Body weight prediction method and device, computer equipment and storage medium

Abstract

The invention relates to a weight prediction method and device, computer equipment and a storage medium. The method comprises the following steps: acquiring a medical image of a detection object; performing feature extraction processing on the medical image to obtain image features of the medical image; and calling a trained body weight prediction model, and processing the image features to obtain the predicted body weight of the detected object. According to the method, the incidence relation between the image features and the body weight is utilized, the body weight prediction model is combined for body weight prediction, and the accuracy and prediction efficiency of the determined body weight can be improved.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and particularly to a method, device, computer device, storage medium, and computer program product for predicting body weight. Background Art

[0002] Body weight is an important parameter for calculating the standardized uptake value (SUV) in the quantitative research of positron emission tomography (PET, a nuclear medicine imaging). When there is a situation of missing or incorrect filling of body weight in the clinical PET data acquisition scenario, it may affect the SUV quantitative calculation of PET images, thereby affecting normal clinical diagnosis and treatment. Therefore, a body weight assessment model is needed to accurately assess the body weight of a human body in whole-body PET / CT (Computed Tomography) data.

[0003] In the related art, the method adopted is the spinal index regression analysis method based on abdominal CT slice 2D images. According to the statistically significant correlation between somatometric parameters (such as height and weight) and various spinal indices, the corresponding spinal indices are calculated from abdominal CT images, and a multiple linear regression method is used to predict somatometric parameters such as human height and weight.

[0004] However, the spinal index mainly estimates body weight based on the morphological characteristics of the spinal vertebrae, while the body weight of a human body is mainly composed of fat, muscle, bone, and other tissues. Therefore, relying solely on the spinal index to estimate body weight may not comprehensively consider the composition of the body. Therefore, the error of the body weight predicted by the existing spinal index method is relatively large. Summary of the Invention

[0005] Based on this, in view of the technical problem that the error of the body weight predicted by the above method is relatively large, it is necessary to provide a method, device, computer device, computer-readable storage medium, and computer program product for predicting body weight.

[0006] In a first aspect, the present application provides a method for predicting body weight. The method includes:

[0007] Obtain a medical image of a detection object;

[0008] Perform feature extraction processing on the medical image to obtain image features of the medical image;

[0009] Call a trained body weight prediction model to process the image features to obtain the predicted body weight of the detection object.

[0010] In one embodiment, the feature extraction process of the medical image to obtain the image features of the medical image further includes:

[0011] Performing a feature extraction process on the medical image to obtain the features corresponding to each voxel in the medical image;

[0012] Filtering out the target features within a preset range from the features corresponding to each voxel; the preset range is determined based on the image features of the detection object;

[0013] Determining the target features as the image features of the medical image.

[0014] In one embodiment, before the feature extraction process of the medical image to obtain the image features of the medical image, it further includes:

[0015] Performing a removal process on the foreign objects in the medical image; the foreign objects refer to the objects other than the detection object in the medical image;

[0016] And / or, adjusting the voxel size of the medical image based on the voxel requirements of the weight prediction model to obtain a preprocessed medical image;

[0017] The feature extraction process of the medical image to obtain the image features of the medical image includes:

[0018] Performing a feature extraction process on the preprocessed medical image to obtain the image features of the medical image.

[0019] In one embodiment, before calling the trained weight prediction model to process the image features to obtain the predicted weight of the detection object, it further includes:

[0020] Determining the image type of the medical image; the image type includes a whole-body image and a non-whole-body image that meets preset conditions;

[0021] Calling the corresponding weight prediction model according to the image type.

[0022] In one embodiment, the medical image includes a computed tomography image; the feature extraction process of the medical image to obtain the image features of the medical image includes:

[0023] Performing a gray value extraction on the medical image to obtain the gray value of each voxel in the medical image;

[0024] Based on the gray value of each voxel, obtaining a computed tomography value histogram;

[0025] Determine the computer tomography value histogram as the image feature of the medical image.

[0026] In one embodiment, the medical image includes a magnetic resonance image; the extracting the image feature of the medical image by performing feature extraction processing on the medical image includes:

[0027] Extract the signal intensity value of each voxel in the medical image;

[0028] Based on the signal intensity value of each voxel, determine the magnetic resonance value of the medical image;

[0029] Determine the magnetic resonance value as the image feature of the medical image.

