Method and device for intelligently analyzing semen and storage medium
Through deep learning model, the sperm images in microscope videos are analyzed, which solves the problem that traditional semen vitality detection methods rely on artificial operations and high cost, and realizes efficient semen detection without fixed volume slides, improving the convenience and accuracy of detection sex.
Patent Information
- Application Number
- CN202410948008.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-16
- Publication Date
- 2024-11-12
AI Technical Summary
Traditional semen vitality detection methods rely on artificial operations and are costly, and the rule-based image processing algorithms are limited in detection accuracy when sperm dense or fast motion, and cannot meet the needs of large-scale detection.
The deep learning model is used to analyze the sperm images in microscope videos. Through pre-processing, feature value extraction, tracking trajectory and index calculation, the semen analysis results are obtained, reducing dependence on fixed volume slides, and improving detection convenience and efficiency.
The convenience and efficiency of semen detection are improved without the need for fixed volume slides. By optimizing training data acquisition and trajectory analysis, the model accuracy and generalization are ensured, and the semen quality analysis is provided at a lower cost.
Smart Images

Figure CN118941512A_ABST
Abstract
Description
Technical Field
[0001] The present disclosure generally relates to the field of semen motility detection technology. More specifically, the present disclosure relates to a method, device and storage medium for intelligent analysis of semen. Background Art
[0002] The detection of semen motility of male animals is an important means to evaluate the reproductive capacity of male animals in the breeding industry. Semen motility directly affects the success rate of insemination and reproductive efficiency. Therefore, accurate and rapid evaluation of semen motility is of great significance to the breeding industry.
[0003] Traditional methods for detecting semen motility mainly include manual microscope analysis and computer-assisted sperm analysis (CASA). However, in the process of manual microscope analysis, it relies on the experience and skills of the operator and is easily affected by human factors, resulting in subjective and inconsistent results. In addition, the operator is required to observe and count sperm one by one, which is a cumbersome and time-consuming process and cannot meet the needs of large-scale testing. Although the objectivity and efficiency of the test are improved in the process of using the computer-assisted sperm analysis system, its equipment cost is high, the cost of each test is high, and the maintenance is complex. In addition, the traditional CASA system mostly uses rule-based image processing algorithms, which have limited processing capabilities for the complexity of sperm morphology and motion trajectory, especially in the case of dense or overlapping sperm, which affects the detection accuracy. At a time when deep learning technology is constantly expanding its application boundaries, the precise use of deep computer vision technology to support sperm kinematic analysis and optimize its efficiency and effectiveness has become a key technical challenge we face.
[0004] Most of the existing deep learning-based male animal semen motility detection algorithms are highly dependent on equipment. Their limitations are that they cannot handle dense and fast sperm movements well, and they need to use a fixed volume slide to reduce the sperm density under the microscope, which increases the cost of each test. At the same time, the algorithm has few functions and cannot guide operators to perform correct operations, which is inconvenient.
[0005] In view of this, there is an urgent need to provide a solution for intelligent analysis of semen so as to improve the convenience and efficiency of semen testing without relying on fixed volume slides. Summary of the invention
[0006] In order to at least solve one or more of the technical problems mentioned above, the present disclosure proposes a solution for intelligent analysis of semen in multiple aspects.
[0007] In a first aspect, the present disclosure provides a method for intelligent analysis of semen, comprising: collecting sperm images that meet characteristic value threshold conditions and sperm quantity threshold conditions to form a training data set; using the training data set to train a sperm prediction frame model; using the trained sperm prediction frame model to analyze multiple input images obtained from a preprocessed microscope video to output sperm prediction frames in each input image; based on the sperm prediction frame, obtaining a tracking trajectory of each sperm in the microscope video; based on the tracking trajectory, deleting sperm in the microscope video that does not meet the frame number threshold condition, and obtaining multiple indicator data corresponding to the remaining sperm; and based on the multiple indicator data corresponding to the remaining sperm, obtaining the analysis results of the semen in the microscope video.
[0008] In one embodiment, the eigenvalue threshold condition is: the eigenvalue of the sperm image is greater than the eigenvalue threshold, wherein the method further includes: converting the sperm image into a grayscale image; and extracting eigenvalues from the grayscale image to obtain the eigenvalue of the sperm image.
