Ultrasonic image processing parameter optimization method
By iterative processing and downsampling technology within the effective parameter range of ultrasound images, combined with doctor selection information, the bias and overfitting problems of parameter optimization in ultrasound image processing are solved, and efficient parameter optimization is achieved.
Patent Information
- Application Number
- CN202510629976.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-05-16
- Publication Date
- 2025-10-10
AI Technical Summary
In existing ultrasound image processing methods, parameter optimization relies on the subjective understanding of algorithm engineers and is prone to bias. In addition, deep learning methods lack high-quality training data, leading to overfitting and difficulty in finding the optimal solution.
By iteratively processing the initialization parameters of the ultrasound image within the effective parameter range and combining it with downsampling technology, an optimized image is generated, and a target parameter set is generated based on the doctor's selection information, avoiding single parameter adjustment by engineers and overfitting of deep learning.
It improves the efficiency and quality of parameter optimization, can efficiently find suboptimal solutions or even optimal solutions, and overcomes the problems of engineer bias and insufficient deep learning data.
Smart Images
Figure CN120765673A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of image parameter optimization, and in particular to an ultrasonic image processing parameter optimization method. BACKGROUND
[0002] Ultrasonic image processing generally refers to digital image processing of two-dimensional gray-scale images of an ultrasonic system to obtain better image resolution and contrast. Image processing generally includes two main goals of noise suppression and boundary enhancement.
[0003] Existing image processing methods mainly include three directions: one is a method based on gray scale, gradient, threshold and other explicit features; two is a machine learning method based on statistics and regression; and three is a deep learning method based on implicit features.
[0004] Regardless of which method is used, the number of super parameters that need to be adjusted can reach tens, and the reasonable value range of each parameter also has several or even dozens, and the permutation and combination can reach an astonishing number of trillions, forming a huge high-dimensional parameter space.
[0005] Due to the huge number of parameter spaces, parameter optimization brings great challenges, and image processing parameter optimization work usually relies on the subjective understanding of algorithm engineers to the algorithm. Such parameters have a strong bias and are most likely not the optimal solution.
[0006] Although the deep learning method can automatically extract features to fit the optimal solution, it is very dependent on the quantity and quality of the training data. However, it is actually difficult to obtain high-quality two-dimensional gray-scale images as training data to meet the requirements. A large amount of data is required for training to avoid overfitting, and it is more difficult to obtain a large number of high-quality gray-scale images.
[0007] Therefore, the present application provides an ultrasonic image processing parameter optimization method. SUMMARY
[0008] The ultrasonic image processing parameter optimization method can efficiently obtain a suboptimal solution of the parameters through multiple iterations of parameter approximation and user confirmation, thereby meeting the user's use requirements.
[0009] The present application provides an ultrasonic image processing parameter optimization method, comprising:
[0010] Step 1: identifying a plurality of initial parameters of an ultrasonic image, determining an effective parameter range corresponding to each of the initial parameters, and matching an initial iteration step for each of the initial parameters within the effective parameter range;
[0011] Step 2: iteratively adjusting the corresponding initialization parameters using the initial iteration step size to determine a plurality of image boundaries contained in the ultrasound image;
[0012] Step 3: downsampling the ultrasound image to obtain downsampling presentation information corresponding to each image boundary, and generating an optimized image of the ultrasound image;
[0013] Step 4: retrieve the corresponding target optimization image according to the selection information uploaded by the doctor, generate a target parameter set according to the target optimization parameters corresponding to the target optimization image, and transmit it to the doctor's terminal for display.
[0014] In one practicable manner,
[0015] The step 1 comprises:
[0016] Step 11: Acquire several basic parameters of the ultrasound image, and determine several initialization parameters of the ultrasound image based on parameter values corresponding to each of the basic parameters;
[0017] Step 12: determining a valid parameter range corresponding to each of the initialization parameters according to the specification information of the ultrasound image;
[0018] Step 13: Compile an iterative pyramid operation script based on the parameter values of the initialization parameters, and determine the initial iteration step length corresponding to each initialization parameter according to the valid parameter range.