[0030] In one embodiment, the weight prediction model is trained in the following manner:

[0031] Obtain a sample data set; the sample data set includes sample medical images of sample objects and the actual weights of the sample objects;

[0032] Perform feature extraction processing on the sample medical images to obtain sample image features of the sample medical images;

[0033] Use the sample image features as input variables, the predicted weight as the output variable, and the actual weight as the supervision information to train the weight prediction model to be trained, and obtain a trained weight prediction model.

[0034] In a second aspect, the present application further provides a weight prediction device. The device includes:

[0035] An image acquisition module, configured to acquire a medical image of a detection object;

[0036] A feature extraction module, configured to perform feature extraction processing on the medical image to obtain the image feature of the medical image;

[0037] A weight prediction module, configured to call the trained weight prediction model to process the image feature to obtain the predicted weight of the detection object.

[0038] In one embodiment, the feature extraction module is further configured to perform feature extraction processing on the medical image to obtain the features corresponding to each voxel in the medical image; screen out target features within a preset range from the features corresponding to each voxel; the preset range is determined based on the image feature of the detection object; determine the target feature as the image feature of the medical image.

[0039] In one embodiment, the device further includes a preprocessing module, configured to remove foreign objects in the medical image; and / or, adjust the voxel size of the medical image based on the voxel requirements of the weight prediction model to obtain a preprocessed medical image; the foreign object refers to an object other than the detection object in the medical image.

[0040] The feature extraction module is further configured to perform feature extraction processing on the preprocessed medical image to obtain the image features of the medical image.

[0041] In one embodiment, the device further includes a model determination module, configured to determine the image type of the medical image; the image type includes a whole-body image and a non-whole-body image that meets preset conditions; and call a corresponding weight prediction model according to the image type.

[0042] In one embodiment, the medical image includes a computed tomography (CT) image; the feature extraction module is further configured to extract the gray value of the medical image to obtain the gray value of each voxel in the medical image; based on the gray value of each voxel, obtain a CT value histogram; and determine the CT value histogram as the image features of the medical image.

[0043] In one embodiment, the medical image includes a magnetic resonance (MR) image; the feature extraction module is further configured to extract the signal intensity value of each voxel in the medical image; based on the signal intensity value of each voxel, determine the MR value of the medical image; and determine the MR value as the image features of the medical image.

[0044] In one embodiment, the device further includes a model training module, configured to obtain a sample data set; the sample data set includes sample medical images of sample objects and the actual weights of the sample objects; perform feature extraction processing on the sample medical images to obtain sample image features of the sample medical images; use the sample image features as input variables, the predicted weight as the output variable, and the actual weight as supervision information to train the weight prediction model to be trained to obtain a trained weight prediction model.

[0045] In a third aspect, the present application further provides a computer device. The computer device includes a memory and a processor, the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

[0046] Obtain a medical image of a detection object;

[0047] Perform feature extraction processing on the medical image to obtain the image features of the medical image;

[0048] Call the trained weight prediction model to process the image features to obtain the predicted weight of the detection object.

[0049] Fourthly, the present application also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the following steps are implemented:

[0050] Obtain a medical image of a detection object;

[0051] Perform feature extraction processing on the medical image to obtain the image features of the medical image;

[0052] Call the trained weight prediction model to process the image features to obtain the predicted weight of the detection object.

[0053] Fifthly, the present application also provides a computer program product. The computer program product includes a computer program, and when the computer program is executed by a processor, the following steps are implemented:

[0054] Obtain a medical image of a detection object;

[0055] Perform feature extraction processing on the medical image to obtain the image features of the medical image;

[0056] Call the trained weight prediction model to process the image features to obtain the predicted weight of the detection object.

[0057] For the above weight prediction method, device, computer device, storage medium, and computer program product, after obtaining the medical image of the detection object, perform feature extraction processing on the medical image to obtain the image features of the medical image; call the trained weight prediction model to process the image features to obtain the predicted weight of the detection object. This method utilizes the correlation between the image features and the weight, and combines the weight prediction model for weight prediction, which can improve the accuracy and prediction efficiency of the determined weight. Description of the Drawings

[0058] Figure 1 It is a schematic flowchart of the weight prediction method in an embodiment;

[0059] Figure 2 It is a schematic diagram of a full-body image and a non-full-body image in an embodiment;

[0060] Figure 3 It is a schematic flowchart of the training steps of the weight prediction model in an embodiment;

[0061] Figure 4 It is a schematic flowchart of the weight prediction method in another embodiment;

