An autonomous scanning method and device for human organs based on bayesian optimal search

Abstract

The application discloses a human organ autonomous scanning method and device based on Bayesian optimal search, and the method comprises the following steps: training an initial agent model according to an initial probe position data set and an initial ultrasound image score data set, and calculating the iteration number and the number of artificial fish of an artificial fish swarm algorithm according to the data volume of the data set; the next probe position is found by using the artificial fish swarm algorithm; the probe of the ultrasound robot is controlled to move to the next probe position, and a new ultrasound image is collected in real time to obtain an updated probe position data set and an updated ultrasound image score data set; the initial agent model is updated based on the updated probe position data set and the updated ultrasound image score data set to obtain an updated agent model; the process of the Bayesian optimal search is re-executed until a preset stop condition is reached to obtain an optimal probe position and a corresponding optimal ultrasound image. The method significantly improves the quality and efficiency of the autonomous scanning of the ultrasound robot.

Description

Technical Field

[0001] This invention relates to a method and apparatus for autonomous scanning of human organs based on Bayesian optimal search. Background Technology

[0002] With the advancement of medical technology, ultrasound examination has been widely used in clinical diagnosis due to its non-invasive, real-time, and cost-effective characteristics. However, traditional ultrasound examination mainly relies on manual operation by doctors, which not only requires doctors to have rich experience and superb skills, but also makes the examination results highly susceptible to the influence of the doctor's individual skill level, resulting in poor consistency and repeatability of the results. To overcome these limitations and improve the accuracy and efficiency of ultrasound examination, ultrasound robotic technology has gradually become a research hotspot. This technology, through automation and intelligent methods, can not only reduce the workload of doctors, but also significantly improve the stability and reliability of examination results, thereby promoting the development of ultrasound examination towards a more precise and efficient direction. Summary of the Invention

[0003] To obtain the optimal ultrasound image of the target organ, this invention provides a method and apparatus for autonomous scanning of human organs based on Bayesian optimal search.

[0004] In a first aspect, embodiments of the present invention provide a method for autonomous scanning of human organs based on Bayesian optimal search, which may include:

[0005] The initial surrogate model is trained based on the acquired initial probe pose dataset and initial ultrasound image scoring dataset;

[0006] Based on the amount of data in the initial ultrasound image scoring dataset, the number of iterations and the number of artificial fish in the artificial fish swarm algorithm are calculated.

[0007] Starting from the probe's initial pose, based on the initial proxy model, the preset acquisition function, the number of iterations of the artificial fish swarm algorithm, and the number of artificial fish, the artificial fish swarm algorithm is used to find the next probe pose;

[0008] The probe of the ultrasound robot is controlled to move to the next probe pose and new ultrasound images are acquired in real time. The initial probe pose dataset and the initial ultrasound image score dataset are updated based on the new ultrasound images to obtain the updated probe pose dataset and the updated ultrasound image score dataset.

[0009] The initial proxy model is updated based on the updated probe pose dataset and the updated ultrasound image scoring dataset to obtain the updated proxy model.

[0010] The Bayesian optimal search process is repeated until the preset stopping condition is met, the optimal probe pose is obtained, and the probe is controlled to acquire the optimal ultrasound image of the target organ at the optimal probe pose.

[0011] In one or more optional embodiments of this application, calculating the number of iterations and the number of artificial fish for the artificial fish swarm algorithm based on the data volume of the initial ultrasound image scoring dataset includes:

[0012] Based on the data volume of the initial ultrasound image scoring dataset, the number of iterations for the artificial fish swarm algorithm is calculated using the following formula:

[0013]

[0014] In the formula, G is the number of iterations of the artificial fish swarm algorithm, G max G is the preset maximum number of iterations. base S is the preset iteration base. p S represents the data size of the initial ultrasound image scoring dataset. max To preset the maximum data volume, It is a rounding function;

[0015] Based on the data volume of the initial ultrasound image scoring dataset, the number of artificial fish in the artificial fish swarm algorithm is calculated using the following formula:

[0016]

[0017] In the formula, F represents the number of artificial fish in the artificial fish swarm algorithm. max F is the preset maximum number of artificial fish to be added. base S is the preset artificial fish population. p S represents the data size of the initial ultrasound image scoring dataset. max To preset the maximum data volume, This is the floor function.

[0018] In one or more optional embodiments of this application, the probe of the controlled ultrasound robot moves to the next probe pose and acquires new ultrasound images in real time. The initial probe pose dataset and the initial ultrasound image scoring dataset are then updated based on the new ultrasound images to obtain an updated probe pose dataset and an updated ultrasound image scoring dataset, including:

[0019] The probe of the ultrasound robot is controlled to move to the next probe pose, and multiple new ultrasound images and the probe pose corresponding to each new ultrasound image are acquired in real time, and the score corresponding to each new ultrasound image is determined.

[0020] The probe pose and score corresponding to the multiple new ultrasound images are added to the initial probe pose dataset and the initial ultrasound image score dataset, respectively. At the same time, sparse downsampling is performed to obtain the updated probe pose dataset and the updated ultrasound image score dataset.

[0021] In one or more optional embodiments of this application, the updated probe pose dataset, the updated ultrasound image scoring dataset, and multiple new ultrasound images are sorted from oldest to newest according to their acquisition time.

[0022] The step of adding the probe pose and score corresponding to the multiple new ultrasound images to the initial probe pose dataset and the initial ultrasound image score dataset, respectively, and simultaneously performing sparse downsampling to obtain the updated probe pose dataset and the updated ultrasound image score dataset, includes:

[0023] Calculate the L2 norm of the probe pose corresponding to the last probe pose and the first new ultrasound image in the initial probe pose dataset, and determine whether the L2 norm is greater than the preset sampling interval.

[0024] If so, the probe pose corresponding to the first new ultrasound image is added to the initial probe pose dataset, and the score corresponding to the first new ultrasound image is added to the initial ultrasound image score dataset.