[0019] In one practicable manner,
[0020] The step 13 includes:
[0021] Step 131: writing the initial iteration parameters corresponding to each of the initialization parameters into the running script to generate an image processing pyramid;
[0022] Step 132: performing verification processing on the ultrasound image using the image processing pyramid to generate a verification image, and obtaining a processing ratio between a verification parameter set corresponding to the verification image and each of the initialization parameters;
[0023] Step 133: adjusting the initial iteration step length according to the processing ratio until all the processing ratios are consistent, thereby obtaining the initial iteration step length corresponding to each initialization parameter.
[0024] In one practicable manner,
[0025] The step 2 comprises:
[0026] Step 21: performing a first iterative process on the initialization parameters using the initial iteration step size to obtain a corresponding first image boundary, performing enhancement and noise suppression on the first image boundary, and generating a first iteration result;
[0027] Step 22: Analyze the doctor's first satisfaction level with the first iteration result according to the operation target uploaded by the doctor. When the satisfaction level is lower than the standard level, adjust the initial iteration step size within the valid parameter range to obtain a second iteration step size.
[0028] Step 23: Perform a second iteration on the result of the first iteration using the second iteration step to generate a second iteration result, and determine whether the second satisfaction level corresponding to the second iteration is higher than the standard level. If not, the iteration result with a higher satisfaction level is regarded as the iteration basis for the next iteration.
[0029] Step 24: Obtain the Nth iteration result with the Nth satisfaction level higher than the standard level, and filter the corresponding Nth image boundary to transmit to the doctor terminal for display.
[0030] In one practicable manner,
[0031] The step 3 comprises:
[0032] Step 31: downsampling the ultrasound image using an image processing pyramid to obtain a plurality of sets of sampling information and marking them in the image processing pyramid, and inputting the image boundaries into corresponding layers of the image processing pyramid in an iterative processing order;
[0033] Step 32: reconstructing the image boundary using the sampling information in each tower layer to obtain down-sampling presentation information corresponding to each image boundary;
[0034] Step 33: Utilize the downsampling presentation information to correct each image boundary contained in the ultrasound image, obtain a corresponding optimized image, and display the image.
[0035] In one practicable manner,
[0036] Also includes:
[0037] The iteration order corresponding to each optimized image is marked respectively, and an image selection task is generated in combination with a preset selection instruction and transmitted to the doctor terminal to wait for the doctor to select a target optimized image.
[0038] In one practicable manner,
[0039] The step 4 comprises:
[0040] Step 41: selecting corresponding target optimization images and target optimization parameters according to selection information uploaded by the doctor;
[0041] Step 42: marking the parameter optimization result corresponding to each target optimization parameter in the target optimization image respectively;
[0042] Step 43: combining each target optimization parameter with the corresponding parameter optimization result respectively to generate a target parameter set and transmit to the doctor terminal for display.
[0043] In an implementable manner,
[0044] Further comprising:
[0045] When the doctor does not upload selection information, comparing the boundary contour information of the optimization images, and according to the comparison result, selecting the corresponding target optimization image and parameter target parameter set and transmitting to the doctor terminal for display.
[0046] The implementable beneficial effects of the above technical solution are: in order to avoid the situation that the engineer tuning is easy to fall into bias, and also in order to overcome the overfitting problem caused by the shortage of high-quality data in deep learning, the image boundary contained in the ultrasound image is determined by iteratively processing the initialization parameters of the ultrasound image in the effective parameter range, and further the downsampling presentation information of each image boundary is determined by the downsampling manner, so as to generate the corresponding optimization image, finally according to the selection information uploaded by the doctor to retrieve the corresponding target optimization image, using the target optimization parameters corresponding to the image to generate the parameter set and transmit it to the doctor terminal for display, in this way, the engineer will not fall into local optimal solution due to single parameter tuning, and the overfitting caused by the lack of high-quality labeled images in the deep learning method can be overcome, the suboptimal solution or even the optimal solution of the parameter space can be efficiently found, and the efficiency and quality of parameter optimization are improved.