[0062] Figure 5 is a structural block diagram of a body weight prediction device in an embodiment;

[0063] Figure 6 is an internal structure diagram of a computer device in an embodiment. Detailed implementation manners

[0064] In order to make the objectives, technical solutions, and advantages of the present application clearer and more understandable, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0065] In an embodiment, as Figure 1 shown, a body weight prediction method is provided. In this embodiment, an example is given where this method is applied to a terminal. It can be understood that this method can also be applied to a server, and can also be applied to a system including a terminal and a server, and is implemented through the interaction between the terminal and the server. Among them, the terminal can be, but is not limited to, various personal computers, laptop computers, smart phones, tablet computers, Internet of Things devices, and portable wearable devices. The Internet of Things devices can be smart speakers, smart TVs, smart air conditioners, smart in-vehicle devices, etc. The portable wearable devices can be smart watches, smart bracelets, head-mounted devices, etc. The server can be implemented by an independent server or a server cluster composed of multiple servers. In this embodiment, the method includes the following steps:

[0066] Step S110, obtain a medical image of a detection object.

[0067] Among them, the detection object refers to the object for which the body weight needs to be determined.

[0068] Among them, the medical image is an image containing three-dimensional volume data information of the detection object. The medical image can have multiple modalities. For example, the medical image can be a Computed Tomography (CT) image, a Magnetic Resonance (MR) image, an attenuation coefficient image, an infrared image, and a CT scout image, etc. Among them, the attenuation coefficient map is an image used to describe the absorption capacity differences of different tissues and structures in a medical image. It reflects the attenuation degree of X-rays or magnetic fields when passing through different tissues and structures in imaging technologies such as X-ray, CT, and MRI. Different tissues and structures have different absorption capacities for X-rays or magnetic fields, so they show different gray values or signal intensities in the image. Among them, the CT scout map refers to the positioning image in Computed Tomography (CT scan), which is usually used to determine the scanning position and range of the detection object. The CT scout map can also be called a positioning image or a positioning scan.

[0069] In specific implementation, medical images of different modalities are obtained by different methods. Specifically, medical images of a detection object can be acquired by medical devices, or medical images acquired by medical devices can be processed to obtain medical images of the detection object.

[0070] For example, for computed tomography images (CT images), they can be acquired by computed tomography devices (CT devices). For magnetic resonance images, they can be acquired by magnetic resonance devices. For CT scout images, they can be reconstructed based on the projection data of X-rays passing through the detection object at different angles. For attenuation coefficient maps, after acquiring CT images, based on the gray values of the CT images, the attenuation coefficients of different tissues and structures can be calculated; and attenuation coefficient maps can be generated according to the attenuation coefficients.

[0071] Step S120: Perform feature extraction processing on the medical image to obtain the image features of the medical image.

[0072] Among them, the image features are features that can characterize the density information of the detection object. For example, the image features can be HU histograms (Hounsfiled Unit histograms), can be statistical features (such as height, body surface area, etc.), can also be Grey Level Co-occurrence Matrix (GLCM) features, or can be the depth features of the image.

[0073] Among them, the HU histogram is a method for statistically analyzing the pixel gray levels in CT images or X-ray images, and can describe the gray distribution of the image by statistically counting the number of voxels at different gray levels in the medical image. In the HU histogram, the X-axis represents different HU values, and the Y-axis represents the number of voxels corresponding to the HU values. By analyzing the HU histogram, the gray feature information of the image can be obtained.

[0074] Among them, each element in the gray level co-occurrence matrix represents the co-occurrence frequency of a specific gray level between two voxels in the image.

[0075] Specifically, for different modalities of image features, the corresponding feature extraction methods will also be different. For example, when the extracted image feature is a gray level co-occurrence matrix feature, the neighborhood size and direction can be determined first, and then each voxel in the medical image is compared with the voxels in its neighborhood, and the frequency of occurrence of different gray levels is statistically counted. Specifically, for each voxel, the frequency of its adjacent voxels having a specific gray value is statistically counted, and thus the gray level co-occurrence matrix is obtained. Another example is that when the extracted image feature is a depth feature, a convolutional neural network can be used for extraction.

[0076] Step S130: Invoke the trained weight prediction model to process the image features and obtain the predicted weight of the detection object.

