Feature combination strategy selection, state detection method and device, and electronic device

Abstract

The present disclosure relates to a method and apparatus for feature combination strategy selection, state detection, and an electronic device. The feature combination strategy selection method includes: obtaining dynamic image features and a state target of cerebral perfusion images in a cerebral perfusion image set; performing feature processing on the dynamic image features based on a preset feature selection method and at least two feature dimensionality reduction methods to obtain first selected features and at least two sets of first dimensionality-reduced features; performing feature combination processing on the first selected features and the first dimensionality-reduced features based on multiple combination strategies to obtain multiple combined features; evaluating the combined features using a classification model and the state target, and determining the combination strategy corresponding to the combined feature with the highest classification score as the optimal feature combination strategy for evaluating the state target. Embodiments of the present disclosure can improve the accuracy of feature selection and the state detection accuracy of cerebral perfusion images.

Description

Technical Field

[0001] The present disclosure relates to the field of image processing technology, and in particular to a feature combination strategy selection, state detection method and device, and electronic equipment. Background Art

[0002] Stroke has become the second leading cause of death in the world, with extremely high mortality and disability rates. Even if patients survive, they often suffer from varying degrees of neurological dysfunction, resulting in impaired quality of life, which places a heavy burden on patients, their families, and society. If stroke can be warned in time and the recovery of neurological function after treatment can be accurately evaluated, it will be conducive to early detection of stroke, selection of personalized treatment plans, and patient rehabilitation, thereby reducing the harm of stroke. Therefore, the abnormal detection of brain tissue status and accurate prediction of prognosis status are key factors in stroke treatment.

[0003] At present, blood flow parameters are one of the important parameters for evaluating the state of brain tissue in clinical practice. The supply of cerebral blood is distributed throughout the human brain. Once a part of the cerebral blood vessels is blocked and causes ischemia in the brain tissue area, the blood flow parameters in this area will be different from those of normal tissue. Therefore, cerebral blood flow parameters have been widely used in the diagnosis of diseases such as brain tumors, stroke, and Alzheimer's disease. Clinically, the maximum tissue residual function Tmax extracted from perfusion weighted imaging (PWI) can be used to detect ischemic stroke lesions (Tmax>6s), and the tissue area with relative cerebral blood flow (rCBF) less than 30% of the contralateral side can be called the core infarct area. It can be seen that the hemodynamic parameters (such as Tmax, CBF, etc.) obtained from medical images have been widely used in the detection of ischemic stroke lesions. However, since the perfusion sequence usually includes dozens of continuously scanned 3D images, the data volume is large, and few algorithms directly process the perfusion sequence. Usually, the dynamic parameters of the perfusion sequence are extracted for related research on stroke diseases. However, the hemodynamic parameters quantified by dynamic perfusion imaging are relatively static blood flow parameters, which are difficult to reflect the dynamic imaging characteristics during blood flow transmission. If the precise dynamic imaging characteristics related to stroke can be explored, it will inevitably provide a basis for accurate analysis of stroke research. Summary of the invention

[0004] The present invention proposes a feature combination strategy selection, state detection method and device, and electronic equipment. By establishing an optimal feature selection strategy, features adapted to the state target of stroke disease are obtained based on dynamic image features, thereby improving feature accuracy and the ability to accurately analyze stroke disease.

[0005] According to one aspect of the present disclosure, a method for selecting a feature combination strategy includes:

[0006] Obtain the dynamic image features and status targets of the brain perfusion images in the brain perfusion image set;

[0007] Perform feature processing on the dynamic image features based on a preset feature selection method and at least two feature dimensionality reduction methods to obtain a first selected feature and at least two sets of first dimensionality-reduced features;

[0008] Perform feature combination processing on the first selected feature and the first dimensionality-reduced features based on multiple combination strategies to obtain multiple combined features;

[0009] Use a classification model and the status target to evaluate the combined features, and determine the combination strategy corresponding to the combined feature with the highest classification score as the optimal feature combination strategy for evaluating the status target.