[0009] In another embodiment, the sperm number threshold condition is: the sperm number in the sperm image is greater than and / or equal to the sperm number threshold.
[0010] In another embodiment, obtaining the tracking trajectory of each sperm in the microscope video based on the sperm prediction frame includes: expanding the boundary of the sperm prediction frame; and inputting the sperm prediction frame with expanded boundary into the tracker to obtain the tracking trajectory of each sperm in the microscope video.
[0011] In another embodiment, deleting sperm in a microscope video that does not meet a frame number threshold condition based on the tracking trajectory includes: obtaining a trajectory coordinate point of each sperm in the microscope video according to the tracking trajectory; counting the total number of frames of each sperm in the microscope video based on the trajectory coordinate points; and deleting sperm in the microscope video that does not meet a frame number threshold condition, wherein the frame number threshold condition is that the total number of frames of the sperm in the microscope video is less than the frame number threshold.
[0012] In another embodiment, the multiple indicator data corresponding to the remaining sperms include: the total distance, total displacement, total time, total average path, curve rate, straight line rate, average path rate, forwardness, linearity, swing and weight of the remaining sperms in the microscope video.
[0013] In yet another embodiment, the analysis results of the semen include: semen motility rate, semen vitality value and the degree of inadequate warming of the semen.
[0014] In another embodiment, when obtaining the semen motility, the following steps are performed: screening out sperm that meet the curve velocity threshold condition from the remaining sperm; and determining the semen motility based on the ratio of the sum of their weights to the sum of the weights of the remaining sperm, wherein the curve velocity threshold condition is: the curve velocity of the sperm is greater than or equal to the curve velocity threshold.
[0015] In another embodiment, when obtaining the motility value of semen, the following steps are performed: screening out sperm from the remaining sperm that meet both the curve velocity threshold condition and the forward direction threshold condition; and determining the motility value of semen based on the ratio of the sum of their weights to the sum of the weights of the remaining sperm, wherein the curve velocity threshold condition is: the curve velocity of the sperm is greater than or equal to the curve velocity threshold, and the forward direction threshold condition is: the forward direction of the sperm is greater than or equal to the forward direction threshold.
[0016] In another embodiment, when obtaining the degree of inadequate warming of semen, the following steps are performed: obtaining the number of sperms among the remaining sperms that meet the linearity threshold condition; obtaining the ratio of the number of sperms to the total number of remaining sperms, wherein the linearity threshold condition is: the linearity of the sperms is less than the linearity threshold; and quantifying the ratio into an integer between -9 and 9 to obtain the degree of inadequate warming of semen.
[0017] In a second aspect, the present disclosure provides a device for intelligent analysis of semen, comprising: a processor and a memory; the memory is used to store at least executable instructions, and the executable instructions are executed by the processor according to the method described in the first aspect above.
[0018] In a third aspect, the present disclosure provides a computer-readable storage medium having stored thereon computer program instructions for performing intelligent analysis on semen, the instructions being executed by one or more processors to implement the method as described in the first aspect above.
[0019] Through the solution for intelligent analysis of semen provided as above, the disclosed embodiment can use the trained deep learning model to intelligently analyze the sperm movement in the microscope video without relying on expensive fixed volume slides, thereby greatly improving the convenience, automation and analysis efficiency of semen detection.