[0047] Other features and advantages of the present application will be set forth in the following description, and in part will become apparent to those skilled in the art from the description, or can be learned by practice of the present application. The objects and other advantages of the present application can be achieved and obtained by the structure particularly pointed out in the written description and the accompanying drawings.
[0048] The technical solutions of the present application will be further described in detail below by means of the accompanying drawings and examples. BRIEF DESCRIPTION OF DRAWINGS
[0049] The accompanying drawings are used to provide a further understanding of the present application, and constitute a part of the specification, together with the embodiments of the present application, used as an explanation of the present application, and do not constitute a limitation of the present application. In the drawings:
[0050] Figure 1A schematic diagram of a work flow of the method for optimizing an ultrasound image processing parameter in an embodiment of the present application is shown in FIG. 1.
[0051] Figure 2 A schematic diagram of a work flow of step 2 of the method for optimizing an ultrasound image processing parameter in an embodiment of the present application is shown in FIG. 2. DETAILED DESCRIPTION
[0052] The preferred embodiments of the present application are described below in conjunction with the accompanying drawings, in which it is to be understood that the preferred embodiments described below are merely used to illustrate and explain the present application, and are not used to limit the present application.
[0053] Embodiment 1
[0054] The present embodiment provides a method for optimizing an ultrasound image processing parameter, as shown in FIG. 1, which comprises the following steps: Figure 1
[0055] Step 1: identifying a plurality of initial parameters of an ultrasound image, determining an effective parameter range corresponding to each of the initial parameters, and matching an initial iteration step for each of the initial parameters within the effective parameter range;
[0056] Step 2: adjusting the initial iteration step for the corresponding initial parameter, and determining a plurality of image boundaries contained in the ultrasound image;
[0057] Step 3: down-sampling the ultrasound image to obtain down-sampling presentation information corresponding to each of the image boundaries, and generating an optimized image of the ultrasound image;
[0058] Step 4: retrieving a target optimized image according to selection information uploaded by a doctor, generating a target parameter set according to a target optimization parameter corresponding to the target optimized image, and transmitting the target parameter set to a doctor terminal for display.
[0059] In this example, the initial parameters represent parameters presented by the ultrasound image, including size, resolution, color mode, color depth, file format, compression, image quality, transparency, and background.
[0060] In this example, the image boundaries represent regions in the ultrasound image where brightness or color changes significantly.
[0061] In this example, the down-sampling presentation information represents characteristics presented by the image boundaries after down-sampling the ultrasound image.
[0062] The working principle and beneficial effects of the above technical solution: In order to avoid the situation where engineers are prone to bias in tuning, and to overcome the overfitting problem caused by the shortage of high-quality data in deep learning, the image boundaries contained in the ultrasound image are determined by iteratively processing the initialization parameters of the ultrasound image within the effective parameter range, and the downsampling presentation information of each image boundary is further determined by downsampling, so as to generate the corresponding optimized image. Finally, the corresponding target optimized image is retrieved according to the selection information uploaded by the doctor, and the target optimization parameters corresponding to the image are used to generate a parameter set and transmit it to the doctor's terminal for display. In this way, it will not fall into the local optimal solution due to the single parameter adjustment of the engineer, and it can also overcome the overfitting caused by the lack of high-quality annotated images in the deep learning method. It can efficiently find the suboptimal solution or even the optimal solution in the parameter space, thereby improving the efficiency and quality of parameter optimization.
[0063] Example 2
[0064] On the basis of Example 1, the ultrasound image processing parameter optimization method, step 1, comprises:
[0065] Step 11: Acquire several basic parameters of the ultrasound image, and determine several initialization parameters of the ultrasound image based on parameter values corresponding to each of the basic parameters;
[0066] Step 12: determining a valid parameter range corresponding to each of the initialization parameters according to the specification information of the ultrasound image;
[0067] Step 13: Compile an iterative pyramid operation script based on the parameter values of the initialization parameters, and determine the initial iteration step corresponding to each of the initialization parameters according to the valid parameter range.