[0010] In some possible implementation manners, the performing feature processing on the dynamic image features based on a preset feature selection method and at least two feature dimensionality reduction methods to obtain a first selected feature and at least two sets of first dimensionality-reduced features includes:

[0011] Obtain the significant features in the dynamic image features;

[0012] Perform feature processing on the significant features based on a preset feature selection method and at least two feature dimensionality reduction methods to obtain a first selected feature and at least two sets of first dimensionality-reduced features;

[0013] and / or

[0014] The method further includes determining the preset feature selection method, which includes:

[0015] Based on at least two feature selection methods, screen out first features that meet the selection conditions of the feature selection methods from the significant features of the dynamic image features;

[0016] Use the classification model to evaluate the first features obtained by each of the feature selection methods, and determine the feature selection method corresponding to the first feature with the highest classification score as the preset feature selection method.

[0017] In some possible implementation manners, the performing feature combination processing on the first selected feature and the first dimensionality-reduced features based on multiple combination strategies to obtain multiple combined features includes at least one of the following manners:

[0018] Combine the first selected feature with each of the first dimensionality-reduced features respectively to obtain multiple combined features;

[0019] Performing feature selection on the first dimension-reduced feature by using the preset selection method to obtain a corresponding second dimension-reduced feature, and combining the first selected feature and the second dimension-reduced feature to obtain a plurality of combined features;

[0020] Evaluating the classification score of the first dimension-reduced feature for the state target by using the classification model, and obtaining a plurality of combined features based on the combination of the preset number of first dimension-reduced features with the highest scores and the first selected feature.

[0021] In some possible implementation manners, the evaluating the combined feature by using the classification model and the state target, and determining the combined strategy corresponding to the combined feature with the highest classification score as the optimal feature combination strategy includes:

[0022] Obtaining scores of a plurality of classification metrics of the combined feature by using the classification model;

[0023] Determining the classification score based on the scores of the plurality of classification metrics;

[0024] And / or

[0025] In the case where there are a plurality of classification models, determining the final classification score based on the mean value of the classification scores corresponding to each classification model.

[0026] In some possible implementation manners, the cerebral perfusion image includes cerebral images at multiple moments, and obtaining the dynamic image features of the cerebral perfusion images in the cerebral perfusion image set includes:

[0027] Determining the brain tissue region of the cerebral perfusion image;

[0028] Based on the multiple moments, respectively performing feature extraction processing on the brain tissue region at each moment to obtain moment image features at each moment;

[0029] For the cerebral perfusion image, respectively combining the moment image features of the brain tissue region at each moment to obtain the dynamic image features of the cerebral perfusion image.

[0030] In some possible implementation manners, the state target includes at least one of a first target indicating whether there is an ischemic stroke lesion, a second target indicating the degree of nerve function damage, and a third target indicating the prognosis state.

[0031] According to a second aspect of the present disclosure, there is provided a feature combination strategy selection device, which includes:

[0032] A first acquisition module, which acquires the dynamic image features and the state target of the cerebral perfusion images in the cerebral perfusion image set;

[0033] A feature processing module, configured to perform feature processing on the dynamic image features based on a preset feature selection method and at least two feature dimensionality reduction methods, to obtain first selected features and at least two groups of first dimensionality-reduced features;

[0034] A first combination module, configured to perform feature combination processing on the first selected features and the first dimensionality-reduced features based on multiple combination strategies, to obtain multiple combined features;

[0035] A first determination module, configured to evaluate the combined features by using a classification model and the state target, and determine the combination strategy corresponding to the combined feature with the highest classification score as the optimal feature combination strategy for evaluating the state target.

[0036] According to a third aspect of the present disclosure, there is provided a state detection method, including:

[0037] Obtaining a cerebral perfusion image;

[0038] Extracting dynamic features of the cerebral perfusion image;

[0039] Performing feature processing on the dynamic features according to a preset feature combination strategy, to obtain combined features, where the feature combination strategy is determined according to the feature combination strategy selection method described in the foregoing embodiment;

[0040] Determining the state corresponding to the cerebral perfusion image by using the combined features.