[0020] Furthermore, this solution adopts a number of optimization measures in the aspects of training data collection, trajectory analysis, and index calculation, which further enhances the technical effect. For example, by combining grayscale transformation and eigenvalue filtering to obtain high-quality training samples, and using the sperm recognition module to strictly control the number of sperm, the disclosed solution can ensure the accuracy and generalization of model training from the source. For another example, the sperm tracking effect is optimized through the expansion of the prior frame, and the reliability of trajectory analysis is further improved by combining the frame number threshold filtering. In addition, by screening the threshold conditions of indicators such as rate, forwardness, and linearity, combined with the weighting of each sperm weight, the overall vitality of semen can be more objectively and comprehensively evaluated, and the degree of inadequate rewarming of semen can be conveniently reflected through ratio quantification. Through the synergistic enhancement of the aforementioned many optimization details, this solution can achieve more intelligent and precise semen quality analysis at a lower cost. BRIEF DESCRIPTION OF THE DRAWINGS
[0021] By reading the detailed description below with reference to the accompanying drawings, the above and other objects, features and advantages of the exemplary embodiments of the present disclosure will become readily understood. In the accompanying drawings, several embodiments of the present disclosure are shown in an exemplary and non-limiting manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein:
[0022] Figure 1 An exemplary flow chart of a method for intelligently analyzing semen according to an embodiment of the present disclosure is shown;
[0023] Figure 2 An exemplary flowchart of a method for collecting sperm images that meet conditions according to an embodiment of the present disclosure is shown;
[0024] Figure 3 An exemplary structural block diagram of an apparatus for intelligently analyzing semen according to an embodiment of the present disclosure is shown. DETAILED DESCRIPTION
[0025] The following will be combined with the drawings in the embodiments of the present disclosure to clearly and completely describe the technical solutions in the embodiments of the present disclosure. Obviously, the described embodiments are part of the embodiments of the present disclosure, not all of the embodiments. Based on the embodiments in the present disclosure, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present disclosure.
[0026] It should be understood that the terms "include" and "comprising" used in the specification and claims of the present disclosure indicate the presence of described features, integers, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or collections thereof.
[0027] It should also be understood that the terms used in this disclosure are only for the purpose of describing specific embodiments and are not intended to limit the disclosure. As used in this disclosure and claims, the singular forms of "a", "an", and "the" are intended to include the plural forms unless the context clearly indicates otherwise. It should also be further understood that the term "and / or" used in this disclosure and claims refers to any combination of one or more of the associated listed items and all possible combinations, including these combinations.
[0028] As used in this specification and claims, the term "if" may be interpreted as "when" or "upon" or "in response to determining" or "in response to detecting," depending on the context. Similarly, the phrase "if it is determined" or "if [described condition or event] is detected" may be interpreted as meaning "upon determination" or "in response to determining" or "upon detection of [described condition or event]" or "in response to detecting [described condition or event]," depending on the context.
[0029] The specific implementation of the present disclosure is described in detail below with reference to the accompanying drawings.
[0030] Exemplary application scenarios
[0031] Deep learning is a subfield of machine learning that uses multi-layer neural networks to simulate the workings of the human brain and process and analyze complex data. Deep learning models are trained with large amounts of data to automatically extract features and perform high-precision classification and prediction. Common deep learning models include convolutional neural networks, recurrent neural networks, and generative adversarial networks. Deep learning has achieved remarkable results in areas such as image recognition, speech recognition, natural language processing, and autonomous driving.
[0032] In the prior art, the male animal semen motility detection algorithm based on deep learning first uses a fixed volume slide to reduce the sperm density under a microscope, and then performs a sperm motility test to directly obtain relevant sperm index data. However, the use of a fixed volume slide increases the cost of each test and reduces the detection efficiency. In addition, only the relevant sperm index data is directly obtained without further analysis of the index data, which makes it impossible for relevant operators to obtain more accurate analysis data and thus unable to guide the operators to perform correct operations.
[0033] Exemplary scheme for intelligent analysis of semen
[0034] In view of this, the disclosed embodiments provide a solution for intelligent analysis of semen, which uses a trained model to analyze sperm images in microscope videos, track sperm trajectories, and obtain semen analysis results. This can improve the convenience and efficiency of semen testing without relying on fixed volume slides.