[0068] In this example, the basic parameters represent the size, resolution, color mode, color depth, file format, compression, image quality, transparency, and background presentation parameters of the ultrasound image;
[0069] In this instance, running the script means writing the code for iterating the pyramid;
[0070] In this example, the iterative pyramid represents an image processing pyramid that performs iterative processing.
[0071] The working principle and beneficial effects of the above technical solution are as follows: the initialization parameters of the ultrasound image are determined by obtaining the basic parameters of the ultrasound image and their parameter values, and then the valid parameter range of each initialization parameter is determined in combination with the specification information of the ultrasound image, and then the initial iteration step of each initialization parameter is determined within the range. At the same time, the running script of the iterative pyramid is written according to the parameter value. In this way, a usable initial iteration step can be obtained, which provides a good start for the iterative work.
[0072] Example 3
[0073] Based on Example 2, the ultrasound image processing parameter optimization method, step 13, includes:
[0074] Step 131: writing the initial iteration parameters corresponding to each of the initialization parameters into the running script to generate an image processing pyramid;
[0075] Step 132: performing verification processing on the ultrasound image using the image processing pyramid to generate a verification image, and obtaining a processing ratio between a verification parameter set corresponding to the verification image and each of the initialization parameters;
[0076] Step 133: adjusting the initial iteration step length according to the processing ratio until all the processing ratios are consistent, thereby obtaining the initial iteration step length corresponding to each initialization parameter.
[0077] In this example, the image processing pyramid represents a pyramid technique for iteratively processing ultrasound images;
[0078] In this example, the verification process refers to a process of performing a processing test on the ultrasound image using the image processing pyramid.
[0079] The working principle and beneficial effects of the above technical solution are as follows: In order to synchronously process all image boundaries in the ultrasound image, the initial iteration parameters are first reduced and written in the running script to construct an image processing pyramid, and the image processing pyramid is further used to verify the ultrasound image. By analyzing the consistency of the processing ratio between the verification parameters of the verification image and the initialization parameters, it is judged whether the initial iteration step size is reasonable. If necessary, corresponding adjustments are made to obtain a usable and synchronized initial iteration step size.
[0080] Example 4
[0081] On the basis of Example 1, the ultrasound image processing parameter optimization method is as follows: Figure 2 As shown, the step 2 includes:
[0082] Step 21: performing a first iterative process on the initialization parameters using the initial iteration step size to obtain a corresponding first image boundary, performing enhancement and noise suppression on the first image boundary, and generating a first iteration result;
[0083] Step 22: Analyze the doctor's first satisfaction level with the first iteration result according to the operation target uploaded by the doctor. When the satisfaction level is lower than the standard level, adjust the initial iteration step size within the valid parameter range to obtain a second iteration step size.
[0084] Step 23: Perform a second iteration on the result of the first iteration using the second iteration step to generate a second iteration result, and determine whether the second satisfaction level corresponding to the second iteration is higher than the standard level. If not, the iteration result with a higher satisfaction level is regarded as the iteration basis for the next iteration.
[0085] Step 24: Obtain the Nth iteration result with the Nth satisfaction level higher than the standard level, and filter the corresponding Nth image boundary to transmit to the doctor terminal for display.
[0086] In this instance, the standard satisfaction level is 80%;
[0087] In this example, the iteration basis represents an iteration result with a high degree of satisfaction, that is, each iteration is performed on the basis of an existing iteration result, and the iteration result is the iteration result with the highest degree of satisfaction among all the iteration results.
[0088] The working principle and beneficial effects of the above technical solution are as follows: by using the initial iteration step to perform the first iterative processing on the initial parameters, and enhancing and noise suppressing the obtained image boundaries, the first iterative result is obtained. If the doctor is not satisfied with the first iterative result, then the second iterative step is selected within the valid parameter range, and then the second iterative processing is performed. If the iterative result at this time still does not meet the doctor's requirements, the iterative result with a higher degree of satisfaction is selected as the basis, and the iterative work is continued until the Nth iterative result that satisfies the doctor is obtained, and the corresponding image boundary is transmitted to the doctor's terminal for display. In this way, a parameter optimization method from coarse to fine, gradually approaching the optimal solution can be achieved, thereby improving the efficiency and quality of parameter optimization.