[0041] According to a fourth aspect of the present disclosure, there is provided a state detection device, including:

[0042] A second acquisition module, configured to acquire a cerebral perfusion image;

[0043] A second extraction module, configured to extract dynamic features of the cerebral perfusion image;

[0044] A second combination module, configured to obtain combined features of the cerebral perfusion image according to a preset feature combination strategy, where the feature combination strategy is determined by the feature combination strategy selection method described in the foregoing embodiment;

[0045] A second determination module, configured to determine the state corresponding to the cerebral perfusion image by using the combined features.

[0046] According to a fifth aspect of the present disclosure, there is provided an electronic device, including:

[0047] A processor;

[0048] A memory for storing instructions executable by the processor;

[0049] Wherein, the processor is configured to call the instructions stored in the memory to execute the method described in any one of the foregoing embodiments.

[0050] According to one aspect of the present disclosure, there is provided a computer-readable storage medium having computer program instructions stored thereon, and when the computer program instructions are executed by a processor, the methods described in the above embodiments are implemented.

[0051] In the embodiments of the present disclosure, a feature selection method and a dimensionality reduction method are used to process the dynamic features extracted from brain perfusion images to obtain first selected features and first dimensionality-reduced features, and different combination strategies are used to combine the obtained first selected features and first dimensionality-reduced features. Each combined feature is evaluated through a set classification model, and the combined feature that best matches the state target is selected, and then the optimal feature combination strategy corresponding to the state target is selected. The embodiments of the present disclosure can adaptively select the optimal combination strategy of dimensionality-reduced features and selected features according to different state targets for the study of stroke diseases, improve the accuracy of analysis, and provide better support for clinical analysis.

[0052] It should be understood that the above general description and the following detailed description are merely exemplary and explanatory, and do not limit the present disclosure.

[0053] Other features and aspects of the present disclosure will become clear from the following detailed description of exemplary embodiments with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] The accompanying drawings herein are incorporated into the specification and constitute a part of the specification. These drawings illustrate embodiments consistent with the present disclosure and, together with the specification, are used to explain the technical solutions of the present disclosure.

[0055] Figure 1 A flowchart showing a method for selecting a feature combination strategy according to an embodiment of the present disclosure;

[0056] Figure 2 A flowchart showing a method for obtaining dynamic image features of brain perfusion images according to an embodiment of the present disclosure;

[0057] Figure 3 A flowchart showing step S30 according to an embodiment of the present disclosure;

[0058] Figure 4 A flowchart showing a state detection method according to an embodiment of the present disclosure;

[0059] Figure 5 A block diagram showing a device for selecting a feature combination strategy according to an embodiment of the present disclosure;

[0060] Figure 6 A block diagram showing a state detection device according to an embodiment of the present disclosure;

[0061] Figure 7A block diagram of an electronic device 800 according to an embodiment of the present disclosure is shown;

[0062] Figure 8 A block diagram of another electronic device 1900 according to an embodiment of the present disclosure is shown. Detailed implementation manners

[0063] Various exemplary embodiments, features, and aspects of the present disclosure will be described in detail below with reference to the accompanying drawings. The same reference numerals in the drawings denote elements having the same or similar functions. Although various aspects of the embodiments are shown in the drawings, the drawings are not necessarily drawn to scale unless otherwise specified.

[0064] The term "exemplary" used herein means "serving as an example, embodiment, or illustration". Any embodiment described as "exemplary" herein is not necessarily to be construed as superior to or better than other embodiments.

[0065] The term "and / or" in this document merely describes an association relationship between associated objects and indicates that three relationships may exist. For example, A and / or B may represent: A exists alone, A and B exist simultaneously, and B exists alone. In addition, the term "at least one" in this document means any one of multiple or any combination of at least two of multiple. For example, including at least one of A, B, and C may represent including any one or more elements selected from the set composed of A, B, and C.

