the network can identify the pathological features of the glomerulus, and the glomerulus cells can be accurately segmented.
Training the neural network through a training set to obtain slice images in the training set, inputting the slice images into the neural network, and training to obtain a preset neural network model;
inputting the slice images in the training set into a neural network for training;
because the slice images in the training set are larger, the slice images in the training set are directly input into a neural network for training, and because of the limitation of computer hardware, a computer reports memory overflow errors, so that training fails; therefore, before the pictures are input into the neural network for training, the slice images in the training set are cut into the size of 256 × 3, then every 8 pictures with the size of 256 × 3 are input into the neural network as a batch for training, the binary images corresponding to the slice images are processed in the same way, and the size of the cut small pictures and the input batch size can be adjusted according to the corresponding hardware level of the computer.
After the training of the glomerular prediction model is completed, the prediction effect of the model is further improved by using a specific post-processing method (including but not limited to corrosion and expansion of the prediction result and screening of the prediction result) on the output result of the model, namely, the prediction result is optimized to obtain the final predicted glomerular label.
Further, inputting the glomerular region slice image into a preset neural network model for identifying the cells in the glomerulus and counting the cells comprises the following steps:
the glomeruli are cut out from the WSI according to labeling of the glomeruli by opennlide, the glomeruli are stored in a jpg format, a glomerular binary labeling picture corresponding to the glomeruli is generated, a background which is a part not including the glomeruli in the binary labeling picture is set to be 0, an abnormal value which is a part including the glomeruli is set to be 1, and cell nuclei in a region except the glomeruli are removed by using a glomerular binary labeling image in the cell counting process.
Training a glomerular intracellular prediction model:
labeling the cell nuclei in the glomerular slices, labeling the cell nuclei in the glomerulus and generating a corresponding cell nucleus binary labeling map, then using the glomerular slices as training data, using the corresponding cell nucleus binary labeling data as labels to input into an improved U-Net model for training, wherein the result of the model is the same as that of the glomerular prediction model, and the difference is that the input is changed into 128 × 3, and the output is changed into 128 × 1.
When the picture is input into the model training, considering that the prediction effect of the segmentation algorithm on the picture edge is not good, before the picture is input into the model training, the glomerular slice picture is cut into smaller size, so that the overlap cutting and prediction can be carried out in the subsequent prediction.
After the neural network is trained, the cell nuclei in the slice images are predicted through a preset neural network model, the cell nuclei in the glomerular slice images are obtained through prediction, the cell nuclei also comprise the cell nuclei of the areas except the glomeruli, at the moment, the binary labeled images corresponding to the glomerular slice images are multiplied by the predicted cell nuclei binary labeled images, so that the areas except the glomeruli are changed into 0, and the numerical values of the areas inside the glomeruli are unchanged.
And then, the nuclear prediction result is optimized by using post-processing methods such as morphology and the like, so that the problems of noise, adhesion and the like in the nuclear prediction result are solved, and the final prediction result is improved. Then, the number of the cell nuclei in the predicted binary image is counted to count the cells.
The method of the invention effectively combines an artificial intelligence method with a traditional post-processing method: the identification of glomeruli and cells thereof is a more complex process, the traditional image processing method is difficult to realize, meanwhile, a deep learning model cannot perform specific correction on a certain detail well, and the traditional image processing method can adjust specific details well. Therefore, the algorithm uses an artificial intelligence method to carry out overall prediction and uses a traditional image processing method to carry out detail adjustment, thereby achieving better cell identification effect.
Further, when a WSI map of renal pathology is predictively diagnosed, the algorithm flow can be summarized as: cutting a WSI interested area; predicting a glomerular region by a glomerular prediction model; correcting a prediction result by a traditional method, and cutting out a glomerular region; predicting cells within the glomerular region by a cell prediction model; the traditional method corrects the prediction result and counts the number of cells in the current glomerular region.
U-Net is a kind of full convolution neural network, it is a end-to-end network, i.e. the input and output are all images. Inputting the pathological image into a U-Net segmentation model, performing downsampling on the pathological image by adopting an activation function after passing through a convolution layer in a contraction path, extracting a feature map of the pathological image, performing upsampling in an expansion path, adding a corresponding feature map obtained in the contraction path during each upsampling, and finally extracting a first abnormal probability matrix of each pixel point in the pathological image.
The input of the model is a three-channel picture of (256, 256, 3), the output is a single-channel picture of (256, 256, 1), and the input size of the picture can be adjusted according to the actual situation.
In order to ensure the accuracy of cell counting, a deep neural network cell detection algorithm is also used, and the algorithm can quickly detect all cells in the glomerulus through a large amount of training. And finally, integrating the results obtained by the two methods to obtain the accurate cell number.