[0025] If not, the process of calculating the L2 norm is repeated based on subsequent new ultrasound images until no further new ultrasound images are available, resulting in an updated probe pose dataset and an updated ultrasound image scoring dataset.

[0026] In one or more optional embodiments of this application, the data in the updated probe pose dataset and the updated ultrasound image scoring dataset are sorted by acquisition time.

[0027] After obtaining the updated probe pose dataset and the updated ultrasound image scoring dataset, the following is also included:

[0028] Retain the maximum number of probe poses and scores that are sorted last in the updated probe pose dataset and the updated ultrasound image scoring dataset, and delete the other data.

[0029] In one or more optional embodiments of this application, the step of finding the next probe pose using an artificial fish swarm algorithm, starting from the probe's initial pose and based on the initial surrogate model, a preset acquisition function, the number of iterations of the artificial fish swarm algorithm, and the number of artificial fish, includes:

[0030] Multiple artificial fish are initialized based on the probe's initial pose and the number of artificial fish; wherein each artificial fish corresponds to a pose;

[0031] Based on the initial agent model and the acquisition function, calculate the sampling potential value for each artificial fish;

[0032] All artificial fish are controlled to perform swarming, tail-chasing, foraging, and random behaviors in the artificial fish swarm algorithm based on their corresponding sampling potential values. All artificial fish are iteratively updated until the number of iterations of the artificial fish swarm algorithm is reached, resulting in multiple updated artificial fish.

[0033] Based on the initial agent model and the acquisition function, calculate the sampling potential value for each of the updated artificial fish;

[0034] The pose of the artificial fish with the highest update sampling potential value is used as the pose of the next probe.

[0035] Secondly, embodiments of the present invention provide an autonomous human organ scanning device based on Bayesian optimal search, which may include:

[0036] The first training module is used to train the initial surrogate model based on the acquired initial probe pose dataset and initial ultrasound image scoring dataset.

[0037] The first calculation module is used to calculate the number of iterations of the artificial fish swarm algorithm and the number of artificial fish based on the amount of data in the initial ultrasound image scoring dataset.

[0038] The first optimization module is used to find the next probe pose starting from the probe's initial pose, based on the initial proxy model, the preset acquisition function, the number of iterations of the artificial fish swarm algorithm, and the number of artificial fish;

[0039] The first update module is used to control the probe of the ultrasound robot to move to the next probe pose, and to acquire new ultrasound images in real time. The module updates the initial probe pose dataset and the initial ultrasound image score dataset based on the new ultrasound images to obtain the updated probe pose dataset and the updated ultrasound image score dataset.

[0040] The second update module is used to update the initial proxy model based on the updated probe pose dataset and the updated ultrasound image scoring dataset to obtain the updated proxy model.

[0041] The first judgment module is used to determine whether the preset stopping condition has been met: if yes, the optimal probe pose is obtained, and the probe is controlled to acquire the optimal ultrasound image of the target organ at the optimal probe pose; if no, the first calculation module is executed again.

[0042] Thirdly, embodiments of the present invention provide a computer-readable storage medium storing a computer program / instruction thereon, which, when executed by a processor, implements the above-described method for autonomous scanning of human organs based on Bayesian optimal search.

[0043] Fourthly, embodiments of the present invention provide a computer program product, including a computer program / instruction, which, when executed by a processor, implements the above-described method for autonomous scanning of human organs based on Bayesian optimal search.

[0044] Fifthly, embodiments of the present invention provide a computer device, including a memory, a processor, and a computer program stored in the memory. When the processor executes the computer program, it implements the above-described method for autonomous scanning of human organs based on Bayesian optimal search.

[0045] The beneficial effects of the above-described technical solutions provided in the embodiments of the present invention include at least the following:

[0046] This invention provides a method for autonomous scanning of human organs based on Bayesian optimal search. The method trains an initial surrogate model and, based on the data volume of the initial ultrasound image scoring dataset, calculates the number of iterations and the number of fish in the Artificial Fish Swarm Algorithm (AFSA). The AFSA is used as an optimization algorithm to find the optimal next probe pose. The ultrasound probe is then controlled to move to the next probe pose, acquiring new ultrasound images during the movement to update the dataset and surrogate model. This process of Bayesian optimal search is repeated until a stopping condition is met, resulting in the optimal probe pose. The probe is then controlled to acquire the optimal ultrasound image of the target organ under this optimal pose. This method establishes a surrogate model using the Bayesian optimal search algorithm to predict scores under different probe poses. The AFSA is used as an optimization algorithm for the Bayesian optimal search algorithm, combined with the acquisition function, to determine the next pose most likely to improve the scanning results, guiding the ultrasound robot device towards the optimal solution. Then, by continuously updating the surrogate model and optimizing the acquisition function, the optimal probe pose is gradually approximated.

[0047] This method can also dynamically adjust the number of artificial fish and the number of iterations in the artificial fish swarm algorithm based on the amount of data in the dataset during each iteration. When the data volume is small and the surrogate model training time is short, the method increases the number of iterations and the number of artificial fish, allowing the artificial fish swarm algorithm to perform a deeper search. Conversely, when the data volume increases and the surrogate model training time is long, the method reduces the number of iterations and the number of artificial fish, thus reducing the search time. This dynamic adjustment mechanism effectively solves the problem of processing time fluctuations in different iteration rounds during Bayesian optimal search, ensuring the stability and efficiency of Bayesian optimal search.

[0048] This method significantly improves the quality and efficiency of ultrasound image acquisition through intelligent and automated means. It not only reduces the workload of ultrasound physicians and improves the utilization efficiency of medical resources, but also ensures the consistency and repeatability of examination results by continuously optimizing the search path and image quality, thereby improving the accuracy and reliability of clinical diagnosis.

[0049] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description and the accompanying drawings.