[0077] Among them, the weight prediction model can adopt a machine learning regression model. For example, the weight prediction model can be a Multilayer-perception (MLP) regression model, a support vector regression model, a random forest regression model, etc. This application does not make specific limitations on this.

[0078] Among them, the weight prediction model can be trained through the sample medical images of the sample objects and the actual weights of the sample objects.

[0079] In specific implementation, the medical images of the sample objects can be obtained in advance as the sample medical images, and the actual weights of the sample objects can be obtained. Then, extract the image features of the sample medical images as the sample image features. Using the sample image features as the input variables, the predicted weight as the output variable, and the actual weight as the supervision information, train the weight prediction model to be trained to obtain the trained weight prediction model.

[0080] In the above weight prediction method, after obtaining the medical image of the detection object, perform feature extraction processing on the medical image to obtain the image features of the medical image; invoke the trained weight prediction model to process the image features to obtain the predicted weight of the detection object. This method utilizes the correlation between the image features and the weight, and combines the weight prediction model for weight prediction, which can improve the accuracy and prediction efficiency of the determined weight.

[0081] In an exemplary embodiment, the step of performing feature extraction processing on the medical image in step S120 to obtain the image features of the medical image further includes:

[0082] Perform feature extraction processing on the medical image to obtain the features corresponding to each voxel in the medical image;

[0083] From the features corresponding to each voxel, screen out the target features within a preset range; the preset range is determined based on the image features of the detection object;

[0084] Based on the target features, obtain the image features of the medical image.

[0085] In specific implementation, the feature extraction of the medical image is specifically to extract the features of each voxel in the medical image. During the acquisition process of the medical image, factors such as air and noise that are not the detection object are likely to be introduced to affect the image features. Therefore, after obtaining the features corresponding to each voxel in the medical image, the features corresponding to each voxel can be further filtered to eliminate the influence of factors such as air and noise that are not the detection object on the image features.

[0086] More specifically, after determining the type of image features to be extracted, the range corresponding to the detection object under this type of image features can be determined as the preset range. Based on this preset range, the features of each voxel of the extracted medical image are screened, and the features within the preset range are selected as the target features. This target feature is used as the image feature for subsequent input into the body weight prediction model.

[0087] For example, if the type of image feature to be extracted is the HU value, the upper limit value and the lower limit value of the range of the HU value of the detection object can be determined, and the preset range is obtained based on this upper limit value and lower limit value. From the HU values of each voxel of the extracted medical image, the HU values within this preset range are selected as the image features for subsequent input into the body weight prediction model.

[0088] In this embodiment, after performing feature extraction processing on the medical image to obtain the features corresponding to each voxel in the medical image, the preset range determined by the image features of the detection object is used to screen each of the extracted features, which can eliminate the influence of non-detection object factors such as air and noise on the image features, improve the accuracy of the determined image features, and thus improve the accuracy of the predicted body weight determined based on this image feature.

[0089] In an exemplary embodiment, before the above step S120 performs feature extraction processing on the medical image to obtain the image features of the medical image, it further includes: removing foreign objects in the medical image; and / or, adjusting the voxel size of the medical image based on the voxel requirements of the body weight prediction model to obtain a preprocessed medical image.

[0090] Correspondingly, step S120 performs feature extraction processing on the medical image to obtain the image features of the medical image, and further includes: performing feature extraction processing on the preprocessed medical image to obtain the image features of the medical image.

[0091] Among them, the foreign object refers to an object other than the detection object in the medical image. For example, the foreign object can be a bed board.

[0092] In a specific implementation, after obtaining the medical image of the detection object and before performing feature extraction processing on the medical image, the foreign objects in the medical image can be removed and / or the voxel size of the medical image can be adjusted. Specifically, to remove the foreign objects in the medical image, the objects other than the detection object existing in the medical image can be detected first, and then removed from the medical image. Adjusting the voxel size of the medical image can specifically be to adjust the voxel sizes of all voxels in the medical image to the same size to meet the requirements of the body weight prediction model. Thus, the preprocessed medical image is obtained, and further feature extraction is performed on the preprocessed medical image to obtain the image features of the medical image.

[0093] In one implementation, in addition to removing foreign objects from the medical image and unifying the voxel size, the medical image can also be subjected to image enhancement processing, normalization processing, and regularization processing to improve the quality of the medical image.