Although the description of the present disclosure has been rather exhaustive and particularly described with respect to several illustrated embodiments, it is not intended to be limited to any such details or embodiments or any particular embodiments, so as to effectively encompass the intended scope of the present disclosure. Furthermore, the foregoing describes the disclosure in terms of embodiments foreseen by the inventor for which an enabling description was available, notwithstanding that insubstantial modifications of the disclosure, not presently foreseen, may nonetheless represent equivalent modifications thereto.
Claims (7)
1. A glomerular cell image identification method based on a deep neural network is characterized by comprising the following steps:
acquiring a pathological image to be detected;
preprocessing a pathological image to obtain a plurality of slice images;
inputting each slice image into a recognition segmentation neural network model for recognition segmentation to obtain a glomerular region slice image;
and inputting the glomerular region slice image into a preset neural network model to identify cells in the glomerulus and count the cells.
2. The method for recognizing the glomerular cell image based on the deep neural network as claimed in claim 1, wherein the method for inputting each slice image into the recognition segmentation neural network model to perform recognition segmentation to obtain the glomerular region slice image comprises the following steps:
the obtaining process of identifying and segmenting the neural network model comprises the following steps:
constructing any one neural network of FastFCN, SegNet, U-net, various variants and RefineNet;
dividing each slice image into a training set and a test set by any one of a leave-out method, a cross-validation method, a leave-one-out method and a self-service method;
inputting the slice images in the training set into a neural network for training to obtain slice images in the training set, inputting the slice images in the training set into the neural network for training to obtain a recognition segmentation neural network model, and inputting each slice image into the recognition segmentation neural network model for recognition segmentation to obtain a slice image containing a glomerular region.
3. The method for identifying the glomerular cell image based on the deep neural network as claimed in claim 2, wherein the neural network is U-Net, and the U-Net is a full convolution neural network which is an end-to-end network, namely, the input and the output of the network are images; inputting the pathological image into U-Net, performing downsampling on the pathological image by adopting an activation function after passing through a convolution layer in a contraction path, extracting a feature map of the pathological image, performing upsampling in an expansion path, adding a corresponding feature map obtained in the contraction path during each upsampling, and finally extracting a first abnormal probability matrix of each pixel point in the pathological image.
4. The method as claimed in claim 3, wherein the U-Net is composed of an encoder and a decoder, the input of the encoder is a slice image, the encoder outputs the feature map output from the fourth layer to the decoder, and outputs the feature maps output from the first, second and third layers to the first, second and third layers of the decoder, respectively; the encoder is used for extracting cell characteristic information: color, morphology, size, position, texture, consisting of 5 levels, each level containing a number of modified ResBlock and downsampling operations, in the neural network, a coder level 1 generates a first layer characteristic diagram through 1 improved ResBlock, a level 2 input is obtained through 1 down sampling, a level 2 generates a second layer characteristic diagram through 1 improved ResBlock, a level 3 input is obtained through 1 down sampling, a third layer generates a third layer characteristic diagram through 1 improved ResBlock, a level 4 input is obtained through 1 down sampling, a fourth layer generates a fourth layer characteristic diagram through 1 improved ResBlock, a level 5 input is obtained through 1 down sampling, and a fifth layer generates a decoder final characteristic diagram through 1 improved ResBlock, thus, the encoder extracts global information of the whole image with low resolution through convolution and downsampling operation level by level; the decoder comprises 4 levels, a fourth level generates a characteristic diagram of a 5 th level of the encoder by upsampling the characteristic diagram, then splices the characteristic diagrams with the same size as the fourth level of the encoder by jump connection, and finally generates a fourth-level characteristic diagram of the decoder by 1 improved ResBlock; similarly, the decoder performs the same process at layer 3, layer 2 and layer 1.
5. The method as claimed in claim 4, wherein the modified ResBlock is obtained by replacing the last summation operation in the original ResBlock with a concat operation.
6. The method for recognizing the glomerular cell image based on the deep neural network as claimed in claim 1, wherein the training process of the preset neural network model comprises the following steps:
labeling the glomeruli in the pathological image, and performing sub-image overlapping cutting on an interest region in the pathological image to obtain sub-image data;
inputting the sub-image data into a neural network for training, and performing image enhancement operation, wherein the image enhancement operation comprises random overturning along an x or y axis of an image matrix, rotation according to a random angle and random contrast adjustment;
segmenting the glomerulus from the pathological image according to the labeling of the glomerulus to obtain a segmented subgraph containing a glomerulus image and generate a binary image corresponding to the segmented subgraph, marking the part, which does not contain the glomerulus, of the binary image as a background mark 0, and marking the part containing the glomerulus as 1; and inputting the slice image into a neural network for training to obtain a preset neural network model.