[0089] Example 5
[0090] Based on Example 1, the ultrasound image processing parameter optimization method, step 3, includes:
[0091] Step 31: down-sampling the ultrasound image by using the image processing pyramid to obtain a plurality of sets of sampling information and mark them in the image processing pyramid, and input the image boundary into the corresponding tower layer of the image processing pyramid in the order of iterative processing;
[0092] Step 32: reconstructing the image boundary in each tower layer by using the sampling information to obtain the down-sampling presentation information corresponding to each image boundary;
[0093] Step 33: correcting each image boundary contained in the ultrasound image by using the down-sampling presentation information to obtain the corresponding optimized image and display it.
[0094] The working principle and beneficial effects of the above technical solution are as follows: by performing multiple down-sampling processing on the ultrasound image, reconstructing the image boundary, obtaining its down-sampling presentation information, and then correcting the image boundary in the ultrasound image by using the down-sampling presentation information, the optimized image is generated and transmitted to the doctor terminal for display. In this way, the processing efficiency of the ultrasound image can be improved, and multiple down-sampling can be performed in a short time, the presentation information of the image boundary is clear, and a strong reference is provided for the doctor's diagnosis.
[0095] Embodiment 6
[0096] On the basis of Embodiment 5, the ultrasound image processing parameter optimization method further comprises:
[0097] The iterative order corresponding to each optimized image is marked respectively, and an image selection task is generated by combining a preset selection instruction and transmitted to the doctor terminal to wait for the doctor to select the target optimized image.
[0098] In this example, the preset selection instruction represents an instruction requesting the doctor to select the target optimized image.
[0099] The working principle and beneficial effects of the above technical solution are as follows: due to the need for the doctor's work to be thorough, in order to avoid errors in the diagnosis results caused by general processing, the selection task is generated for the doctor to select the target optimized image, and humanistic care is shown.
[0100] Embodiment 7
[0101] On the basis of Embodiment 1, the ultrasound image processing parameter optimization method, the step 4 comprises:
[0102] Step 41: selecting the corresponding target optimized image and target optimization parameter according to the selection information uploaded by the doctor;
[0103] Step 42: marking the parameter optimization result corresponding to each target optimization parameter in the target optimized image.
[0104] Step 43: Combine each of the target optimization parameters with the corresponding parameter optimization result to generate a target parameter set and transmit it to the doctor terminal for display.
[0105] The working principle and beneficial effects of the above technical solution are as follows: the parameter optimization result corresponding to each target optimization parameter is determined based on the target optimization image and target optimization parameters selected by the doctor, and then the target optimization parameters and parameter optimization results are combined to construct a target parameter set for the doctor's reference.
[0106] Example 8
[0107] Based on Example 1, the ultrasound image processing parameter optimization method further includes:
[0108] When the doctor has not uploaded the selection information, the boundary contour information of the optimized image is compared, and the corresponding target optimized image and parameter target parameter set are selected according to the comparison result and transmitted to the doctor terminal for display.
[0109] The working principle and beneficial effects of the above technical solution: In order to reduce the waiting time of subsequent patients, when the doctor has not selected information for a long time, the optimal target optimized image is automatically selected and transmitted to the doctor's terminal for display, and then the subsequent images can be processed in turn, which improves the intelligence and convenience of the method.
[0110] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if such changes and modifications fall within the scope of the claims and their equivalents, the present invention is intended to include such changes and modifications.
Claims
1. A method for optimizing ultrasound image processing parameters, characterized in that: include: Step 1: Identify several initialization parameters of the ultrasound image, determine the valid parameter range corresponding to each of the initialization parameters, and match a corresponding initial iteration step length for each of the initialization parameters within the valid parameter range; Step 2: iteratively adjusting the corresponding initialization parameters using the initial iteration step size to determine a plurality of image boundaries contained in the ultrasound image; Step 3: downsampling the ultrasound image to obtain downsampling presentation information corresponding to each image boundary, and generating an optimized image of the ultrasound image; Step 4: retrieve the corresponding target optimization image according to the selection information uploaded by the doctor, generate a target parameter set according to the target optimization parameters corresponding to the target optimization image, and transmit it to the doctor's terminal for display.