[0050] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0051] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0052] Figure 1 A flowchart illustrating the autonomous human organ scanning method based on Bayesian optimal search provided in an embodiment of the present invention;

[0053] Figure 2 This is a schematic diagram of an ultrasonic robot device provided in an embodiment of the present invention;

[0054] Figure 3 A framework diagram of the autonomous human organ scanning method based on Bayesian optimal search provided in an embodiment of the present invention;

[0055] Figure 4 A schematic diagram of the structure of the autonomous human organ scanning device based on Bayesian optimal search provided in the embodiments of this application. Detailed Implementation

[0056] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0057] The inventors discovered that, in existing technologies, due to significant differences in physical conditions among individuals, it is difficult to construct a universally applicable human body model to effectively plan ultrasound scanning paths. When scanning a specific organ, the professional skills and experience of the ultrasound physician are still required to find the optimal ultrasound probe position, which not only increases the physician's workload but also significantly consumes limited medical resources. Based on this, the inventors, through further research and development, created this invention, providing a method and device for autonomous scanning of human organs based on Bayesian optimal search.

[0058] Example 1

[0059] Embodiment 1 of this invention provides a method for autonomous scanning of human organs based on Bayesian optimal search, referring to... Figure 1 As shown, the method may include the following steps S101-S106:

[0060] S101: Train the initial surrogate model based on the acquired initial probe pose dataset and initial ultrasound image scoring dataset.

[0061] S102: Based on the amount of data in the initial ultrasound image scoring dataset, calculate the number of iterations and the number of artificial fish for the artificial fish swarm algorithm.

[0062] S103: Starting from the probe's initial pose, based on the initial surrogate model, the preset acquisition function, the number of iterations of the artificial fish swarm algorithm, and the number of artificial fish, the artificial fish swarm algorithm is used to find the next probe pose.

[0063] S104: Control the probe of the ultrasound robot to move to the next probe pose, and acquire new ultrasound images in real time. Update the initial probe pose dataset and the initial ultrasound image score dataset based on the new ultrasound images to obtain the updated probe pose dataset and the updated ultrasound image score dataset.

[0064] S105: Update the initial surrogate model based on the updated probe pose dataset and the updated ultrasound image scoring dataset to obtain the updated surrogate model.

[0065] S106: Determine if the preset stopping condition has been met: If yes, obtain the optimal probe pose and control the probe to acquire the optimal ultrasound image of the target organ at the optimal probe pose. If no, repeat steps S102-S105 above.

[0066] This invention provides a method for autonomous scanning of human organs based on Bayesian optimal search. The method trains an initial surrogate model and, based on the data volume of the initial ultrasound image scoring dataset, calculates the number of iterations and the number of fish in an artificial fish swarm algorithm. Using the artificial fish swarm algorithm as an optimization algorithm, it finds the optimal next probe pose. The ultrasound probe is then controlled to move to the next probe pose, acquiring new ultrasound images during the movement to update the dataset and surrogate model. This process of Bayesian optimal search is repeated until a stopping condition is met, resulting in the optimal probe pose. The probe is then controlled to acquire the optimal ultrasound image of the target organ under this optimal pose. This method establishes a surrogate model using the Bayesian optimal search algorithm to predict scores under different probe poses. The artificial fish swarm algorithm is used as an optimization algorithm for the Bayesian optimal search algorithm. Combined with the acquisition function, the next pose most likely to improve the scanning results is determined, guiding the ultrasound robot device to approach the optimal probe pose. Then, by continuously updating the surrogate model and optimizing the acquisition function, the optimal probe pose is gradually approximated.

[0067] This method can also dynamically adjust the number of artificial fish and the number of iterations in the artificial fish swarm algorithm based on the amount of data in the dataset during each iteration. When the data volume is small and the surrogate model training time is short, the method increases the number of iterations and the number of artificial fish, allowing the artificial fish swarm algorithm to perform a deeper search. Conversely, when the data volume increases and the surrogate model training time is long, the method reduces the number of iterations and the number of artificial fish, thus reducing the search time. This dynamic adjustment mechanism effectively solves the problem of processing time fluctuations in different iteration rounds during Bayesian optimal search, ensuring the stability and efficiency of Bayesian optimal search.

[0068] This method significantly improves the quality and efficiency of ultrasound image acquisition through intelligent and automated means. It not only reduces the workload of ultrasound physicians and improves the utilization efficiency of medical resources, but also ensures the consistency and repeatability of examination results by continuously optimizing the search path and image quality, thereby improving the accuracy and reliability of clinical diagnosis.

[0069] To facilitate understanding of this solution by those skilled in the art, an example diagram of the ultrasonic robot in the embodiments of this application is given below: The ultrasonic robot device is as follows... Figure 2 As shown, the ultrasound robot device includes a robotic arm and a computer. In this method, the computer will be deployed to control the movement of the robotic arm to perform autonomous scanning of target organs in the human body.

[0070] In this embodiment of the application, before performing step S101, step S107 needs to be performed to obtain the initial probe pose dataset and the initial ultrasound image scoring dataset.

[0071] Specifically, this can be achieved by controlling the probe to move along a random path within the scanning area of ​​the target organ, acquiring ultrasound images in real time during the movement, and recording the probe pose at each acquisition. The images are then sorted by acquisition time from oldest to newest to obtain an initial ultrasound image dataset and an initial probe pose dataset. Then, based on each ultrasound image in the initial ultrasound image dataset, a score is obtained for that ultrasound image, resulting in an initial ultrasound image score dataset.

[0072] The scoring method for ultrasound images varies depending on the target organ. For example, when the target organ is the carotid artery, the score is determined based on the area and length of the carotid artery, while when the target organ is the liver, the clarity of the liver's edges may also need to be considered.