[0066] In addition, to better illustrate the present disclosure, numerous specific details are given in the following detailed implementation manners. Those skilled in the art should understand that the present disclosure can be implemented without some specific details. In some instances, methods, means, elements, and circuits well known to those skilled in the art are not described in detail so as to highlight the gist of the present disclosure.

[0067] The execution subject of the feature combination strategy selection method and the state target discrimination method according to the embodiments of the present disclosure may be an image processing device. For example, the image processing method may be executed by a terminal device, a server, or other processing devices. Among them, the terminal device may be a user equipment (UE), a mobile device, a user terminal, a terminal, a cellular phone, a cordless phone, a personal digital assistant (PDA), a handheld device, a computing device, a vehicle-mounted device, a wearable device, etc. In some possible implementation manners, it may also be implemented by a processor invoking computer-readable instructions stored in a memory.

[0068] Figure 1 A flowchart of a feature combination strategy selection method according to an embodiment of the present disclosure is shown, as Figure 1As shown, the feature combination strategy selection method includes:

[0069] S10: Obtain the dynamic image features and state targets of the brain perfusion images in the brain perfusion image set;

[0070] In some possible implementation manners, the brain perfusion image set may include multiple brain perfusion images (such as at least 100), and the type of the brain perfusion image may be at least one of magnetic resonance perfusion weighted imaging (PWI), computed tomography perfusion imaging (CTP), and arterial spin labeling perfusion imaging (ASL-MRI). Additionally, the brain perfusion image includes multiple groups of brain images, and each group of brain images can be used as the brain images scanned within a time range. The multiple groups of brain images in the brain perfusion image can be the brain images scanned at consecutive times (as multiple moments). Additionally, the brain perfusion images in the embodiments of the present disclosure can be the images of patients with brain diseases, such as the perfusion images of patients with ischemic stroke, or can also be the images of patients with glioma. The present disclosure does not specifically limit the type, but the disease types of the brain perfusion images in the perfusion image set are the same. For example, the present disclosure is a perfusion image set of brain ischemia patients. Additionally, the state targets in the embodiments of the present disclosure include at least one of a state target indicating whether there is an ischemic stroke lesion, a state target indicating the degree of nerve function damage, and a state target indicating the prognosis state.

[0071] The embodiments of the present disclosure can select the optimal feature combination strategy based on the above state targets, thereby improving the accuracy of the evaluation of the state targets.

[0072] Since the brain perfusion images in the embodiments of the present disclosure are brain images at multiple moments collected in a time series, image features can be extracted for each moment of the brain perfusion images respectively, and dynamic image features are formed by combining based on the moment information. By analyzing the dynamic image features, dynamic blood flow information related to the state target can be extracted.

[0073] S20: Perform feature processing on the dynamic image features based on a preset feature selection method and at least two feature dimension reduction methods to obtain a first selected feature and at least two groups of first dimension-reduced features;

[0074] In some possible implementation manners, a preset feature selection method and a dimension reduction method can be used to perform feature processing on the dynamic image features to obtain corresponding selected features and dimension-reduced features.

[0075] S30: Perform feature combination processing on the first selected feature and the first dimension-reduced features based on multiple combination strategies to obtain multiple combined features;

[0076] In some possible embodiments, various combination strategies may be adopted to combine the first selected feature and each first dimension-reduced feature to obtain corresponding combined features. Among them, the combined features at least include the first selected feature.

[0077] S40: Use the classification model and the state target to evaluate the combined features, and determine the combination strategy corresponding to the combined feature with the highest classification score as the optimal feature combination strategy for evaluating the state target.

[0078] In some possible embodiments, for each state target, the embodiments of the present disclosure may use a classification model to evaluate each combined feature, obtain an evaluation score (classification score) corresponding to the combined feature, and determine the combined feature with the highest classification score as the best combined feature for evaluating the state target. The corresponding feature combination strategy is the optimal feature combination strategy.