7. The method for identifying the glomerular cell image based on the deep neural network as claimed in claim 6, wherein the method for identifying and counting the cells in the glomerulus by inputting the glomerular region slice image into a preset neural network model comprises the following steps:
marking a cell nucleus area in the glomerular area slice image, and taking a binary image after marking the cell nucleus area as a label corresponding to the glomerular cell nucleus;
marking a cell nucleus region in the glomerular region slice image and generating a corresponding binary image, taking the segmentation subgraph as training data, and inputting the corresponding binary image as a label into a neural network for training;
after the neural network training is finished, predicting cell nucleuses in the segmentation subgraph by using the trained neural network, wherein the predicted cell nucleus region is the cell nucleus in the whole segmentation subgraph and also comprises the cell nucleus of a region except for the glomerulus, and at the moment, multiplying the binary image of the glomerulus in the corresponding segmentation subgraph by the predicted cell nucleus binary image to change the region except for the glomerulus into 0, and keeping the value of the region inside the glomerulus unchanged;
optimizing the prediction result by a post-processing method to obtain a finally predicted cell nucleus label, wherein the post-processing method comprises the steps of corroding the prediction result and screening the prediction result by a threshold value; and finally, counting the number of the cell nuclei in the predicted binary image so as to count the cells.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010670136.9A CN111951221B (en) | 2020-07-13 | 2020-07-13 | Glomerular cell image recognition method based on deep neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010670136.9A CN111951221B (en) | 2020-07-13 | 2020-07-13 | Glomerular cell image recognition method based on deep neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111951221A true CN111951221A (en) | 2020-11-17 |
| CN111951221B CN111951221B (en) | 2023-10-31 |
Family
ID=73341298
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010670136.9A Active CN111951221B (en) | 2020-07-13 | 2020-07-13 | Glomerular cell image recognition method based on deep neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111951221B (en) |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112233112A (en) * | 2020-12-07 | 2021-01-15 | 深圳大学 | Prediction method of CK19 of hepatocellular carcinoma based on deep learning analysis |
| CN112562855A (en) * | 2020-12-18 | 2021-03-26 | 深圳大学 | Hepatocellular carcinoma postoperative early recurrence risk prediction method |
| CN113192077A (en) * | 2021-04-15 | 2021-07-30 | 华中科技大学 | Automatic classification method and system for pathological graphs of cells and regional levels |
| CN113222944A (en) * | 2021-05-18 | 2021-08-06 | 湖南医药学院 | Cell nucleus segmentation method, system and device and cancer auxiliary analysis system and device based on pathological image |
| CN113256577A (en) * | 2021-05-18 | 2021-08-13 | 湖南医药学院 | Cancer auxiliary analysis system and device based on HE staining pathological image |
| CN113362294A (en) * | 2021-05-27 | 2021-09-07 | 同济大学 | Puncture needle identification method, system and equipment containing puncture needle ultrasonic blood vessel image |
| CN113409293A (en) * | 2021-06-30 | 2021-09-17 | 上海市第一人民医院 | Pathology image automatic segmentation system based on deep learning |
| CN113591919A (en) * | 2021-06-29 | 2021-11-02 | 复旦大学附属中山医院 | AI-based analysis method and system for prognosis of postoperative recurrence of early hepatocellular carcinoma |
| CN114219807A (en) * | 2022-02-22 | 2022-03-22 | 成都爱迦飞诗特科技有限公司 | Mammary gland ultrasonic examination image grading method, device, equipment and storage medium |
| CN114387201A (en) * | 2021-04-08 | 2022-04-22 | 透彻影像科技(南京)有限公司 | Cytopathic image auxiliary diagnosis system based on deep learning and reinforcement learning |
| CN114943723A (en) * | 2022-06-08 | 2022-08-26 | 北京大学口腔医学院 | Method for segmenting and counting irregular cells and related equipment |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20190147215A1 (en) * | 2017-11-16 | 2019-05-16 | General Electric Company | System and method for single channel whole cell segmentation |
| JP2019135939A (en) * | 2018-02-07 | 2019-08-22 | 澁谷工業株式会社 | Cell count method, cell count device and cell culture method |