[0035] Figure 1 FIG. 1 shows an exemplary flow chart of a method 100 for intelligently analyzing semen according to an embodiment of the present disclosure. Figure 1 As shown in, in step S110, sperm images that meet the characteristic value threshold condition and the sperm number threshold condition are collected to form a training data set. In one embodiment, the characteristic value threshold condition is: the characteristic value of the sperm image is greater than the characteristic value threshold, and the sperm number threshold condition is: the number of sperm in the sperm image is greater than and / or equal to the sperm number threshold. Specifically, the sperm image that forms the training data set must satisfy both that the characteristic value is greater than the characteristic value threshold and that the number of sperm is greater than and / or equal to the sperm number threshold. In some embodiments, the collected sperm image is converted into a grayscale image, and the characteristic value of the grayscale image is extracted to obtain the characteristic value of the aforementioned sperm image. The collected sperm image is input into a sperm recognition module to obtain the number of sperm in the image. Specifically, the sperm recognition module can use a convolutional neural network for sperm recognition. By collecting sperm images that meet the characteristic value threshold condition and the sperm number threshold condition to form a training data set, it is possible to ensure that the image quality in the training data set is high, avoid model training errors caused by poor image quality or insufficient sperm number, and thus improve the accuracy and reliability of the model of subsequent training.
[0036] After forming the training data set, in step S120, the sperm prediction box model is trained using the training data set. In one embodiment, the object category in the image of the training data set is defined as sperm, each image in the training data set is annotated, and the sperm prediction box model is trained using the annotated training data set. In one embodiment of the present invention, the sperm prediction box model uses the YOLOv5 pre-trained model. YOLOv5 is the 5th version of YOLO, which is a real-time target detection system that uses a convolutional neural network to directly predict bounding boxes and category probabilities in an image. Unlike traditional target detection algorithms, YOLO significantly improves detection speed and efficiency by inputting the entire image into the network at one time. At the same time, the YOLO model divides the image into grids and predicts multiple bounding boxes and corresponding category probabilities in each grid, which can achieve real-time detection while maintaining high accuracy, and is widely used in video surveillance, autonomous driving, robot vision and other fields. By adopting the YOLOv5 pre-trained model, the hardware requirements for the equipment and the operating requirements for the operators are reduced. The device side supports the use of common slides and glass rods instead of fixed volume slides and pipettes, which greatly reduces the operation complexity and usage cost.
[0037] In another embodiment of the present invention, the sperm prediction box model may also adopt other versions of the YOLO series models, FastRCNN, semantic segmentation models (e.g., U-Net), etc.
[0038] In step S130, the trained sperm prediction frame model is used to analyze multiple input images obtained from the preprocessed microscope video to output the sperm prediction frame in each input image. In one embodiment, OpenCV is used to read the microscope video and obtain each frame image corresponding to the microscope video. Each frame image obtained is then converted into RGB mode. Then, the converted image is normalized by dividing it by 255.0 using OpenCV. Finally, OpenCV is used to scale the normalized image to the size of the input image required by the sperm prediction frame model and input it into the trained sperm prediction frame model.
[0039] Specifically, in the process of using the trained sperm prediction box model to output the sperm prediction box in each input image, the sperm prediction box model will perform a series of processing and analysis for each input image. Specifically, the trained sperm prediction box model is used to identify the sperm that may exist in the image and calculate the probability value of the specific position of these sperm in the image. At the same time, the model will also generate a prediction box for each possible sperm position, which is used to mark the approximate area of the sperm in the image. Then, the sperm prediction box model will compare all the calculated probability values and select the one with the highest probability. The prediction box corresponding to the highest probability value will be regarded as the most likely location for sperm in the image. Finally, the prediction box with the highest probability is selected as the final output. In other words, the sperm prediction box model outputs the most likely position of sperm in the image.
[0040] In one embodiment of the present invention, a microscope video of a certain length (e.g., 3 seconds) is recorded and uploaded to the cloud through the http protocol. Thus, the cloud server uses OpenCV to read the microscope video. Specifically, since the microscope video is to be used for subsequent analysis, there are basic requirements for the quality of the microscope video. The semen under the lens should be correctly focused, without flow, with few impurities, and the sperm should be clear.
[0041] After outputting the sperm prediction frame in each input image, in step S140, the tracking track of each sperm in the microscope video is obtained based on the sperm prediction frame. Specifically, the boundary of the sperm prediction frame is first expanded, and then the sperm prediction frame with the expanded boundary is input into the tracker to obtain the tracking track of each sperm in the microscope video. In one embodiment, in the process of expanding the boundary of the sperm prediction frame, the sperm prediction frame is increased by 30 pixels in the upper, lower, left and right directions, thereby increasing the tracking range.