[0079] Based on the above configuration, in the embodiments of the present disclosure, a feature selection method and a dimension reduction method may be used to process the dynamic features extracted from cerebral perfusion images to obtain the first selected feature and the first dimension-reduced feature, and different combination strategies may be used to perform feature combination on the obtained first selected feature and the first dimension-reduced feature. Each combined feature is evaluated through a set classification model, and the combined feature that best matches the state target is selected, and then the optimal feature combination strategy corresponding to the state target is selected. The embodiments of the present disclosure can adaptively select the optimal combination strategy of dimension-reduced features and selected features according to different state targets for the study of stroke diseases, improve the accuracy of analysis, and provide better support for clinical analysis.

[0080] The embodiments of the present disclosure will be described in detail below with reference to the accompanying drawings. First, the embodiments of the present disclosure may obtain a cerebral perfusion image set, and the manner of obtaining the cerebral perfusion image set may include at least one of the following manners:

[0081] A1) Directly use a medical imaging acquisition device to acquire cerebral perfusion images; in the embodiments of the present disclosure, the medical imaging acquisition device may be a nuclear magnetic resonance device, but this is not a specific limitation of the present disclosure.

[0082] A2) Transmit and receive a cerebral perfusion image set including multiple cerebral perfusion images through an electronic device; the embodiments of the present disclosure may receive cerebral perfusion images transmitted by other electronic devices through a communication method, and the communication method may include wired communication and / or wireless communication, which are not specifically limited in the present disclosure.

[0083] A3) Read the cerebral perfusion images stored in the database; the embodiments of the present disclosure may read the cerebral perfusion images stored locally or on the server according to the received data reading instruction to obtain the cerebral perfusion image set, which is not specifically limited in the present disclosure.

[0084] It should be noted here that the cerebral perfusion images in the embodiments of the present disclosure can be perfusion images acquired by the same device or different devices. Those skilled in the relevant technical fields can select corresponding devices according to requirements, and no specific limitations are made here.

[0085] The embodiments of the present disclosure can perform bone removal processing on the cerebral perfusion images to obtain the brain tissue region in the cerebral perfusion images. The embodiments of the present disclosure aim to analyze the characteristics of the brain tissue region to obtain feature information matching the target state and corresponding feature combination strategies. Therefore, the position of the brain tissue region in the cerebral perfusion images can be extracted first. The brain tissue region can include gray matter, white matter, and cerebrospinal fluid. The embodiments of the present disclosure can use FSL software to perform bone removal processing on the cerebral perfusion images, so as to obtain the position information of the brain tissue region. Through bone removal processing, the influence of skull pixels in the cerebral perfusion images on subsequent processing can be avoided, and the accuracy of feature extraction can be improved.

[0086] In the case of obtaining the brain tissue region, the cerebral perfusion images can also be preprocessed, or subsequent processing can be directly performed according to the determined brain tissue region in the cerebral perfusion images.

[0087] Among them, the preprocessing of the cerebral perfusion images includes at least one of the following methods: performing registration processing on the time series of the cerebral perfusion images; performing smoothing processing on the time series of the cerebral perfusion images.

[0088] Performing registration on the time series of the cerebral perfusion images includes: performing rigid registration on the brain images at multiple moments in the cerebral perfusion images to eliminate the motion deviation generated by the examined object during the image acquisition process due to movement.

[0089] Performing smoothing processing on the time series of the cerebral perfusion images includes: obtaining the gray values of the same pixel points in the brain tissue at multiple moments to form a time-gray sequence; performing smoothing processing on the time-gray sequence. Among them, the smoothing processing can include three-time moving average processing. The embodiments of the present disclosure can use a 1*3 moving window to smooth the time-gray sequence, and the present disclosure does not make specific limitations on this. Through smoothing processing, the noise in the cerebral perfusion images can be reduced, and the image quality can be improved.

