High-quality ultrasound image acquisition method based on artificial intelligence

Abstract

The present invention discloses a method for acquiring high-quality ultrasound images based on artificial intelligence, and relates to the technical field of ultrasound image processing. The method extracts characteristic parameters of ultrasound images of the same diseased lesion site, determines an ultrasound image quality evaluation index, classifies image feature types and performs forward normalization processing on the characteristic parameters, calculates the quantitative value of the ultrasound image quality evaluation index, constructs a neural network model using detection process parameters as input samples and the quantitative value of the ultrasound image quality evaluation index as output samples, obtains a correlation relationship from the detection process parameter samples to the quantitative value samples of the evaluation index through the constructed neural network model, calculates the quantitative value of the quality evaluation index of the ultrasound image in the current detection process in real time, determines the optimal value of each detection process parameter through the obtained correlation relationship, and adjusts the detection process parameters until the optimal value is reached, thereby acquiring an ultrasound image of the current diseased lesion site.

Description

Technical Field

[0001] The present invention relates to the technical field of ultrasonic image processing, and in particular to a method for acquiring high-quality ultrasonic images based on artificial intelligence. Background Art

[0002] During ultrasound examination, the physician places the ultrasound probe on the outer surface of the lesion site on the patient's body. During the examination process, the physician needs to rely on his or her clinical experience to judge the quality of the obtained ultrasound image, and then appropriately adjust the pressure and tilt angle of the probe to obtain a satisfactory ultrasound image.

[0003] Because ultrasound image quality is difficult to quantitatively assess, differences in physician experience can lead to significant discrepancies in diagnostic results. Furthermore, changes in the ultrasound probe's acquisition state and environmental conditions can lead to misleading changes in the acquired ultrasound image. For example, even if the patient's anatomy remains unchanged, variations in the contact force and angle between the ultrasound probe and the patient's body surface can lead to significant differences in ultrasound imaging results, making it difficult for physicians to obtain optimal, high-quality ultrasound images. To address this, we propose an artificial intelligence-based method for acquiring high-quality ultrasound images. Summary of the Invention

[0004] The main purpose of the present invention is to provide a high-quality ultrasound image acquisition method based on artificial intelligence, which can effectively solve the problems in the background technology.

[0005] In order to achieve the above purpose, the technical solution adopted by the present invention is:

[0006] High-quality ultrasound image acquisition methods based on artificial intelligence, including:

[0007] Acquire multiple ultrasound images of the same lesion and extract the characteristic parameters of each ultrasound image ,in, Expressed as the jth feature parameter of the kth ultrasound image, the ultrasound image quality evaluation index is determined. According to the correlation between the feature parameter and the quality evaluation index, the image feature type is divided into four categories: positive feature, negative feature, interval feature, and threshold feature.

[0008] The ultrasound image quality evaluation index is any one of image clarity and image uniformity;

[0009] The principle of image feature type classification is:

[0010] If the quantitative value of the quality evaluation index tends to the optimal value when the value of the rth feature parameter increases, then the feature is a positive feature, where r∈m;

[0011] If the quantitative value of the quality evaluation index tends to the worst value when the value of the rth feature parameter increases, then the feature is a reverse feature;

[0012] If the quantitative value of the quality evaluation index tends to the optimal value when the value of the rth feature parameter tends to a certain interval, then the feature is an interval feature;

[0013] If the quantitative value of the quality evaluation index tends to the optimal value when the value of the rth feature parameter tends to a fixed value, then the feature is a threshold-type feature;

[0014] The characteristic parameters of the reverse feature, interval feature, and threshold feature are forward processed to calculate the quantitative value of the k-th ultrasound image quality evaluation index , using the quantitative value of ultrasound image quality evaluation index Construct a feature parameter set Y, where ; Among them, the forward processing formula is:

[0015] For reverse type characteristic parameters: ;

[0016] For threshold-type feature parameters: ;

[0017] For interval-type characteristic parameters: ;