[0042] In one embodiment, the aforementioned tracker may adopt a BoT-SORT tracker or a ByteTrack (Tracking by Detection) tracker. The BoT-SORT tracker adopts the BoT-SORT multi-target tracking method. The BoT-SORT multi-target tracking method manages the state of the target (for example, appearance, disappearance, and reappearance, etc.) through byte-level tracking, thereby improving the robustness and continuity of tracking. The method uses a Kalman filter to predict and update the motion state of the target, making the tracking process smoother and more accurate. At the same time, the BoT-SORT multi-target tracking method adopts the Hungarian algorithm to solve the association problem between the tracking target and the detection result. Specifically, the Hungarian algorithm achieves efficient target matching by minimizing the association cost function. The BoT-SORT multi-target tracking method uses a dual-branch network to extract the appearance features and motion features of the target matched by the Hungarian algorithm, and then uses these features to track the target.
[0043] After obtaining the tracking tracks of each sperm in the microscope video, in step S150, based on the tracking tracks, the sperm that does not meet the frame number threshold condition in the microscope video is deleted, and multiple indicator data corresponding to the remaining sperm are obtained. Specifically, the track coordinate points of each sperm in the microscope video are obtained according to the tracking tracks. Then, the total number of frames of each sperm in the microscope video is counted based on the track coordinate points. Then, the sperm that does not meet the frame number threshold condition in the microscope video is deleted.
[0044] In one embodiment, the frame number threshold condition is that the total frame number of sperm in the microscope video is less than the frame number threshold. By deleting sperm in the microscope video that do not meet the frame number threshold condition, sperm that move too fast or too slow can be deleted, reducing mismatches caused by frame number issues, ensuring that sperm data used for subsequent analysis and evaluation is more accurate and reliable, and reducing the workload of subsequent processing and analysis.
[0045] In one embodiment, the multiple indicator data corresponding to the remaining sperms include: the total distance, total displacement, total time, total average path, curvilinear velocity (VCL), straight-line / rectilinear velocity (VSL), average path velocity (VAP), straightness (STR), linearity (LIN), wobble (WOB) and weight of the remaining sperms in the microscope video.
[0046] Specifically, the curve rate is the average rate of the sperm head along its actual curve, that is, the two-dimensional motion trajectory under the microscope, reflecting the motility of sperm. The linear rate is the average rate of the linear motion of the sperm head between the position at the beginning of the test and the final position. The average path rate is the average rate of the sperm head moving along its average path, which is calculated by smoothing the actual curve trajectory according to the algorithm of the CASA instrument. Propulsion is the linearity of the average path, and its calculation formula is: STR = VSL / VAP, where STR is propulsion, VSL is the linear rate, and VAP is the average path rate. Linearity is the linearity of the curve path, and its calculation formula is: LIN = VSL / VCL, where LIN is linearity, VSL is the linear rate, and VCL is the curve rate. Wobble is the swing value of the actual curve path about the average path, and its calculation formula is: WOB = VAP / VCL, where WOB is wobble, VAP is the average path rate, and VCL is the curve rate. The weight is the number of frames in which each remaining sperm exists in the video divided by the total number of frames in which the remaining sperm exist in the video.
[0047] By deleting sperm that do not meet the frame number threshold conditions in the microscope video and obtaining the weights of the remaining sperm in the microscope video, the fault tolerance of sperm images is further improved, making the results more objective and accurate, and realizing the automation, efficiency, accuracy and low cost of sperm intelligent analysis.
[0048] In one embodiment of the present invention, the aforementioned frame number threshold is 5 frames.
[0049] After obtaining the multiple indicator data corresponding to the remaining sperm, in step S160, the analysis results of the semen in the microscope video are obtained based on the multiple indicator data corresponding to the remaining sperm. In one embodiment of the present invention, the analysis results of the semen include: semen motility, semen vitality value and the degree of inadequate rewarming of the semen. By further analyzing the multiple indicator data corresponding to the remaining sperm, it can be more accurately determined whether the quality and quantity of the sperm of the male animal meet the standards.