[0090] After preprocessing or obtaining the brain tissue region, feature processing can be further performed on the brain tissue region. First, the dynamic image features of the cerebral perfusion images can be extracted. In the embodiments of the present disclosure, the cerebral perfusion images can include brain images at multiple moments. Figure 2 The flowchart showing the acquisition of the dynamic image features of the cerebral perfusion images according to the embodiments of the present disclosure is shown. The acquisition of the dynamic image features of the cerebral perfusion images in the cerebral perfusion image set includes:

[0091] S11: Determine the brain tissue region of the cerebral perfusion image;

[0092] S12: Based on the multiple moments, perform feature extraction processing on the brain tissue region at each moment respectively to obtain the moment image features at each moment;

[0093] S13: For the cerebral perfusion image, combine the moment image features of the brain tissue region at each moment respectively to obtain the dynamic image features of the cerebral perfusion image.

[0094] In the embodiments of the present disclosure, the cerebral perfusion image may include t groups of brain images, and the t groups of brain images respectively correspond to t moments. For example, in the embodiments of the present disclosure, t may be an integer greater than 1 and less than or equal to 50, but it is not a specific limitation of the present disclosure. The embodiments of the present disclosure may perform feature extraction processing on the brain tissue regions in the brain images at t moments respectively.

[0095] In one example, the feature extraction processing may include: performing at least one image transformation on the brain tissue region, obtaining an augmented set of the brain tissue region based on the brain tissue region and its image transformation result; and extracting at least one of the first-order gradient feature, shape feature, and texture feature of any image in the augmented set. Wherein, the image transformation includes at least one of Fourier transform, Gabor transform, Gaussian-Laplacian transform, wavelet transform, square root filtering, and exponential function filtering. The embodiments of the present disclosure may use the original brain tissue region and the results after image transformation to form an augmented set, and perform feature extraction on each brain tissue region in the augmented set to obtain richer image features. The extracted first-order gradient features may include features describing single pixels or single voxels such as the gray mean value, maximum gray value, minimum gray value, variance, percentile (14 and 15) of the brain tissue region, skewness and kurtosis features describing the shape of the data intensity distribution, and histogram quotient and energy information. Among them, skewness reflects the asymmetry of the data distribution curve to the left (negative skew, below the mean) or to the right (positive skew, above the mean); while kurtosis reflects the tailing of the data distribution due to outliers relative to the Gaussian distribution. Shape features may include surface- and volume-based features such as compactness and sphericity features. Texture features may include Absolute Gradient, Gray-Level Co-occurrence Matrix (GLCM), Gray-Level Run-Length Matrix (GLRLM), Gray-Level Size Zone Matrix (GLSZM), Gray-Level Size Zone Matrix (GLSZM), and Gray-Level Dependence Matrix (GLDM).

[0096] In some possible embodiments, the above-mentioned feature extraction process may be performed by using the method of extracting radiomics to obtain the moment image features corresponding to the brain tissue regions at each moment. By combining the moment image features at each moment, the dynamic image features of the brain tissue region are obtained. In the embodiments of the present disclosure, 65,800 dynamic image features can be calculated (3D brain images at 50 moments × 1,316 moment image features). These dynamic image features are divided into 9 groups: (1) Shape features × 50 = 700, (2) First-order gradient features: 18 × 50 = 900, (3) Gray-level co-occurrence matrix GLCM (24 × 50 = 1,200), (4) Gray-level run length matrix GLRLM (16 × 50 = 800), (5) Gray-level size zone matrix GLSZM (16 × 50 = 800), (6) Neighboring gray tone difference matrix NGTDM (5 × 50 = 250), (7) Gray-level dependence matrix GLDM (14 × 50 = 700), (8) Laplace transform (465 × 50 = 23,250), (9) Wavelet transform (744 × 50 = 37,200). In the embodiments of the present disclosure, each moment image feature may be defined as the combination of the name of the radiomics feature itself and the moment value of the 3D brain image, where t is the time value corresponding to the 3D image. For example, "Log-sigma-1-0-mm-3d_firstorder_skewness_17" represents the moment image feature "Log-sigma-1-0-mm-3d_firstorder_skewness" at the 17th moment in the DSC-PWI brain perfusion image.