[0018] Where, Represents the data value before normalization; It represents the data value after positive transformation; M represents the maximum value of the data before positive transformation; m represents the minimum value of the data before positive transformation; 、 They represent the lower and upper limits of the data interval of the intermediate parameters before normalization; 、 They represent the lower and upper limits of the optimal data interval for the intermediate parameters respectively;

[0019] Quantitative value of the kth ultrasound image quality evaluation index The calculation formula is:

[0020]

[0021] Where, is the distance between the characteristic parameters of the kth ultrasound image and the optimal characteristic parameters, ; is the distance between the characteristic parameter of the kth ultrasound image and the worst characteristic parameter, ; is the weight of the jth feature, m is the feature type, and ; 、 are the optimal and worst values of the jth feature parameter in k ultrasound images respectively; is the normalized value of the jth characteristic parameter of the kth ultrasound image, and the normalization formula is: , n is the total number of ultrasound images, k=1,2,...,n;

[0022] The detection process parameters during the ultrasound image acquisition process are obtained, and the feature parameter set X is constructed based on the obtained image feature parameters, where: , denoted as the i-th detection process parameter in the k-th ultrasound image acquisition process, wherein the detection process parameters include probe parameters and environmental parameters, wherein the probe parameters include the tilt angle and positive pressure value between the ultrasound probe and the patient's body; and the environmental parameters include the temperature and humidity of the environment in which the ultrasound probe is located.

[0023] The obtained detection process parameters are used as input samples, and the quantitative values of ultrasound image quality evaluation indicators are used as A neural network model is constructed for the output sample. The neural network model is trained and the number of neurons in the hidden layer of the neural network model is adjusted according to the training results until the prediction result reaches the expected value. The neural network model is a BP neural network model with a three-layer structure of input layer, middle layer and output layer. The number of neurons in the middle layer is determined by the following formula:

[0024]

[0025] Where s represents the number of neurons in the middle layer; p represents the number of neurons in the input layer; q represents the number of neurons in the output layer; Expressed as a pair The value of is rounded up; a is a constant coefficient, and a is an integer in the interval [1,9]; the number of neurons in the input layer is equal to the type of detection process parameters; in the neural network model constructed in this scheme, p=4; q=1;

[0026] The calculation formula of the expected value is:

[0027]

[0028] in, Expressed as expected value; Represents the number of neural network input samples; Represented as the output function of the neural network; Represented as the neural network output samples;

[0029] The constructed neural network model is used to obtain the correlation relationship between the detection process parameter samples and the quantitative value samples of the evaluation index, where the expression of the correlation relationship is: , It is represented as the i-th detection process parameter in the ultrasound image acquisition process;

[0030] Real-time calculation of the quantitative value of the quality evaluation index of the ultrasound image in the current detection process, through the obtained correlation relationship The optimal values of various detection process parameters are determined, and the detection process parameters are adjusted until the optimal values are reached, and an ultrasonic image of the current lesion focus is obtained.

[0031] The present invention has the following beneficial effects:

[0032] Compared with the existing technology, by acquiring multiple ultrasound images of the same lesion site, using convolutional neural network CNN to extract feature parameters of each ultrasound image, determining ultrasound image quality evaluation index, according to the correlation between feature parameters and quality evaluation index, the image feature types are divided into four categories: forward feature, reverse feature, interval feature, and threshold feature, the feature parameters of reverse feature, interval feature, and threshold feature are forward processed, the quantitative value of the k-th ultrasound image quality evaluation index is calculated, the detection process parameters in the ultrasound image acquisition process are acquired, and the feature parameter set is constructed based on the acquired image feature parameters, the acquired detection process parameters are used as input samples, and the quantized value of the ultrasound image quality evaluation index is used as output sample to construct a neural network model, and the neural network is constructed by The model is trained, and the number of neurons in the hidden layer of the neural network model is adjusted according to the training results until the prediction result reaches the expected value. The constructed neural network model obtains the correlation relationship from the detection process parameter samples to the quantitative value samples of the evaluation index, and the quantitative value of the quality evaluation index of the ultrasound image in the current detection process is calculated in real time. The optimal value of each detection process parameter is determined by the obtained correlation relationship, and the detection process parameters are adjusted to reach the optimal value, and the ultrasound image of the current diseased lesion site is obtained. By using artificial intelligence technology to analyze the impact of the ultrasound image quality, it is helpful to assist the detection doctor to adjust various parameters including probe pressure, probe angle, environmental parameters, etc. during the ultrasound detection process to the optimal state, so as to obtain high-quality ultrasound images under the optimal detection process parameter state. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 This is a flow chart of the method for acquiring high-quality ultrasound images based on artificial intelligence of the present invention;