[0050] Specifically, in the process of obtaining the semen motility, first, the sperms that meet the curve velocity threshold condition are screened out from the remaining sperms, and then the semen motility is determined according to the ratio of the sum of their weights to the sum of the weights of the remaining sperms. More specifically, the curve velocity threshold condition is: the curve velocity of the sperm is greater than or equal to the curve velocity threshold.
[0051] In one embodiment of the present invention, the aforementioned curve rate threshold is set to 10. In actual application, the user may adjust the aforementioned curve rate threshold according to specific scenarios or requirements.
[0052] Specifically, in the process of obtaining the vitality value of semen, first, the sperms that meet both the curve velocity threshold condition and the forwardness threshold condition are screened out from the remaining sperms, and then the vitality value of semen is determined according to the ratio of the sum of their weights to the sum of the weights of the remaining sperms. More specifically, the curve velocity threshold condition is: the curve velocity of the sperm is greater than or equal to the curve velocity threshold, and the forwardness threshold condition is: the forwardness of the sperm is greater than or equal to the forwardness threshold.
[0053] In one embodiment of the present invention, the aforementioned curve rate threshold is set to 15, and the forwardness threshold is set to 20%. In actual application, the user can adjust the aforementioned curve rate threshold and forwardness threshold according to specific scenarios or requirements.
[0054] Specifically, in the process of obtaining the degree of inadequate warming of semen, first, the number of sperms in the remaining sperms that meet the linearity threshold condition is obtained, then the ratio of the number of sperms to the total number of remaining sperms is obtained, and finally, the ratio is quantified into an integer between -9 and 9 to obtain the degree of inadequate warming of semen. More specifically, the linearity threshold condition is: the linearity of the sperm is less than the linearity threshold.
[0055] In one embodiment of the present invention, the linearity threshold is set to 0.15.
[0056] In one embodiment of the present invention, in the process of quantizing the ratio to an integer between -9 and 9, first, the ratio is multiplied by 100 and then divided by 5 to obtain a first value, and then, the integer part of the first value is taken, and 4 is subtracted from the integer part to obtain an integer distributed between -9 and 9.
[0057] In summary, through the solution for intelligent analysis of semen provided above, the disclosed embodiment uses a trained model to analyze sperm images in microscope videos, track sperm trajectories, and obtain semen analysis results, thereby improving the convenience and efficiency of semen testing without relying on fixed volume slides.
[0058] Furthermore, this solution adopts a number of optimization measures in the aspects of training data collection, trajectory analysis, and index calculation, which further enhances the technical effect. For example, by combining grayscale transformation and eigenvalue filtering to obtain high-quality training samples, and using the sperm recognition module to strictly control the number of sperm, the disclosed solution can ensure the accuracy and generalization of model training from the source. For another example, the sperm tracking effect is optimized through the expansion of the prior frame, and the reliability of trajectory analysis is further improved by combining the frame number threshold filtering. In addition, by screening the threshold conditions of indicators such as rate, forwardness, and linearity, combined with the weighting of each sperm weight, the overall vitality of semen can be more objectively and comprehensively evaluated, and the degree of inadequate rewarming of semen can be conveniently reflected through ratio quantification. Through the synergistic enhancement of the aforementioned many optimization details, this solution can achieve more intelligent and precise semen quality analysis at a lower cost.
[0059] Figure 2 FIG. 2 shows an exemplary flow chart of a method 200 for collecting sperm images that meet the conditions according to an embodiment of the present disclosure. Figure 2 As shown in FIG. 2 , in step S210 , a sperm image is collected.
[0060] Then, in step S220, the sperm image currently collected is converted from an RGB image into a grayscale image. Specifically, the sperm image can be converted into a grayscale image by OpenCV, and the sperm image can also be converted into a grayscale image by a weighted average method, a maximum method, a minimum method or an average method. OpenCV is an open source cross-platform computer vision and machine learning software library that provides a large number of computer vision and machine learning algorithms. In terms of machine learning, OpenCV also provides many functions, such as machine learning classifiers (such as SVM, KNN, decision trees, etc.), feature extraction (such as SIFT, SURF, ORB, etc.) and clustering algorithms, etc. These functions make OpenCV useful for processing image and video data.
[0061] Next, in step S230, feature value extraction is performed on the converted grayscale image.