[0097] In addition, when performing feature extraction processing on the brain tissue region in the embodiments of the present disclosure, the time series can also be optimized to reduce the number of time values and improve the operation efficiency. Specifically, in the embodiments of the present disclosure, the time t of the cerebral perfusion image can be divided into three groups. For example, the first group is the preparation stage, the second group is the reaction stage, and the third group is the recovery stage. Among them, the preparation stage is the stage when the brain image is not affected by the contrast agent during the perfusion imaging process, the reaction stage is the stage when the contrast agent flows through the blood vessels and causes the gray value of the pixel points to change, and the recovery stage is the process when the gray value of the contrast agent leaving the pixel points returns to the initial state. In the embodiments of the present disclosure, t is 50 time points, where the first group is from time point 1 to 10, the second group is from time point 11 to 30, and the third group is from time point 31 to 50. The above is only an example of the present disclosure and is not a specific limitation. In the case of obtaining the three groups of time points, the mean processing can be performed on the brain images corresponding to the first group of time points, and the mean processing can also be performed on the brain images corresponding to the third group of time points. The brain images obtained by the mean processing of the first group, the brain images corresponding to the second group of time points, and the brain images obtained by the mean processing of the third group are used as the new cerebral perfusion images for feature extraction processing. Thus, while ensuring comprehensive information, the amount of calculation is reduced and the feature extraction efficiency is improved.

[0098] The above embodiments can perform image feature extraction on the brain tissue region of the three-dimensional brain image at each time point, so as to obtain dynamic image features at different times from the perspective of multi-time three-dimensional images.

[0099] In addition, in some other embodiments of the present disclosure, the features of each layer of the brain image in the cerebral perfusion image at time t can also be analyzed as a whole to obtain dynamic image features. Among them, the step of respectively extracting the dynamic image features of the brain tissue region based on the multiple time points may further include: generating first brain images respectively based on the brain images of the same layer at different time points, the number of the first brain images being the same as the number of layers of the brain images, and the number of layers of the first brain images being the same as the number of time points; performing feature extraction processing on the brain tissue region in the first brain images to obtain layer image features; and obtaining the dynamic image features of the cerebral perfusion image based on the combination of the layer image features of the brain tissue region for the cerebral perfusion image.

[0100] The brain images in the cerebral perfusion images of the embodiments of the present disclosure are 3D images, and each brain image has the same dimension. It may include multiple layers of brain images. The 3D image may include brain images in three directions, such as the coronal plane, the sagittal plane, and the transverse plane, and the images of each layer in each direction can be used as the object for feature extraction in the embodiments of the present disclosure.

[0101] In one example, the dimension of the brain perfusion image can be expressed as t*C*W*H, where t represents the number of time instances, C represents the number of layers of the brain image, and W and H respectively represent the width and height of the brain image. In the embodiments of the present disclosure, according to the order from the first layer to the C-th layer, t brain images at t time instances are extracted and combined into the first brain image. Each first brain image corresponds to the number of layers of the brain image, the number of the first brain images is the same as the number of layers of the brain image, and the number of layers of the first brain image is the same as the number of time instances. Therefore, for each brain perfusion image, the dimension of the obtained first brain image is t*W*H, and the number is C.

[0102] In the case of obtaining the first brain image, feature extraction processing can be performed on the brain tissue region in the first brain image to obtain layer image features. The feature extraction processing is the same as the configuration in the above embodiments, including: performing at least one image transformation on the brain tissue region, obtaining an augmented set of the brain tissue region based on the brain tissue region and the result of its image transformation; and extracting at least one of the first-order gradient feature, shape feature, and texture feature of any image in the augmented set. For each layer, after obtaining the first brain image and performing the feature extraction processing on the brain tissue region, the corresponding layer image features can be obtained. By combining the layer image features of each layer, dynamic image features can be obtained. When the number of layers of the brain image is 20, the number of dynamic image features can be 26320 (1316*20), but this is not a specific limitation of the present disclosure.

[0103] Similarly, in the embodiments of the present disclosure, the time instance value can be optimized before extracting the layer image features to reduce the time instance value, and the specific method refers to the foregoing embodiments.