| CN110151181A (en) * | 2019-04-16 | 2019-08-23 | 杭州电子科技大学 | Rapid magnetic resonance imaging method based on the U-shaped network of recurrence residual error |
| CN110427970A (en) * | 2019-07-05 | 2019-11-08 | 平安科技（深圳）有限公司 | Image classification method, device, computer equipment and storage medium |
| CN110490840A (en) * | 2019-07-11 | 2019-11-22 | 平安科技（深圳）有限公司 | A kind of cell detection method, device and the equipment of glomerulus pathology sectioning image |
| US20200097701A1 (en) * | 2016-11-16 | 2020-03-26 | Ventana Medical Systems, Inc. | Convolutional neural networks for locating objects of interest in images of biological samples |
| CN111080575A (en) * | 2019-11-22 | 2020-04-28 | 东南大学 | Thalamus segmentation method based on residual error dense U-shaped network model |
| WO2020093042A1 (en) * | 2018-11-02 | 2020-05-07 | Deep Lens, Inc. | Neural networks for biomedical image analysis |
| CN111275191A (en) * | 2020-02-26 | 2020-06-12 | 上海商汤智能科技有限公司 | Method and apparatus for detecting cell, electronic device, and storage medium |
| CN111340711A (en) * | 2020-05-21 | 2020-06-26 | 腾讯科技（深圳）有限公司 | Super-resolution reconstruction method, device, equipment and storage medium |
| US20200372648A1 (en) * | 2018-05-17 | 2020-11-26 | Tencent Technology (Shenzhen) Company Limited | Image processing method and device, computer apparatus, and storage medium |
-
2020
- 2020-07-13 CN CN202010670136.9A patent/CN111951221B/en active Active
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20200097701A1 (en) * | 2016-11-16 | 2020-03-26 | Ventana Medical Systems, Inc. | Convolutional neural networks for locating objects of interest in images of biological samples |
| US20190147215A1 (en) * | 2017-11-16 | 2019-05-16 | General Electric Company | System and method for single channel whole cell segmentation |
| JP2019135939A (en) * | 2018-02-07 | 2019-08-22 | 澁谷工業株式会社 | Cell count method, cell count device and cell culture method |
| US20200372648A1 (en) * | 2018-05-17 | 2020-11-26 | Tencent Technology (Shenzhen) Company Limited | Image processing method and device, computer apparatus, and storage medium |
| WO2020093042A1 (en) * | 2018-11-02 | 2020-05-07 | Deep Lens, Inc. | Neural networks for biomedical image analysis |
| CN110151181A (en) * | 2019-04-16 | 2019-08-23 | 杭州电子科技大学 | Rapid magnetic resonance imaging method based on the U-shaped network of recurrence residual error |
| CN110427970A (en) * | 2019-07-05 | 2019-11-08 | 平安科技（深圳）有限公司 | Image classification method, device, computer equipment and storage medium |
| CN110490840A (en) * | 2019-07-11 | 2019-11-22 | 平安科技（深圳）有限公司 | A kind of cell detection method, device and the equipment of glomerulus pathology sectioning image |
| CN111080575A (en) * | 2019-11-22 | 2020-04-28 | 东南大学 | Thalamus segmentation method based on residual error dense U-shaped network model |
| CN111275191A (en) * | 2020-02-26 | 2020-06-12 | 上海商汤智能科技有限公司 | Method and apparatus for detecting cell, electronic device, and storage medium |
| CN111340711A (en) * | 2020-05-21 | 2020-06-26 | 腾讯科技（深圳）有限公司 | Super-resolution reconstruction method, device, equipment and storage medium |
Non-Patent Citations (2)
| Title |
|---|
| GLORIA BUENO 等: "Glomerulosclerosis identification in whole slide images using semantic segmentation", 《COMPUTER METHODS AND PROGRAMS IN BIOMEDICINE》, vol. 184, no. 105273, pages 1 - 10 * |
| 裘静韬 等: "应用化肝脏病灶分割的预测方案", 《计算机工程与应用》, vol. 56, no. 20, pages 173 - 178 * |
Cited By (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112233112B (en) * | 2020-12-07 | 2021-06-01 | 深圳大学 | Prediction method of CK19 of hepatocellular carcinoma based on deep learning analysis |
| CN112233112A (en) * | 2020-12-07 | 2021-01-15 | 深圳大学 | Prediction method of CK19 of hepatocellular carcinoma based on deep learning analysis |
| CN112562855B (en) * | 2020-12-18 | 2021-11-02 | 深圳大学 | Hepatocellular carcinoma postoperative early recurrence risk prediction method, medium and terminal equipment |
| CN112562855A (en) * | 2020-12-18 | 2021-03-26 | 深圳大学 | Hepatocellular carcinoma postoperative early recurrence risk prediction method |
| CN114387201A (en) * | 2021-04-08 | 2022-04-22 | 透彻影像科技(南京)有限公司 | Cytopathic image auxiliary diagnosis system based on deep learning and reinforcement learning |
| CN113192077A (en) * | 2021-04-15 | 2021-07-30 | 华中科技大学 | Automatic classification method and system for pathological graphs of cells and regional levels |