[0073] In one specific embodiment, when the target organ is the carotid artery, the score of the ultrasound image can be determined based on the area and length of the carotid artery in the ultrasound image. Specifically, each ultrasound image in the initial ultrasound image dataset can be input into a pre-trained image segmentation model to obtain the contour of the carotid artery in the ultrasound image. Then, the area and length of the carotid artery in the ultrasound image are calculated based on OpenCV (Open Source Computer Vision Library). The area and length of the carotid artery are weighted, summed, and normalized to obtain the score of the ultrasound image.

[0074] Those skilled in the art can select a suitable neural network for pre-training based on the detailed description of the prior art to obtain a pre-trained image segmentation model. The training process may specifically include:

[0075] The first step is to collect carotid ultrasound images and label the contours of the carotid arteries in the images. After image preprocessing, a carotid contour dataset is obtained.

[0076] The second step is to select a suitable neural network model as the initial image segmentation network, such as the U-Net model or the SegNet model.

[0077] The third step is to divide the carotid artery contour dataset into a training set and a test set.

[0078] The fourth step is to define the loss function (such as cross-entropy loss function, mean squared error loss function, etc.) and the optimization algorithm (such as Adam, SGD, etc.).

[0079] The fifth step is to train the initial image segmentation network using the training set of the carotid artery contour dataset to obtain the trained image segmentation model.

[0080] Repeat the training process of the image segmentation model described above until the preset conditions are met, then stop training to obtain the pre-trained image segmentation model. The preset conditions can be set to reach a preset number of iterations, achieve a certain accuracy threshold, or ensure that the accuracy remains unchanged within a preset number of iterations. No specific limitations are required here.

[0081] In this embodiment of the application, after performing the above step S107 to obtain the initial probe pose dataset and the initial ultrasound image scoring dataset, it is also necessary to perform step S108 to perform sparse downsampling on the initial probe pose dataset and the initial ultrasound image scoring dataset.

[0082] Specifically, it can be done by traversing all probe poses in the initial probe pose dataset, calculating the L2 norm between every two adjacent probe poses, and if the L2 norm is less than the preset sampling interval, deleting the probe pose that is ranked later in the two adjacent probe poses, and deleting the score data corresponding to that probe pose in the initial ultrasound image scoring dataset.

[0083] In this embodiment of the application, after performing the above step S107 to obtain the initial probe pose dataset and the initial ultrasound image scoring dataset, it is also necessary to perform step S109 to perform data time distance filtering on the initial probe pose dataset and the initial ultrasound image scoring dataset.

[0084] Specifically, this can involve retaining a preset maximum number of probe poses and scores from the initial probe pose dataset and the initial ultrasound image scoring dataset, i.e., the latest preset maximum number of data entries, while deleting other data from older time periods. The preset maximum number can be set to 200 for example.

[0085] In step S101 above, an initial surrogate model is trained based on the acquired initial probe pose dataset and initial ultrasound image scoring dataset.

[0086] Specifically, the initial probe pose dataset can be used as training data, and the initial ultrasound image score dataset can be used as the corresponding label to train an initial surrogate model. This initial surrogate model can learn the mapping relationship between different probe poses and scores, providing a foundation for subsequent optimization search.

[0087] The initial surrogate model can be implemented based on existing technologies such as Gaussian regression, random forest, and neural networks, and no restrictions are imposed here.

[0088] In step S102 above, the number of iterations and the number of artificial fish in the artificial fish swarm algorithm are calculated based on the data volume of the initial ultrasound image scoring dataset. Specifically, this includes the following steps S1021-S1022:

[0089] S1021: Based on the data volume of the initial ultrasound image scoring dataset, the number of iterations of the artificial fish swarm algorithm is calculated according to the following formula:

[0090]

[0091] In the formula, G is the number of iterations of the artificial fish swarm algorithm, G max G is the preset maximum number of iterations. base S is the preset iteration base. p S represents the data size of the initial ultrasound image scoring dataset. max To preset the maximum data volume, This is the floor function. Where G... max It can be set to 8, G as an example. base It can be set to 4 as an example.

[0092] S1022: Based on the data volume of the initial ultrasound image scoring dataset, the number of artificial fish in the artificial fish swarm algorithm is calculated using the following formula 2:

[0093]

[0094] In the formula, F represents the number of artificial fish in the artificial fish swarm algorithm. max F is the preset maximum number of artificial fish to be added. base S is the preset artificial fish population. p S represents the data size of the initial ultrasound image scoring dataset. max To preset the maximum data volume, This is the floor function. Where F... max For example, it can be set to 10, F base It can be set to 5 as an example.

[0095] In this embodiment of the application, during the Bayesian optimal search process, the training time of the surrogate model is relatively short in the early stages of iteration due to the small amount of data. However, as the search progresses, the amount of data gradually increases in the later stages of iteration, leading to a significant increase in the training time of the surrogate model. This fluctuation in training time affects the duration of each iteration of the Bayesian optimal search, thereby affecting the stability and efficiency of the algorithm.

[0096] To address the aforementioned issues, this method achieves dynamic adjustment through step S102, using the cosine functions in formulas 1 and 2 to adjust the iteration count G and the number of artificial fish F in the artificial fish swarm algorithm according to the data volume S. p The rise and fall of the value gradually reduces the number of iterations G in the artificial fish swarm algorithm. base and G base +G max Between F, the number of artificially raised fish F base and Fbase +F max Therefore, in the early stages of the Bayesian optimal search iteration, this method increases the number of iterations of the artificial fish swarm algorithm and the number of artificial fish to extend the search time and make full use of the faster training speed at this time for a deeper search. In the later stages of the Bayesian optimal search iteration, as the amount of data increases and the training time of the surrogate model increases, the number of iterations of the artificial fish swarm algorithm and the number of artificial fish can be reduced accordingly to reduce the search time, thereby maintaining the relative stability of the iteration time of each round and improving the stability and overall efficiency of the Bayesian optimal search.