[0034] Figure 2 A schematic diagram of the structure of the neural network model constructed in the solution of the present invention;

[0035] Figure 3 Schematic diagram of the ultrasonic probe during the detection process. DETAILED DESCRIPTION

[0036] The present invention will be further described below in conjunction with specific embodiments. The accompanying drawings are for illustrative purposes only and represent only schematic diagrams rather than actual drawings. They should not be understood as limiting the present invention. In order to better illustrate the specific embodiments of the present invention, some parts of the drawings may be omitted, enlarged or reduced, and do not represent the size of the actual product.

[0037] The specific implementation process of the technical solution of the present invention includes the following steps:

[0038] Step 1: Obtain multiple ultrasound images of the same lesion site and extract the characteristic parameters of each ultrasound image ,in, Expressed as the jth feature parameter of the kth ultrasound image;

[0039] The image feature parameters can be extracted using the convolutional neural network model CNN in the prior art. The main steps of the extraction process include:

[0040] Data preprocessing

[0041] Before using a convolutional neural network for image feature extraction, data preprocessing is required. First, the image data must be converted to a suitable format. Ultrasound images are typically formatted in DICOM. Image processing software can be used to convert DICOM ultrasound images into common formats, including JPEG and PNG. Second, the image must be normalized, converting pixel values to a range of 0-1 to facilitate better image processing by the network. Furthermore, data augmentation operations, such as random cropping, rotation, and flipping, are necessary to increase data diversity and robustness.

[0042] Building a convolutional neural network model

[0043] Building a convolutional neural network model is a key step in image feature extraction. A convolutional neural network consists of multiple convolutional layers, pooling layers, and fully connected layers. Convolutional layers extract local image features through convolution operations, pooling layers reduce the size of feature maps through downsampling, and fully connected layers perform classification by connecting all feature maps. When building the model, it is necessary to determine parameters such as the number of network layers, the size of the convolution kernel for each layer, and the activation function. The appropriate loss function and optimization algorithm should be selected based on the task requirements.

[0044] Network training and optimization

[0045] After building a convolutional neural network model, it needs to be trained and optimized. During training, the dataset is first divided into a training set and a validation set. The training set is used to train the model, and the validation set is used to evaluate model performance. Parameters are then adjusted and optimized based on the validation set results. During training, optimization algorithms such as batch gradient descent or stochastic gradient descent can be used to update the network density. Regularization methods such as LI regularization and L2 regularization can also be used to prevent overfitting.

[0046] Feature extraction and classification

[0047] After network training, the trained convolutional neural network model is used for feature extraction and classification. Feature extraction involves extracting high-level image features through convolutional and pooling layers. Common methods include using the output of the convolutional layer as a feature vector and using global semi-averaged pooling. Image classification is performed based on the extracted features, using classification algorithms such as support vector machines and logistic regression.

[0048] Model evaluation and tuning

[0049] After using a convolutional neural network to extract image features, the model needs to be evaluated and tuned. Model performance is evaluated using metrics such as accuracy, recall, and F1 score. If the model performance is unsatisfactory, optimization can be performed by adjusting the network structure, increasing the amount of training data, and adjusting hyperparameters.