| CN113192077B (en) * | 2021-04-15 | 2022-08-02 | 华中科技大学 | Automatic classification method and system for pathological graphs of cells and regional levels |
| CN113256577B (en) * | 2021-05-18 | 2022-06-28 | 湖南医药学院 | Cancer auxiliary analysis system and device based on HE staining pathological image |
| CN113256577A (en) * | 2021-05-18 | 2021-08-13 | 湖南医药学院 | Cancer auxiliary analysis system and device based on HE staining pathological image |
| CN113222944A (en) * | 2021-05-18 | 2021-08-06 | 湖南医药学院 | Cell nucleus segmentation method, system and device and cancer auxiliary analysis system and device based on pathological image |
| CN113362294A (en) * | 2021-05-27 | 2021-09-07 | 同济大学 | Puncture needle identification method, system and equipment containing puncture needle ultrasonic blood vessel image |
| CN113591919A (en) * | 2021-06-29 | 2021-11-02 | 复旦大学附属中山医院 | AI-based analysis method and system for prognosis of postoperative recurrence of early hepatocellular carcinoma |
| CN113591919B (en) * | 2021-06-29 | 2023-07-21 | 复旦大学附属中山医院 | Analysis method and system for prognosis of early hepatocellular carcinoma postoperative recurrence based on AI |
| CN113409293A (en) * | 2021-06-30 | 2021-09-17 | 上海市第一人民医院 | Pathology image automatic segmentation system based on deep learning |
| CN114219807A (en) * | 2022-02-22 | 2022-03-22 | 成都爱迦飞诗特科技有限公司 | Mammary gland ultrasonic examination image grading method, device, equipment and storage medium |
| CN114943723A (en) * | 2022-06-08 | 2022-08-26 | 北京大学口腔医学院 | Method for segmenting and counting irregular cells and related equipment |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111951221B (en) | 2023-10-31 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111951221B (en) | Glomerular cell image recognition method based on deep neural network | |
| CN106056595B (en) | Based on the pernicious assistant diagnosis system of depth convolutional neural networks automatic identification Benign Thyroid Nodules | |
| CN108898595B (en) | Construction method and application of positioning model of focus region in chest image | |
| CN109636808B (en) | Lung lobe segmentation method based on full convolution neural network | |
| CN112529894B (en) | Thyroid nodule diagnosis method based on deep learning network | |
| Liu et al. | A framework of wound segmentation based on deep convolutional networks | |
| CN110934606A (en) | Cerebral apoplexy early-stage flat-scan CT image evaluation system and method and readable storage medium | |
| CN109035283B (en) | Emphysema accurate detection and quantitative analysis method based on randomly selected partitions | |
| CN110974306B (en) | System for discernment and location pancreas neuroendocrine tumour under ultrasonic endoscope | |
| CN111402254A (en) | CT image pulmonary nodule high-performance automatic detection method and device | |
| CN104318565B (en) | Interactive method for retinal vessel segmentation based on bidirectional region growing of constant-gradient distance | |
| CN112085113B (en) | Severe tumor image recognition system and method | |
| CN112102332A (en) | Cancer WSI segmentation method based on local classification neural network | |
| CN115546605A (en) | Training method and device based on image labeling and segmentation model | |
| CN115206495A (en) | Renal cancer pathological image analysis method and system based on CoAtNet deep learning and intelligent microscopic device | |
| CN106960199A (en) | A kind of RGB eye is as the complete extraction method in figure white of the eye region | |
| EP4118617A1 (en) | Automated detection of tumors based on image processing | |
| CN115880266B (en) | Intestinal polyp detection system and method based on deep learning | |
| Xu et al. | RUnT: A network combining residual U-Net and transformer for vertebral edge feature fusion constrained spine CT image segmentation | |
| Zhou et al. | Computer aided diagnosis for diabetic retinopathy based on fundus image | |
| CN112348796B (en) | Cerebral hemorrhage segmentation method and system based on multi-model combination | |
| CN111210436B (en) | Lens segmentation method, device and storage medium | |
| Wen et al. | A novel lesion segmentation algorithm based on U-net network for tuberculosis CT image | |
| CN113838020A (en) | Molybdenum target image-based lesion area quantification method | |
| CN112819042A (en) | Method, system and medium for processing esophageal squamous dysplasia image |
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 |