[0097] In step S103 above, starting from the probe's initial pose, based on the initial surrogate model, the preset acquisition function, the number of iterations of the artificial fish swarm algorithm, and the number of artificial fish, the artificial fish swarm algorithm is used to find the next probe pose. Specifically, this includes the following steps S1031-S1035:

[0098] S1031: Initialize multiple artificial fish according to the probe's initial pose and the number of artificial fish, with each artificial fish corresponding to a pose.

[0099] Specifically, it can be that the search space is determined based on the probe's initial pose and the preset maximum step size, and multiple candidate points are randomly selected within the search space as the poses corresponding to multiple artificial fish.

[0100] S1032: Calculate the sampling potential value for each artificial fish based on the initial agent model and the acquisition function.

[0101] Specifically, the pose corresponding to each artificial fish can be input into the initial surrogate model to obtain the prediction score and variance. The prediction score, variance, and pose corresponding to the artificial fish can then be input into the acquisition function to obtain the sampling potential value of the artificial fish.

[0102] The acquisition function can be set to Expected Improvement or Upper Confidence Bound, etc., and is not limited here.

[0103] S1033: Control all artificial fish to perform swarming, tail-chasing, foraging, and random behaviors in the artificial fish swarm algorithm based on the corresponding sampling potential value, and iteratively update all artificial fish until the number of iterations of the artificial fish swarm algorithm is reached, resulting in multiple updated artificial fish.

[0104] Specifically, this can be done by first having all artificial fish perform swarming behavior, followed by tail-chasing behavior. After performing swarming and tail-chasing behavior, the sampling potential value of each artificial fish is calculated. If no artificial fish has a sampling potential value exceeding the maximum sampling potential value before swarming and tail-chasing behavior, i.e. no better solution has been found, then all artificial fish continue to perform foraging behavior. If no better solution has been found, then random behavior is performed. This completes one iteration of the artificial fish swarming algorithm. The artificial fish swarming algorithm is then re-executed until the number of iterations of the artificial fish swarming algorithm is reached, resulting in multiple updated artificial fish and the pose corresponding to each updated artificial fish.

[0105] The artificial fish swarm algorithm includes four basic behaviors: swarming, tail-chasing, foraging, and random behavior. Specifically, these include:

[0106] Swarming Behavior: Artificial fish swarms tend to act collectively. Each artificial fish decides whether to follow other fish based on the location of nearby artificial fish and food concentration, maintaining the group's concentration and avoiding dispersion. This helps to evenly distribute the fish throughout the search space, avoiding local optima traps. The pose of the artificial fish after swarming behavior is calculated based on the following formula 3:

[0107]

[0108] In the formula, X next X represents the pose of the artificially induced fish after they start swarming, and X represents the pose of the artificially induced fish before they start swarming. c S is the center location of the artificial fish in the neighborhood. f To preset the maximum step size, U(0,1) represents a random number within the interval 0-1. Where S... f It can be set to 0.3 as an example.

[0109] The location X of the center of the neighboring artificial fish in Formula 3 above. c The following formula 4 is used to calculate:

[0110]

[0111] In the formula, X c N represents the pose corresponding to the center position of the neighboring artificial fish. f The number of other artificial fish within the field of vision of the artificial fish. Let represent the pose of the i-th artificial fish within its field of vision.

[0112] It should be noted that for the artificial fish to exhibit the above-mentioned swarming behavior, they must meet the swarming conditions, which are shown in Formula 5 below:

[0113]

[0114] In the formula, δ is the crowding factor, Y is the sampling potential value of the artificial fish, and N is the sampling potential value of the artificial fish. f Y represents the number of other artificial fish within the artificial fish's field of vision. c The pose X corresponding to the center position of the neighboring artificial fish c The corresponding sampling potential value. Here, δ can be set to 0.2 for example.

[0115] Tail-chasing behavior: In an artificial fish swarm, if one artificial fish discovers that another artificial fish has a better sampling potential value, it may follow that fish, using its exploration results to accelerate the search. The pose of the artificial fish after the tail-chasing behavior is calculated based on the following formula 6:

[0116]

[0117] In the formula, X next X represents the pose of the artificial fish after the tail-chasing behavior, and X represents the pose of the artificial fish before the tail-chasing behavior. o S represents the pose of the artificial fish with the highest sampling potential value within the artificial fish's field of vision. f To preset the maximum step size, U(0,1) represents a random number within the interval 0-1. Where S... f It can be set to 0.3 as an example.

[0118] It should be noted that for the artificial fish to exhibit the aforementioned swarming behavior, the tail-chasing condition must be met, as shown in Formula 7 below:

[0119]

[0120] In the formula, δ is the crowding factor, Y is the sampling potential value of the artificial fish, and N is the sampling potential value of the artificial fish. f Y represents the number of other artificial fish within the artificial fish's field of vision. o This represents the highest sampling potential value within the field of vision of the artificial fish. δ can be set to 0.2, for example.

[0121] Foraging behavior: The artificial fish senses the current food concentration (i.e., the objective function value) within its field of vision, and then determines its direction of movement based on the food concentration. If the food concentration ahead is high, it moves in that direction; otherwise, it randomly chooses another direction to continue foraging. The pose of the artificial fish after foraging behavior is calculated based on the following formula 8:

[0122]

[0123] In the formula, X next X represents the pose of the artificially foraged fish after the foraging behavior, and X represents the pose of the artificially foraged fish before the foraging behavior. j S represents the random pose within the line of sight of the artificial fish. fTo preset the maximum step size, U(0,1) represents a random number within the interval 0-1. Where S... f It can be set to 0.3 as an example.

[0124] The random pose X within the line of sight of the artificial fish in Formula 8 above. j The following formula 9 is used for calculation:

[0125] X j =X+V f *U(0,1) Formula 9

[0126] In the formula, X represents the pose of the artificial fish before its foraging behavior. j V represents the random pose within the line of sight of the artificial fish. f Let V be the line of sight of the artificial fish, and U(0,1) represent a random number in the interval 0-1. f It can be set to 0.5 as an example.