[0094] It can be understood that if foreign objects such as a bed board are not removed, since the density of the bed board is fixed and occupies a relatively stable gray level distribution, that is, the weight occupied by the bed board is relatively stable, therefore, the bed board correction of the body weight can be corrected using a linear model. And the body weight prediction model adopted in this application can also be regarded as a generalized linear model. Therefore, the bed board body weight will be fitted and corrected during the training process of the body weight prediction model. Thus, the foreign object correction processing of the medical image will also be realized during the training process of the body weight prediction model.

[0095] In this embodiment, by removing foreign objects in the medical image and adjusting the voxel size of the medical image, the quality of the preprocessed medical image is improved, thereby improving the accuracy of the image features extracted based on the preprocessed medical image. Furthermore, the accuracy of the predicted body weight determined based on the image features is improved. For example, referring to Figure 2 , such as Figure 2 as shown in (a) of Figure 2As shown in (b) of , if foreign objects such as a bed board in the medical image are not removed, the predicted weight is 74.15 kg, and the error between the predicted weight and the actual weight of 75 kg is 75 kg - 74.15 kg = 0.85 kg. If foreign objects such as a bed board in the medical image are removed, the predicted weight is 75 kg, and the error between the predicted weight and the actual weight of 75 kg is 75 kg - 75 kg = 0 kg. It can be seen that the accuracy of the predicted weight obtained after foreign object processing of the medical image is significantly higher than that of the predicted weight obtained without foreign object processing.

[0096] In one exemplary embodiment, before calling the trained weight prediction model in step S130 above to process the image features and obtain the predicted weight of the detection object, it further includes: determining the image type of the medical image; and calling the corresponding weight prediction model according to the image type.

[0097] Among them, the image type includes the full-body image of the detection object and the non-full-body image that meets the preset conditions. For example, as Figure 2 shown, Figure 2 in (a) of is a schematic diagram of a non-full-body image, Figure 2 in (b) of is a schematic diagram of a full-body image.

[0098] Specifically, the medical image can be the full-body image and the non-full-body image of the detection object. When it is a non-full-body image, it needs to meet the preset conditions. For example, the detection object included in the non-full-body image needs to have a main body part, and the integrity ratio is greater than the threshold, such as 80% or the like.

[0099] It can be understood that for different image types, the prediction accuracy will be different when training the weight prediction model. Specifically, the weight prediction accuracy based on the full-body image is higher than that based on the non-full-body image. Therefore, to ensure the accuracy of the predicted weight for the detection object, a weight prediction model can be trained separately for the full-body image and the non-full-body image. Before predicting the weight of the detection object, first determine the image type of the medical image of the detection object obtained, and call the corresponding weight prediction model according to the image type. Specifically, if the image type is a full-body image, then call the weight prediction model trained based on the sample medical image with the image type of the full-body image. If the image type is a non-full-body image, then call the weight prediction model trained based on the sample medical image with the image type of the non-full-body image.

[0100] In this embodiment, by training the corresponding weight prediction models for the full-body image and the non-full-body image respectively. When actually predicting the weight, according to the type of the medical image obtained, call the corresponding weight prediction model to predict the weight, which can further improve the accuracy of the predicted weight.

[0101] In an exemplary embodiment, the medical image includes a computed tomography (CT) image; in step S120, feature extraction processing is performed on the medical image to obtain the image features of the medical image, including: extracting the gray value of the medical image to obtain the gray value of each voxel in the medical image; generating a CT value histogram based on the gray value of each voxel; and determining the CT value histogram as the image features of the medical image.

[0102] In a specific implementation, when the medical image is a CT image, the corresponding image features are the CT histogram, specifically the HU histogram. Therefore, the feature extraction of the medical image can be performed according to the method for obtaining the HU histogram. Specifically, the gray value of the medical image can be extracted first to obtain the gray value of each voxel in the medical image. Further, based on the gray value of each voxel, the HU histogram is generated. The HU histogram is used as the image features of the medical image.

[0103] In this embodiment, by first extracting the gray value of each voxel in the medical image and generating a CT value histogram based on the gray value of each voxel as the image features of the medical image, the determination of the image features of the medical image when the medical image is a CT image is realized, so as to facilitate the subsequent body weight prediction model to process the image features and achieve accurate prediction of body weight.

[0104] In an exemplary embodiment, the medical image includes a magnetic resonance (MR) image; in step S120, the signal intensity value of each voxel in the medical image is extracted; based on the signal intensity value of each voxel, the MR value of the medical image is determined; and the MR value is determined as the image features of the medical image.