[0062] Then, in step S240, it is determined whether the characteristic value of the grayscale image is greater than the characteristic value threshold. When the characteristic value of the grayscale image is not greater than the characteristic value threshold, the process returns to step S210 and the sperm image is collected again until the characteristic value of the grayscale image converted from the collected sperm image is greater than the characteristic value threshold.
[0063] When the eigenvalue of the grayscale image is greater than the eigenvalue threshold, in step S250, the sperm image is uploaded to the cloud, and the cloud calls a sperm recognition module to detect the number of sperm in the sperm image.
[0064] In step S260, it is determined whether the number of sperm in the sperm image is greater than or equal to the sperm number threshold. When the number of sperm in the sperm image is less than the sperm number threshold, the process returns to step S210 and the sperm image is re-collected until the eigenvalue of the grayscale image converted from the collected sperm image is greater than the eigenvalue threshold and the number of sperm in the collected sperm image is greater than or equal to the sperm number threshold.
[0065] When the number of sperm in the sperm image is greater than or equal to the sperm number threshold, in step S270, the sperm image is added to the training data set.
[0066] Figure 3 FIG. 2 shows an exemplary structural block diagram of a device 200 for intelligently analyzing semen according to an embodiment of the present disclosure. Figure 3 As shown, the device 200 may include a processor 210 and a memory 220, the aforementioned memory 220 may include but is not limited to a resistive random access memory RRAM (Resistive Random Access Memory), a dynamic random access memory DRAM (Dynamic Random Access Memory), a static random access memory SRAM (Static Random Access Memory), an enhanced dynamic random access memory EDRAM (Enhanced Dynamic Random Access Memory), the aforementioned processor 210 may include, for example, a general-purpose processor ("CPU"), a dedicated graphics processor ("GPU") or a microcontroller ("MCU"), and the aforementioned memory 220 stores at least executable instructions, and the executable instructions are executed by the aforementioned processor 21 as described above for the method for intelligent analysis of semen. As the method has been described in detail in the previous text, it will not be repeated here.
[0067] In another aspect of the present invention, a computer storage medium is provided, on which computer program instructions for intelligent analysis of semen are stored, and the instructions are executed by one or more processors to implement the method for intelligent analysis of semen as described above. Specifically, the aforementioned readable storage medium can be any tangible medium containing or storing a program, which can be used by an instruction execution system, device or device or used in combination with it. For example, the computer readable storage medium can be any appropriate magnetic storage medium or magneto-optical storage medium, such as resistive random access memory RRAM (Resistive Random Access Memory), dynamic random access memory DRAM (Dynamic Random Access Memory), static random access memory SRAM (Static Random-Access Memory), enhanced dynamic random access memory EDRAM (Enhanced Dynamic Random Access Memory), high-bandwidth memory HBM (High-Bandwidth Memory), hybrid memory cube HMC (Hybrid Memory Cube), etc., or any other medium that can be used to store the required information and can be accessed by an application, a module or both. Any such computer storage medium can be part of a device or accessible or connectable to a device. Any application or module described herein may be implemented using computer-readable / executable instructions that may be stored or otherwise maintained by such computer-readable media.
[0068] Although multiple embodiments of the present disclosure have been shown and described herein, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Those skilled in the art may think of many changes, modifications, and alternatives without departing from the thought and spirit of the present disclosure. It should be understood that in the process of practicing the present disclosure, various alternatives to the embodiments of the present disclosure described herein may be adopted. The attached claims are intended to define the scope of protection of the present disclosure, and therefore cover equivalents or alternatives within the scope of these claims.
[0069] The collection and acquisition of various data in this application complies with relevant laws and regulations and is authorized by the data provider. Any organization or individual that needs to obtain external data must obtain authorization in accordance with the law and ensure data security. It is not allowed to illegally collect, use, process, or transmit unauthorized or unprotected data, or to illegally buy, sell, provide, or disclose unauthorized or unprotected data.