[0127] Random behavior: If the above three behaviors do not yield a better solution, the artificial fish can randomly choose a direction to explore, increasing the diversity of the search space and avoiding getting trapped in local optima. The pose of the artificial fish after random behavior is calculated based on the following formula 10:

[0128] X next =X+V f *U(0,1) Formula 10

[0129] In the formula, X next Let X be the pose of the artificial fish after its random actions, and V be the pose of the artificial fish before its random actions. f Let V be the line of sight of the artificial fish, and U(0,1) represent a random number in the interval 0-1. f It can be set to 0.5 as an example.

[0130] S1034: Calculate the sampling potential value for each updated artificial fish based on the initial agent model and the acquisition function.

[0131] Specifically, the pose corresponding to each updated artificial fish can be input into the initial surrogate model to obtain the prediction score and variance. The prediction score, variance, and pose corresponding to the updated artificial fish can then be input into the acquisition function to obtain the sampling potential value of the updated artificial fish.

[0132] S1035: Update the pose of the artificial fish with the highest sampling potential value and use it as the pose of the next probe.

[0133] In this embodiment, the Bayesian optimal search requires an optimization algorithm to be executed in each iteration. If the optimization algorithm used is inefficient, the iteration cycle of the Bayesian optimal search algorithm will be significantly prolonged, leading to problems such as slow robot movement and disordered movement rhythm, interfering with the normal operation of other programs and significantly reducing the overall scanning efficiency. Furthermore, the gradient ascent method commonly used in the prior art is prone to getting trapped in local optima. The artificial fish swarm algorithm used in this method has a fast convergence speed and strong global search capability, effectively exploring the search space and avoiding getting trapped in local optima. Its parallel operation is suitable for various computing environments, improving solution efficiency. Moreover, this algorithm is applicable to multi-objective, continuous, and discrete optimization problems, exhibiting high robustness and scalability, and is suitable for the high real-time requirements of global optimization problems in this method.

[0134] In step S104 above, the probe of the ultrasound robot is controlled to move to the next probe pose, and new ultrasound images are acquired in real time. The initial probe pose dataset and the initial ultrasound image scoring dataset are then updated based on the new ultrasound images to obtain the updated probe pose dataset and the updated ultrasound image scoring dataset. Specifically, this includes the following steps S1041-S1042:

[0135] S1041: Control the probe of the ultrasound robot to move to the next probe pose, and acquire multiple new ultrasound images and the probe pose corresponding to each new ultrasound image in real time, and determine the score corresponding to each new ultrasound image.

[0136] The method for determining the corresponding score based on the new ultrasound image has been explained in detail in step S107 above, and will not be repeated here.

[0137] S1042: Add the probe pose and score corresponding to multiple new ultrasound images to the initial probe pose dataset and the initial ultrasound image score dataset, respectively. Simultaneously, perform sparse downsampling to obtain the updated probe pose dataset and the updated ultrasound image score dataset. Specifically, this includes the following steps S10421-S10423:

[0138] S10421: Calculate the L2 norm of the probe pose corresponding to the last probe pose and the first new ultrasound image in the initial probe pose dataset, and determine whether the L2 norm is greater than the preset sampling interval: if yes, proceed to step S10422; if no, proceed to step S10423.

[0139] Among them, the earliest new ultrasound image refers to the newest ultrasound image acquired earliest among multiple new ultrasound images.

[0140] S10422: Add the probe pose corresponding to the first new ultrasound image to the initial probe pose dataset, and add the score corresponding to the first new ultrasound image to the initial ultrasound image score dataset.

[0141] S10423: Re-execute the above L2 norm calculation process based on subsequent new ultrasound images until no subsequent new ultrasound images exist, to obtain the updated probe pose dataset and the updated ultrasound image scoring dataset.

[0142] To facilitate understanding of this solution by those skilled in the art, the specific implementation process of S1042 provided in the embodiments of this application is described more clearly and completely below: The probe poses corresponding to multiple new ultrasound images are represented as P = {p1, p2} 2, …,p i ,...,p n The corresponding score is E={e1,e}. 2, …,e i ,...,e n Let the initial probe pose dataset be represented as P. x ={...,p0}, the initial ultrasound image scoring dataset is represented as E x ={...,e0}. Where, p i e represents the probe pose corresponding to the i-th new ultrasound image. i Let p represent the score corresponding to the i-th new ultrasound image, p0 represent the last probe pose in the initial probe pose dataset, and e0 represent the last score in the initial ultrasound image score dataset. Set the current probe pose p... current =p0, then perform the following calculations:

[0143] d i =‖p i -p current ‖2 for i=1,2,…,n

[0144] if d i >τ P x =P x ∪{p i E x =E x ∪{e i} p current =p i

[0145] Where, d i τ represents the L2 norm of the two probe poses, and τ is the preset sampling interval, which can be set to 0.03 for example.

[0146] In this embodiment of the application, after performing the above step S104 to obtain the updated probe pose dataset and the updated ultrasound image scoring dataset, it is also necessary to perform step S109 to perform data time distance filtering on the updated probe pose dataset and the updated ultrasound image scoring dataset.

[0147] Specifically, this can involve retaining a preset maximum number of probe poses and scores from the updated probe pose dataset and the updated ultrasound image scoring dataset—that is, the latest preset maximum number of data entries—while deleting other data entries from older time periods. The preset maximum number can be set to 200 for example.

[0148] The specific algorithm is as follows:

[0149]

[0150] Among them, S max To preset the maximum quantity, S P The amount of data to update the probe pose dataset or the ultrasound image scoring dataset, where k is the index. Where S... max An example setting is 200.