2. The ultrasonic image processing parameter optimization method according to claim 1, wherein: The step 1 comprises: Step 11: Acquire several basic parameters of the ultrasound image, and determine several initialization parameters of the ultrasound image based on parameter values corresponding to each of the basic parameters; Step 12: determining a valid parameter range corresponding to each of the initialization parameters according to the specification information of the ultrasound image; Step 13: Compile an iterative pyramid operation script based on the parameter values of the initialization parameters, and determine the initial iteration step length corresponding to each initialization parameter according to the valid parameter range.
3. The ultrasonic image processing parameter optimization method according to claim 2, wherein: The step 13 includes: Step 131: writing the initial iteration parameters corresponding to each of the initialization parameters into the running script to generate an image processing pyramid; Step 132: performing verification processing on the ultrasound image using the image processing pyramid to generate a verification image, and obtaining a processing ratio between a verification parameter set corresponding to the verification image and each of the initialization parameters; Step 133: adjusting the initial iteration step length according to the processing ratio until all the processing ratios are consistent, thereby obtaining the initial iteration step length corresponding to each initialization parameter.
4. The ultrasonic image processing parameter optimization method according to claim 1, wherein: The step 2 comprises: Step 21: performing a first iterative process on the initialization parameters using the initial iteration step size to obtain a corresponding first image boundary, performing enhancement and noise suppression on the first image boundary, and generating a first iteration result; Step 22: Analyze the doctor's first satisfaction level with the first iteration result according to the operation target uploaded by the doctor. When the satisfaction level is lower than the standard level, adjust the initial iteration step size within the valid parameter range to obtain a second iteration step size. Step 23: performing a second iteration on the first iteration result using the second iteration step size to generate a second iteration result, and determining whether a second satisfaction level corresponding to the second iteration is higher than the standard level; if not, taking the iteration result with a higher satisfaction level as the iteration basis for the next iteration; Step 24: Obtain the Nth iteration result with the Nth satisfaction level higher than the standard level, and filter the corresponding Nth image boundary to transmit to the doctor terminal for display.
5. The ultrasonic image processing parameter optimization method according to claim 1, wherein: The step 3 comprises: Step 31: downsampling the ultrasound image using an image processing pyramid to obtain a plurality of sets of sampling information and marking them in the image processing pyramid, and inputting the image boundaries into corresponding layers of the image processing pyramid in an iterative processing order; Step 32: reconstructing the image boundary using the sampling information in each tower layer to obtain down-sampling presentation information corresponding to each image boundary; Step 33: Utilize the downsampling presentation information to correct each image boundary contained in the ultrasound image, obtain a corresponding optimized image, and display the image.
6. The ultrasonic image processing parameter optimization method according to claim 5, characterized in that: Also includes: The iteration order corresponding to each optimized image is marked respectively, and an image selection task is generated in combination with a preset selection instruction and transmitted to the doctor terminal to wait for the doctor to select a target optimized image.
7. The method for optimizing ultrasound image processing parameters according to claim 1, wherein: The step 4 comprises: Step 41: Selecting the corresponding target optimization image and target optimization parameters according to the selection information uploaded by the doctor; Step 42: Mark the parameter optimization results corresponding to each target optimization parameter in the target optimization image; Step 43: Combine each of the target optimization parameters with the corresponding parameter optimization result to generate a target parameter set and transmit it to the doctor terminal for display.
8. The method for optimizing ultrasound image processing parameters according to claim 1, wherein: Also includes: When the doctor has not uploaded the selection information, the boundary contour information of the optimized image is compared, and the corresponding target optimized image and parameter target parameter set are selected according to the comparison result and transmitted to the doctor terminal for display.