Claims
1. A method for intelligent analysis of semen, characterized in that: include: Collect sperm images that meet the characteristic value threshold conditions and sperm quantity threshold conditions to form a training data set; Using the training data set to train the sperm prediction frame model; Using the trained sperm prediction box model, multiple input images obtained from the preprocessed microscope video are analyzed to output sperm prediction boxes in each input image; Obtain the tracking trajectory of each sperm in the microscope video based on the sperm prediction frame; Based on the tracking trajectory, sperm that do not meet the frame number threshold condition in the microscope video are deleted, and multiple indicator data corresponding to the remaining sperm are obtained; as well as Based on the multiple indicator data corresponding to the remaining sperms, the analysis results of the semen in the microscope video are obtained.
2. The method according to claim 1, characterized in that The characteristic value threshold condition is: the characteristic value of the sperm image is greater than the characteristic value threshold, wherein the method further comprises: converting the sperm image into a grayscale image; and The feature value of the grayscale image is extracted to obtain the feature value of the sperm image.
3. The method according to claim 1, characterized in that The sperm count threshold condition is: the sperm count in the sperm image is greater than and / or equal to the sperm count threshold.
4. The method according to claim 1, characterized in that: Obtaining the tracking trajectory of each sperm in the microscope video based on the sperm prediction frame includes: Expanding the boundaries of the sperm prediction box; and The sperm prediction box with expanded boundaries is input into the tracker to obtain the tracking trajectory of each sperm in the microscope video.
5. The method according to claim 1, characterized in that Sperm that do not meet the frame number threshold conditions in the microscope video are deleted based on the tracking trajectory, including: According to the tracking trajectory, the trajectory coordinate points of each sperm in the microscope video are obtained; Counting the total number of frames of each sperm in the microscope video based on the trajectory coordinate points; and Spermatozoa that do not meet a frame number threshold condition in the microscope video are deleted, wherein the frame number threshold condition is that the total frame number of the spermatozoa in the microscope video is less than the frame number threshold.
6. The method according to claim 1, characterized in that The multiple indicator data corresponding to the remaining sperms include: the total distance, total displacement, total time, total average path, curve rate, straight line rate, average path rate, forwardness, straightness, swing and weight of the remaining sperms in the microscope video.
7. The method according to claim 6, characterized in that The analysis results of semen include: semen viability, semen vitality value and the degree of inadequate warming of semen.
8. The method according to claim 7, characterized in that To obtain semen motility, perform the following steps: Screening out sperms that meet the curve velocity threshold condition from the remaining sperms; and The semen motility is determined according to the ratio of the sum of its weights to the sum of the weights of the remaining sperms, wherein the curve velocity threshold condition is: the curve velocity of the sperm is greater than or equal to the curve velocity threshold.
9. The method according to claim 7, characterized in that: To obtain the motility value of semen, perform the following steps: Sperms that meet both the curve velocity threshold condition and the forwardness threshold condition are selected from the remaining sperms; as well as The motility value of semen is determined according to the ratio of the sum of its weights to the sum of the weights of the remaining sperm, wherein the curve velocity threshold condition is: the curve velocity of the sperm is greater than or equal to the curve velocity threshold, and the forwardness threshold condition is: the forwardness of the sperm is greater than or equal to the forwardness threshold.
10. The method according to claim 7, characterized in that When the semen is not warmed up enough, perform the following steps: Obtain the number of sperms that meet the linearity threshold condition among the remaining sperms; The ratio of the total number of sperm to the total number of remaining sperm is obtained, wherein the linearity threshold condition is: the linearity of the sperm is less than the linearity threshold; and The ratio is quantified as an integer between -9 and 9 to obtain the degree of inadequate warming of the semen.
11. A device for intelligent analysis of semen, comprising: Processor and memory; The memory is used to store at least executable instructions, and the executable instructions are executed by the processor to perform the method according to any one of claims 1 to 10.
12. A computer-readable storage medium having stored thereon computer program instructions for intelligent analysis of semen, characterized in that: The instruction is executed by one or more processors to implement the method according to any one of claims 1 to 10.
Citation Information
Patent Citations
Sperm recognition and multi-target trajectory tracking method
CN112580476A
Semen automatic detection method and system
CN115184243A
Computer-assisted sperm analysis method for animals
CN115452826A
Multi-target tracking method and device
CN116012418A
Image processing method for automatic identification and classification of stem cells
CN118230319A