[0151] In this embodiment, if all data is sampled during probe data acquisition, the dataset becomes too large, the surrogate model training time is too long, and the scanning efficiency is reduced. Furthermore, due to unavoidable movement and deformation of the human body during respiration, the condition of the target organ may change over time, and the reliability of data from further time intervals decreases. Therefore, in step S104, this method effectively reduces the dataset size, data redundancy, and computational load through sparse data downsampling and temporal distance filtering, thereby improving processing efficiency. By filtering out invalid data with long time intervals, the timeliness and relevance of the data are ensured. This also reduces noise interference and improves the accuracy and efficiency of the search.

[0152] In step S105 above, the initial surrogate model is updated based on the updated probe pose dataset and the updated ultrasound image scoring dataset to obtain the updated surrogate model.

[0153] Specifically, it can be that the hyperparameters of the initial surrogate model are re-optimized based on the updated probe pose dataset and the updated ultrasound image scoring dataset to obtain the updated surrogate model.

[0154] In step S106 above, it is determined whether the preset stopping condition has been met: if yes, the optimal probe pose is obtained, and the probe is controlled to acquire the optimal ultrasound image of the target organ at the optimal probe pose; if no, steps S102-S105 above are executed again.

[0155] The preset stopping conditions can be reaching the maximum number of iterations, or the ultrasound image score reaching a preset score threshold, etc.

[0156] To facilitate understanding of this solution by those skilled in the art, the specific implementation process of the autonomous human organ scanning method based on Bayesian optimal search provided in the embodiments of this application is described more clearly and completely below: (Refer to...) Figure 3As shown in the figure, taking the carotid artery as the target organ as an example, the overall framework of this method is illustrated:

[0157] The upper left of the figure shows the process of obtaining the initial ultrasound image scoring dataset and the initial probe pose dataset corresponding to step S107.

[0158] The right side of the figure is a schematic diagram of the Bayesian optimal search steps described in steps S102-S106 of this method. Specifically, it includes: starting the iterative process of Bayesian optimal search based on the initial ultrasound image scoring dataset and the initial probe pose dataset. The preset stopping condition is whether the maximum number of iterations has been reached. If yes, the optimal probe pose is obtained. If no, the surrogate model is updated, the acquisition function is optimized, the next probe pose is output, the ultrasound robot is controlled to move, the image scores and corresponding values ​​of the moving process are acquired, the probe pose is updated, and the image evaluation score dataset and the probe pose dataset are updated.

[0159] The lower left corner of the figure shows a schematic diagram of controlling the probe to acquire the optimal ultrasound image at the optimal probe pose after obtaining the optimal probe pose.

[0160] Example 2

[0161] Based on the same inventive concept, embodiments of the present invention also provide an autonomous human organ scanning device based on Bayesian optimal search, referring to... Figure 4 As shown, the device includes:

[0162] The first training module 101 is used to train the initial surrogate model based on the acquired initial probe pose dataset and initial ultrasound image scoring dataset.

[0163] The first calculation module 102 is used to calculate the number of iterations of the artificial fish swarm algorithm and the number of artificial fish based on the amount of data in the initial ultrasound image scoring dataset.

[0164] The first optimization module 103 is used to find the next probe pose starting from the probe's initial pose, based on the initial proxy model, the preset acquisition function, the number of iterations of the artificial fish swarm algorithm, and the number of artificial fish.

[0165] The first update module 104 is used to control the probe of the ultrasound robot to move to the next probe pose, and to acquire new ultrasound images in real time, and to update the initial probe pose dataset and the initial ultrasound image score dataset according to the new ultrasound images, so as to obtain the updated probe pose dataset and the updated ultrasound image score dataset.

[0166] The second update module 105 is used to update the initial proxy model based on the updated probe pose dataset and the updated ultrasound image scoring dataset to obtain the updated proxy model.

[0167] The first judgment module 106 is used to determine whether the preset stop condition has been met: if yes, the optimal probe pose is obtained, and the probe is controlled to acquire the optimal ultrasound image of the target organ at the optimal probe pose; if no, the first calculation module is re-executed.

[0168] Example 3

[0169] Based on the same inventive concept, embodiments of the present invention also provide a computer-readable storage medium storing a computer program / instruction thereon, which, when executed by a processor, implements the autonomous human organ scanning method based on Bayesian optimal search as described in Embodiment 1 above.

[0170] Example 4

[0171] Based on the same inventive concept, this embodiment of the invention also provides a computer program product, including a computer program / instruction, which, when executed by a processor, implements the autonomous human organ scanning method based on Bayesian optimal search as described in Embodiment 1 above.

[0172] Example 5

[0173] Based on the same inventive concept, this embodiment of the invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory. When the processor executes the computer program, it implements the autonomous human organ scanning method based on Bayesian optimal search as described in Embodiment 1 above.

[0174] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage and optical storage) containing computer-usable program code.

[0175] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1A device that provides the functions specified in one or more boxes.

[0176] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0177] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0178] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for autonomous scanning of human organs based on Bayesian optimal search, characterized in that, include: The initial surrogate model is trained based on the acquired initial probe pose dataset and initial ultrasound image scoring dataset; Based on the data volume of the initial ultrasound image scoring dataset, the number of iterations of the artificial fish swarm algorithm and the number of artificial fish in the artificial fish swarm algorithm are calculated using the following formulas: In the formula, This represents the number of iterations in the artificial fish swarm algorithm. The preset maximum number of iterations, The preset iteration base, This represents the number of artificial fish in the artificial fish swarm algorithm. To set the maximum number of artificial fish to be added, The pre-set artificial fish population, The data size of the initial ultrasound image scoring dataset. To preset the maximum data volume, It is a rounding function; Starting from the probe's initial pose, based on the initial proxy model, the preset acquisition function, the number of iterations of the artificial fish swarm algorithm, and the number of artificial fish, the artificial fish swarm algorithm is used to find the next probe pose; The probe of the ultrasound robot is controlled to move to the next probe pose and new ultrasound images are acquired in real time. The initial probe pose dataset and the initial ultrasound image score dataset are updated based on the new ultrasound images to obtain the updated probe pose dataset and the updated ultrasound image score dataset. The initial proxy model is updated based on the updated probe pose dataset and the updated ultrasound image scoring dataset to obtain the updated proxy model. The Bayesian optimal search process is repeated until the preset stopping condition is met, the optimal probe pose is obtained, and the probe is controlled to acquire the optimal ultrasound image of the target organ at the optimal probe pose.