[0105] In a specific implementation, when the medical image is an MR image, the corresponding image features are the MR value. Therefore, the feature extraction of the medical image can be performed according to the method for obtaining the MR value. Specifically, the signal intensity value of each voxel in the medical image can be extracted first, and based on the signal intensity value of each voxel, the MR value of the medical image is determined as the image features of the medical image.

[0106] In this embodiment, by first extracting the signal intensity value of each voxel in the medical image and obtaining the MR value based on the signal intensity value of each voxel as the image features of the medical image, the determination of the image features of the medical image when the medical image is an MR image is realized, so as to facilitate the subsequent body weight prediction model to process the image features and achieve accurate prediction of body weight.

[0107] In an exemplary embodiment, as Figure 3 shown, the body weight prediction model in step S130 is trained in the following manner:

[0108] Step S310, obtain a sample data set; the sample data set includes sample medical images of sample objects and the actual weights of the sample objects;

[0109] Step S320, perform feature extraction processing on the sample medical images to obtain sample image features of the sample medical images;

[0110] Step S330, use the sample image features as input variables, the predicted weight as the output variable, and the actual weight as the supervision information to train the weight prediction model to be trained, and obtain a trained weight prediction model.

[0111] In specific implementation, sample medical images and actual weights of several sample objects can be obtained to form a sample data set. Image features of each sample medical image are respectively extracted as sample image features. In each training process, the image features of a sample medical image are input into the weight prediction model to be trained to obtain a predicted weight. Based on the predicted weight and the actual weight corresponding to the sample medical image, the weight prediction model to be trained is trained to obtain a trained weight prediction model.

[0112] More specifically, the loss value between the predicted weight and the actual weight corresponding to the sample medical image can be calculated. Specifically, a loss function can be used to calculate the loss value, and then the weight prediction model to be trained is trained with the aim of reducing the loss value to obtain a trained weight prediction model. Specifically, after a loss value is obtained by training with a set of sample data, if the loss value does not meet the preset accuracy, the hyperparameters of the weight prediction model are adjusted to obtain a new weight prediction model, and the new weight prediction model is trained with the next set of sample data to obtain a new loss value. If the new loss value still does not meet the preset accuracy, the hyperparameters of the weight prediction model are continuously adjusted, and so on, until the loss value meets the preset accuracy or reaches the preset number of iterations to obtain the optimal hyperparameters, which are used as the hyperparameters of the weight prediction model. The hyperparameters are assigned to the weight prediction model as the trained weight prediction model. Among them, the hyperparameters can include the initial learning rate, the number of hidden layers, the number of neurons, etc.

[0113] In this embodiment, the weight prediction model to be trained is trained through the sample medical images and actual weights of the sample objects to obtain a trained weight prediction model, so as to predict the weight of the detection object through the trained weight prediction model, which can improve the prediction efficiency and the accuracy of the prediction result.

[0114] In one embodiment, to facilitate the understanding of the embodiments of the present application by those skilled in the art, the following will be described with specific examples of the accompanying drawings. Refer to Figure 4 , which shows a specific flowchart of a weight prediction method. In this embodiment, the method includes the following steps:

[0115] Step S401: Obtain a medical image of the detection object;

[0116] Step S402: Remove foreign objects in the medical image; and / or, adjust the voxel size of the medical image based on the voxel requirements of the weight prediction model to obtain a preprocessed medical image;

[0117] Step S403: Perform feature extraction on the preprocessed medical image to obtain the features corresponding to each voxel in the medical image;

[0118] Step S404: Select target features within a preset range from the features corresponding to each voxel; the preset range is determined based on the image features of the detection object;

[0119] Step S405: Determine the target features as the image features of the medical image;

[0120] Step S406: Determine the image type of the medical image; according to the image type, call the corresponding weight prediction model; the image type includes a whole-body image and a non-whole-body image meeting preset conditions;

[0121] Step S407: Input the image features of the medical image into the weight prediction model to obtain the predicted weight of the detection object.

[0122] The weight prediction method provided by this application has a simple principle, high evaluation accuracy and strong interpretability, and can be used for accurate weight evaluation of various special cases. For example, patients with abnormal Body Mass Index (BMI), amputees, metal artifacts, and infant data. And through experiments, it is found that the weight prediction model provided by this application has an error less than 2.5 kg in non-whole-body CT data, and its relative impact on the standardized uptake value (SUV) of the predicted weight does not exceed 5%. For example, referring to Figure 2 the data shown in (a) in, the error between the predicted weight and the actual weight is: 75 - 74.71 = 0.29 kg. In the whole-body CT dataset, the error between the predicted weight and the actual weight is less than 1 kg. For example, referring to Figure 2 the data shown in (b) in, the error between the predicted weight and the actual weight is: 75 - 75 = 0 kg.