[0050] Step 2: Determine the ultrasound image quality evaluation index. Based on the correlation between the feature parameters and the quality evaluation index, the image feature types are divided into four categories: positive feature, negative feature, interval feature, and threshold feature. The ultrasound image quality evaluation index can be either image clarity or image uniformity.

[0051] The principle of image feature type classification is:

[0052] If the quantitative value of the quality evaluation index tends to the optimal value when the value of the rth feature parameter increases, then the feature is a positive feature, where r∈m;

[0053] If the quantitative value of the quality evaluation index tends to the worst value when the value of the rth feature parameter increases, then the feature is a reverse feature;

[0054] If the quantitative value of the quality evaluation index tends to the optimal value when the value of the rth feature parameter tends to a certain interval, then the feature is an interval feature;

[0055] If the quantitative value of the quality evaluation index tends to the optimal value when the value of the rth feature parameter tends to a fixed value, then the feature is a threshold-type feature.

[0056] Step 3: Perform forward processing on the characteristic parameters of the reverse feature, interval feature, and threshold feature, and calculate the quantitative value of the k-th ultrasound image quality evaluation index , using the quantitative value of ultrasound image quality evaluation index Construct a feature parameter set Y, where ; Among them, the forward processing formula is:

[0057] For reverse type characteristic parameters: ;

[0058] For threshold-type feature parameters: ;

[0059] For interval-type characteristic parameters: ;

[0060] Where, Represents the data value before normalization; It represents the data value after positive transformation; M represents the maximum value of the data before positive transformation; m represents the minimum value of the data before positive transformation; 、 They represent the lower and upper limits of the data interval of the intermediate parameters before normalization; 、 They represent the lower and upper limits of the optimal data interval for the intermediate parameters respectively;

[0061] Quantitative value of the kth ultrasound image quality evaluation index The calculation formula is:

[0062]

[0063] Where, is the distance between the characteristic parameters of the kth ultrasound image and the optimal characteristic parameters, ; is the distance between the characteristic parameter of the kth ultrasound image and the worst characteristic parameter, ; is the weight of the jth feature, m is the feature type, and ; 、 are the optimal and worst values of the jth feature parameter in k ultrasound images respectively; is the normalized value of the jth characteristic parameter of the kth ultrasound image, and the normalization formula is: , n is the total number of ultrasound images, k=1,2,...,n;

[0064] It should be noted that the optimal and worst values of the characteristic parameters can be determined based on the correlation between the characteristic parameters and the quality evaluation indicators. That is, when the other characteristic parameters are fixed values, by changing the value of the characteristic parameter and obtaining the quantitative value of the corresponding quality evaluation indicator during the change process, when the quantitative value of the quality evaluation indicator takes the maximum value, the corresponding value of the characteristic parameter is the optimal value. Conversely, when the quantitative value of the quality evaluation indicator takes the minimum value, the corresponding value of the characteristic parameter is the worst value.

[0065] Step 4: Obtain the detection process parameters during the ultrasound image acquisition process, and construct a feature parameter set X based on the obtained image feature parameters, where: , It is represented as the i-th detection process parameter in the k-th ultrasound image acquisition process. The detection process parameters include probe parameters and environmental parameters. The probe parameters include the tilt angle and positive pressure value between the ultrasound probe and the patient's body, such as Figure 3 As shown; environmental parameters include the temperature and humidity of the environment where the ultrasound probe is located;

[0066] Step 5: Take the acquired detection process parameters as input samples and the quantitative values of ultrasound image quality evaluation indicators A neural network model is constructed for the output samples. The neural network model is trained and the number of neurons in the hidden layer of the neural network model is adjusted according to the training results until the prediction result reaches the expected value. The neural network model is a BP neural network model with a three-layer structure of input layer, middle layer and output layer. The number of neurons in the middle layer is determined by the following formula:

[0067]