2. The method according to claim 1, characterized in that, The probe of the controlled ultrasound robot moves to the next probe pose and acquires new ultrasound images in real time. Based on these new ultrasound images, the initial probe pose dataset and the initial ultrasound image scoring dataset are updated to obtain the updated probe pose dataset and the updated ultrasound image scoring dataset, including: The probe of the ultrasound robot is controlled to move to the next probe pose, and multiple new ultrasound images and the probe pose corresponding to each new ultrasound image are acquired in real time, and the score corresponding to each new ultrasound image is determined. The probe pose and score corresponding to the multiple new ultrasound images are added to the initial probe pose dataset and the initial ultrasound image score dataset, respectively. At the same time, sparse downsampling is performed to obtain the updated probe pose dataset and the updated ultrasound image score dataset.

3. The method according to claim 2, characterized in that, The updated probe pose dataset, the updated ultrasound image scoring dataset, and multiple new ultrasound images are sorted by acquisition time. The step of adding the probe pose and score corresponding to the multiple new ultrasound images to the initial probe pose dataset and the initial ultrasound image score dataset, respectively, and simultaneously performing sparse downsampling to obtain the updated probe pose dataset and the updated ultrasound image score dataset, includes: Calculate the L2 norm of the probe pose corresponding to the last probe pose and the first new ultrasound image in the initial probe pose dataset, and determine whether the L2 norm is greater than the preset sampling interval. If so, the probe pose corresponding to the first new ultrasound image is added to the initial probe pose dataset, and the score corresponding to the first new ultrasound image is added to the initial ultrasound image score dataset. If not, the process of calculating the L2 norm is repeated based on subsequent new ultrasound images until there are no unprocessed new ultrasound images, resulting in an updated probe pose dataset and an updated ultrasound image scoring dataset.

4. The method according to claim 2, characterized in that, The data in the updated probe pose dataset and the updated ultrasound image scoring dataset are sorted by acquisition time. After obtaining the updated probe pose dataset and the updated ultrasound image scoring dataset, the following is also included: The maximum number of probe poses and scores that are sorted last in the updated probe pose dataset and the updated ultrasound image scoring dataset are retained.

5. The method according to claim 1, characterized in that, Starting from the probe's initial pose, based on the initial surrogate model, the preset acquisition function, the number of iterations of the artificial fish swarm algorithm, and the number of artificial fish, the artificial fish swarm algorithm is used to find the next probe pose, including: Multiple artificial fish are initialized based on the probe's initial pose and the number of artificial fish; wherein each artificial fish corresponds to a pose; Based on the initial agent model and the acquisition function, calculate the sampling potential value for each artificial fish; All artificial fish are controlled to perform swarming, tail-chasing, foraging, and random behaviors in the artificial fish swarm algorithm based on their corresponding sampling potential values. All artificial fish are iteratively updated until the number of iterations of the artificial fish swarm algorithm is reached, resulting in multiple updated artificial fish. Based on the initial agent model and the acquisition function, calculate the sampling potential value for each of the updated artificial fish; The pose of the artificial fish with the highest update sampling potential value is used as the pose of the next probe.

6. A human organ autonomous scanning device based on Bayesian optimal search, characterized in that, include: The first training module is used to train the initial surrogate model based on the acquired initial probe pose dataset and initial ultrasound image scoring dataset. The first calculation module is used to calculate the number of iterations of the artificial fish swarm algorithm and the number of artificial fish in the artificial fish swarm algorithm based on the data volume of the initial ultrasound image scoring dataset and the following formula: In the formula, This represents the number of iterations in the artificial fish swarm algorithm. The preset maximum number of iterations, The preset iteration base, This represents the number of artificial fish in the artificial fish swarm algorithm. To set the maximum number of artificial fish to be added, The pre-set artificial fish population, The data size of the initial ultrasound image scoring dataset. To preset the maximum data volume, It is a rounding function; The first optimization module is used to find the next probe pose starting from the probe's initial pose, based on the initial proxy model, the preset acquisition function, the number of iterations of the artificial fish swarm algorithm, and the number of artificial fish; The first update module is used to control the probe of the ultrasound robot to move to the next probe pose, and to acquire new ultrasound images in real time. The module updates the initial probe pose dataset and the initial ultrasound image score dataset based on the new ultrasound images to obtain the updated probe pose dataset and the updated ultrasound image score dataset. The second update module is used to update the initial proxy model based on the updated probe pose dataset and the updated ultrasound image scoring dataset to obtain the updated proxy model. The first judgment module is used to determine whether the preset stopping condition has been met: if yes, the optimal probe pose is obtained, and the probe is controlled to acquire the optimal ultrasound image of the target organ at the optimal probe pose; if no, the first calculation module is executed again.

7. A computer-readable storage medium having a computer program / instructions stored thereon, characterized in that, When the computer program / instruction is executed by the processor, it implements the autonomous human organ scanning method based on Bayesian optimal search as described in any one of claims 1-5.

8. A computer program product comprising a computer program / instructions, characterized in that, When the computer program / instruction is executed by the processor, it implements the autonomous human organ scanning method based on Bayesian optimal search as described in any one of claims 1-5.

9. A computer device, comprising a memory, a processor, and a computer program stored in the memory, characterized in that, The processor executes the computer program to implement the autonomous human organ scanning method based on Bayesian optimal search as described in any one of claims 1-5.

Images