[0123] It should be understood that although the steps in the flowcharts involved in the above-described embodiments are sequentially shown according to the indications of the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless there is a clear indication in this article, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-described embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily executed at the same moment, but can be executed at different moments, and the execution order of these steps or stages is not necessarily sequential, but can be executed alternately or in turn with at least a part of other steps or steps or stages in other steps.

[0124] Based on the same inventive concept, an embodiment of the present application further provides a weight prediction device for implementing the above-mentioned weight prediction method. The solution provided by this device to solve the problem is similar to the solution described in the above method. Therefore, the specific limitations in one or more embodiments of the weight prediction device provided below can refer to the limitations on the weight prediction method in the above text, and will not be repeated here.

[0125] In one embodiment, as Figure 5 shown, a weight prediction device is provided, including: an image acquisition module 510, a feature extraction module 520, and a weight prediction module 530, where:

[0126] The image acquisition module 510 is configured to acquire a medical image of a detection object;

[0127] The feature extraction module 520 is configured to perform feature extraction processing on the medical image to obtain the image features of the medical image;

[0128] The weight prediction module 530 is configured to call a trained weight prediction model to process the image features to obtain the predicted weight of the detection object.

[0129] In one of the embodiments, the feature extraction module 520 is further configured to perform feature extraction processing on the medical image to obtain the features corresponding to each voxel in the medical image; screen out the target features within a preset range from the features corresponding to each voxel; the preset range is determined based on the image features of the detection object; and determine the target features as the image features of the medical image.

[0130] In one of the embodiments, the device further includes a preprocessing module, configured to remove foreign objects in the medical image; and / or, adjust the voxel size of the medical image based on the voxel requirements of the weight prediction model to obtain a preprocessed medical image; the foreign object refers to an object other than the detection object in the medical image;

[0131] The feature extraction module 520 is further configured to perform feature extraction processing on the preprocessed medical image to obtain the image features of the medical image.

[0132] In one embodiment, the apparatus further includes a model determination module, configured to determine the image type of the medical image; the image type includes a whole-body image and a non-whole-body image that meets a preset condition; and according to the image type, call a corresponding body weight prediction model.

[0133] In one embodiment, the medical image includes a computed tomography (CT) image; the feature extraction module 520 is further configured to extract the gray value of the medical image to obtain the gray value of each voxel in the medical image; based on the gray value of each voxel, obtain a CT value histogram; and determine the CT value histogram as the image features of the medical image.

[0134] In one embodiment, the medical image includes a magnetic resonance (MR) image; the feature extraction module 520 is further configured to extract the signal intensity value of each voxel in the medical image; based on the signal intensity value of each voxel, determine the MR value of the medical image; and determine the MR value as the image features of the medical image.

[0135] In one embodiment, the apparatus further includes a model training module, configured to obtain a sample data set; the sample data set includes the sample medical images of sample objects and the actual body weights of the sample objects; perform feature extraction processing on the sample medical images to obtain the sample image features of the sample medical images; use the sample image features as input variables, the predicted body weight as an output variable, and the actual body weight as supervision information to train the body weight prediction model to be trained, and obtain a trained body weight prediction model.

[0136] Each module in the above body weight prediction apparatus can be implemented in whole or in part by software, hardware, and their combination. The above modules can be embedded in the processor in the computer device in the form of hardware or be independent of the processor, or can be stored in the memory in the computer device in the form of software, so as to facilitate the processor to call and execute the operations corresponding to the above modules.

[0137] In one embodiment, a computer device is provided. The computer device can be a terminal, and its internal structure diagram can be as Figure 6As shown in the figure. The computer device includes a processor, a memory, a communication interface, a display screen, and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, a mobile cellular network, NFC (Near Field Communication), or other technologies. When the computer program is executed by the processor, it implements a body weight prediction method. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a button, a trackball, or a touchpad provided on the housing of the computer device, or an external keyboard, touchpad, or mouse, etc.