[0068] In the formula, s represents the number of neurons in the middle layer; p represents the number of neurons in the input layer; q represents the number of neurons in the output layer. Expressed as a pair The value of is rounded up; a is a constant coefficient, and a is an integer in the interval [1,9]; wherein, the number of neurons in the input layer is equal to the type of detection process parameters; in the neural network model constructed in this scheme, p=4, q=1; the structural form of the neural network model constructed in this scheme is as follows Figure 2 As shown;

[0069] The formula for calculating the expected value is,

[0070]

[0071] in, Expressed as expected value; Represents the number of neural network input samples; Represented as the output function of the neural network; Represented as the neural network output samples;

[0072] Step 6: Obtain the association relationship between the detection process parameter samples and the quantitative value samples of the evaluation index through the constructed neural network model, where the expression of the association relationship is: , It is represented as the i-th detection process parameter in the ultrasound image acquisition process;

[0073] Step 7: Real-time calculation of the quantitative value of the quality evaluation index of the ultrasound image in the current detection process, through the obtained correlation relationship Determine the optimal value of each detection process parameter, and adjust the detection process parameters until the optimal value is reached, and obtain an ultrasound image of the current lesion site;

[0074] It should be noted that the detection process parameters include probe parameters and environmental parameters. The probe parameters include the tilt angle and positive pressure between the ultrasound probe and the patient's body, and the environmental parameters include the temperature and humidity of the environment in which the ultrasound probe is located. In actual situations, each detection process parameter has an upper limit and a lower limit.

[0075] Regarding the tilt angle between the ultrasound probe and the patient's body in the probe parameters, when the ultrasound probe is perpendicular to the patient's body surface, the tilt angle is 90 degrees, and when the ultrasound probe is parallel to the patient's body surface, the tilt angle is 0 degrees. The ultrasound probe can circle 360 degrees around the detection point. During actual operation, a positive direction can be set and the set positive direction can be used as a reference to obtain the tilt angle value of the ultrasound probe;

[0076] For the positive pressure value in the probe parameters, when the ultrasound probe is separated from the patient's body, the positive pressure value is 0. In actual operation, when the ultrasound probe contacts the patient's body, if the pressure is too great, it will cause discomfort to the patient. Therefore, the upper limit of the positive pressure value is artificially set. A questionnaire survey is conducted to select several volunteers for pressure tests. The average positive pressure that caused discomfort among the volunteers participating in the test is used as the upper limit of the positive pressure value, and the lower limit of the positive pressure value is 0.

[0077] Environmental parameters include the temperature and humidity of the environment where the ultrasonic probe is located. Generally, the environment during the detection process is an indoor environment, and the temperature and humidity of the environment where the ultrasonic probe is located can be controlled by adjusting the temperature and humidity of the indoor environment. When the optimal value of a certain detection process parameter is determined to be not within the range of its upper and lower limits, the optimal value of the detection process parameter is its upper or lower limit.

[0078] The basic principles, main features, and advantages of the present invention are shown and described above. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions are merely illustrative of the principles of the present invention. Various changes and modifications may be made to the present invention without departing from the spirit and scope of the present invention. Such changes and modifications are intended to fall within the scope of the present invention. The scope of protection claimed in the present invention is defined by the appended claims and their equivalents.