[0138] Those skilled in the art can understand that Figure 6 the structure shown in the figure is only a block diagram of some structures related to the solution of this application, and does not constitute a limitation on the computer device to which the solution of this application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine some components, or have different component arrangements.

[0139] In one embodiment, a computer device is further provided, including a memory and a processor. A computer program is stored in the memory, and when the processor executes the computer program, the steps in the above method embodiments are implemented.

[0140] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by the processor, the steps in the above method embodiments are implemented.

[0141] In one embodiment, a computer program product is provided, including a computer program. When the computer program is executed by the processor, the steps in the above method embodiments are implemented.

[0142] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use, and processing of relevant data need to comply with the relevant laws, regulations, and standards of relevant countries and regions.

[0143] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, database, or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memories. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc. The databases involved in the embodiments provided in the present application can include at least one of relational databases and non-relational databases. Non-relational databases can include distributed databases based on blockchain, etc., without limitation. The processors involved in the embodiments provided in the present application can be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, data processing logics based on quantum computing, etc., without limitation.

[0144] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope recorded in this specification.

[0145] The above-described embodiments merely represent several implementation manners of the present application. The description is relatively specific and detailed, but it should not be construed as a limitation on the patent scope of the present application. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present application, several modifications and improvements can still be made, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the appended claims.

Claims

1. A method for predicting body weight, characterized in that, the method comprises: obtaining a medical image of a detection object; performing feature extraction processing on the medical image to obtain image features of the medical image; invoking a trained body weight prediction model to process the image features to obtain the predicted body weight of the detection object.

2. The method according to claim 1, characterized in that, the performing feature extraction processing on the medical image to obtain image features of the medical image includes: performing feature extraction processing on the medical image to obtain features corresponding to each voxel in the medical image; screening out target features within a preset range from the features corresponding to each voxel; the preset range is determined based on the image features of the detection object; determining the target features as the image features of the medical image.

3. The method according to claim 1, characterized in that, before performing feature extraction processing on the medical image to obtain image features of the medical image, it further includes: performing removal processing on foreign objects in the medical image; the foreign objects refer to objects other than the detection object in the medical image; and / or, performing adjustment processing on the voxel size of the medical image based on the voxel requirements of the body weight prediction model to obtain a preprocessed medical image; the performing feature extraction processing on the medical image to obtain image features of the medical image includes: performing feature extraction processing on the preprocessed medical image to obtain image features of the medical image.

4. The method according to claim 1, characterized in that, before invoking a trained body weight prediction model to process the image features to obtain the predicted body weight of the detection object, it further includes: determining the image type of the medical image; the image type includes a whole body image and a non-whole body image meeting preset conditions; invoking a corresponding body weight prediction model according to the image type.

5. The method according to any one of claims 1-4, characterized in that, the medical image includes a computed tomography image; the performing feature extraction processing on the medical image to obtain image features of the medical image includes: extracting the gray value of the medical image to obtain the gray value of each voxel in the medical image; obtaining a computed tomography value histogram based on the gray value of each voxel; determining the computed tomography value histogram as the image features of the medical image.

6. The method according to any one of claims 1-4, characterized in that, the medical image includes a magnetic resonance image; the performing feature extraction processing on the medical image to obtain image features of the medical image includes: extracting the signal intensity value of each voxel in the medical image; determining the magnetic resonance value of the medical image based on the signal intensity value of each voxel; determining the magnetic resonance value as the image features of the medical image.

7. The method according to claim 1, characterized in that, the body weight prediction model is trained through the following method: Obtain a sample data set; the sample data set includes sample medical images of sample objects and the actual weights of the sample objects; Perform feature extraction processing on the sample medical images to obtain sample image features of the sample medical images; Use the sample image features as input variables, the predicted weight as the output variable, and the actual weight as the supervision information to train the weight prediction model to be trained, and obtain a trained weight prediction model.

8. A weight prediction device, Characterized in that, The device includes: An image acquisition module for acquiring medical images of a detection object; A feature extraction module for performing feature extraction processing on the medical images to obtain image features of the medical images; A weight prediction module for calling the trained weight prediction model to process the image features to obtain the predicted weight of the detection object.

9. A computer device, comprising a memory and a processor, the memory stores a computer program, Characterized in that, When the processor executes the computer program, the steps of the weight prediction method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium, on which a computer program is stored, Characterized in that, When the computer program is executed by a processor, the steps of the weight prediction method according to any one of claims 1 to 7 are implemented.

Images