Claims

1. A high-quality ultrasound image acquisition method based on artificial intelligence, characterized in that: include: Acquire multiple ultrasound images of the same lesion and extract the characteristic parameters of each ultrasound image ,in, Expressed as the jth feature parameter of the kth ultrasound image, the ultrasound image quality evaluation index is determined. According to the correlation between the feature parameter and the quality evaluation index, the image feature types are divided into four categories: positive feature, negative feature, interval feature, and threshold feature. The characteristic parameters of the reverse feature, interval feature, and threshold feature are forward processed to calculate the quantitative value of the k-th ultrasound image quality evaluation index , using the quantitative value of ultrasound image quality evaluation index Construct a feature parameter set Y, where ; The detection process parameters during the ultrasonic image acquisition process are obtained, and the characteristic parameter set X is constructed based on the obtained detection process parameters, where: , It is represented as the i-th detection process parameter in the k-th ultrasound image acquisition process; The obtained detection process parameters are used as input samples, and the quantitative values of ultrasound image quality evaluation indicators are used as Constructing a neural network model for the output sample, training the neural network model, and adjusting the number of neurons in the hidden layer of the neural network model according to the training results until the prediction result reaches the expected value; The constructed neural network model is used to obtain the correlation relationship between the detection process parameter samples and the quantitative value samples of the evaluation index, where the expression of the correlation relationship is: , It is represented as the i-th detection process parameter in the ultrasound image acquisition process; Calculate the quantitative value of the quality evaluation index of the ultrasound image in the current detection process in real time, and obtain the correlation relationship Determining optimal values of various detection process parameters, and adjusting the detection process parameters until the optimal values are reached, obtaining an ultrasound image of the current lesion site; The detection process parameters include probe parameters and environmental parameters, wherein the probe parameters include the tilt angle and positive pressure value between the ultrasound probe and the patient's body; the environmental parameters include the temperature and humidity of the environment in which the ultrasound probe is located; The ultrasound image quality evaluation index is any one of image clarity and image uniformity; The principle of image feature type classification is: If the quantitative value of the quality evaluation index tends to the optimal value when the value of the rth feature parameter increases, then the feature is a positive feature, where r∈m; If the quantitative value of the quality evaluation index tends to the worst value when the value of the rth feature parameter increases, then the feature is a reverse feature; If the quantitative value of the quality evaluation index tends to the optimal value when the value of the rth feature parameter tends to a certain interval, then the feature is an interval feature; If the quantitative value of the quality evaluation index tends to the optimal value when the value of the rth feature parameter tends to a fixed value, then the feature is a threshold-type feature.

2. The method for acquiring high-quality ultrasound images based on artificial intelligence according to claim 1, characterized in that: Quantitative value of the kth ultrasound image quality evaluation index The calculation formula is: ； Where, is the distance between the characteristic parameters of the kth ultrasound image and the optimal characteristic parameters, ; is the distance between the characteristic parameter of the kth ultrasound image and the worst characteristic parameter, ; is the weight of the jth feature, m is the feature type, and ; 、 are the optimal and worst values of the jth feature parameter in k ultrasound images respectively; is the normalized value of the jth characteristic parameter of the kth ultrasound image, and the normalization formula is: , n is the total number of ultrasound images, k=1,2,...,n.

3. The method for acquiring high-quality ultrasound images based on artificial intelligence according to claim 1, characterized in that: The neural network model is a BP neural network model with a three-layer structure of an input layer, an intermediate layer, and an output layer, wherein the number of neurons in the intermediate layer is determined by the following formula: ； Where s represents the number of neurons in the middle layer; p represents the number of neurons in the input layer; q represents the number of neurons in the output layer; Expressed as a pair The value of is rounded up; a is a constant coefficient, and a is an integer in the interval [1,9]; the number of neurons in the input layer is equal to the type of detection process parameters.

4. The method for acquiring high-quality ultrasound images based on artificial intelligence according to claim 1, characterized in that: The forward processing formula is: For reverse type characteristic parameters: ; For threshold-type feature parameters: ; For interval-type characteristic parameters: ; Where, Represents the data value before normalization; It represents the data value after positive transformation; M represents the maximum value of the data before positive transformation; m represents the minimum value of the data before positive transformation; 、 They represent the lower and upper limits of the data interval of the intermediate parameters before normalization; 、 They represent the lower and upper limits of the optimal data interval for the intermediate parameters respectively.

5. The method for acquiring high-quality ultrasound images based on artificial intelligence according to claim 1, characterized in that: The calculation formula of the expected value is: ； in, Expressed as expected value; Represents the number of neural network input samples; Represented as the output function of the neural network; Represented as the neural network output samples.

Images