[0104] Based on the above registration, the embodiments of the present disclosure can construct a three-dimensional image from the perspective of time instances, extract layer image features, and further enrich the extracted dynamic image features. In some other embodiments, the dynamic image features obtained in the embodiments of the present disclosure can be a combination of the above two types of dynamic image features, but this is not a specific limitation of the present disclosure.

[0105] In the case of obtaining rich dynamic image features, feature screening and feature dimensionality reduction can be further performed to obtain features that can accurately classify the state target. The embodiments of the present disclosure adopt various combination strategies to establish an effective combination between the selected features and the dimensionality-reduced features, and optimize the feature accuracy.

[0106] Figure 3 The flowchart of step S30 in the embodiments of the present disclosure is shown. Among them, the feature combination processing is performed on the first selected feature and the first dimensionality-reduced feature based on various combination strategies to obtain a plurality of combined features, including:

[0107] S31: Obtain the significant features in the dynamic image features;

[0108] S32: Perform feature processing on the significant features based on a preset feature selection method and at least two feature dimensionality reduction methods to obtain first selected features and at least two groups of first dimensionality-reduced features.

[0109] In some possible implementation manners, standardization processing may be first performed on the obtained dynamic image features to reduce the influence of the numerical span of the features themselves. Each row of the dynamic image features of the cerebral perfusion image set obtained in the embodiments of the present disclosure represents the feature values of different feature items of the same patient, and each column represents the feature values of the same feature of different patients. When performing feature standardization, standardization processing is respectively performed on each column of the dynamic image features. For example, the standardization processing in the embodiments of the present disclosure may be mean-variance standardization, so that the standardized features have a mean of 0 and a variance of 1. In other implementation manners, the ratio of each column of features to the maximum value of the column of features may also be used as the standardized feature value. Then, the feature processing may be performed using the standardized dynamic image features.

[0110] In some possible implementation manners, in the process of significant feature extraction in the embodiments of the present disclosure, on the one hand, dimensionality reduction processing of high-dimensional features is achieved, and on the other hand, the accuracy of the selected features is improved. In the embodiments of the present disclosure, significance analysis may be performed on each dynamic image feature through different states of the state target, and the p-value (assumed value) between different state groups of features is calculated. When the p-value is less than the significance threshold, it is determined that the feature is a significant feature. The significance threshold is 0.05, and the p-value calculation method includes a T-test. The above is only an exemplary illustration and does not constitute a specific limitation of the present disclosure.

[0111] In one example, when the state target is whether a patient has a stroke, the state target may include a first target indicating having a stroke and a second target indicating not having a stroke. The dynamic image features of the cerebral perfusion images corresponding to the first target and the dynamic image features of the cerebral perfusion images corresponding to the second target in the cerebral perfusion image set are grouped, and the significance value p-value (assumed value) between the same feature items in the two groups of features is calculated. When the p-value is less than the significance threshold, it is determined that the feature is a significant feature. In the embodiments of the present disclosure, the correlation coefficient between the two groups of features may also be calculated. When the correlation coefficient of the feature is higher than the coefficient threshold and the p-value is less than the significance threshold, it is determined that the feature is a significant feature. The coefficient threshold may be a value greater than 0.6.

[0112] In another example, when the status target indicates whether the cranial nerve function is damaged, the status target may include a first target indicating damaged nerve function (NIHSS score greater than 0) and a second target indicating undamaged nerve function (NIHSS score equal to 0). Group the dynamic image features of the cerebral perfusion images corresponding to the first target and the dynamic image features of the cerebral perfusion images corresponding to the second target in the cerebral perfusion image set, calculate the significance value p-value (assumed value) of the same feature item between the two groups of features. When the p-value is less than the significance threshold, determine that this feature is a significant feature. The embodiments of the present disclosure can also calculate the correlation coefficient between the two groups of features. When the correlation coefficient of the feature is higher than the coefficient threshold and the p-value is less than the significance threshold, determine that this feature is a significant feature. The coefficient threshold can be a value greater than 0.6.

[0113] In another example, when the status target represents the prognosis status of a stroke patient, the status target may include a first target indicating good prognosis (90daymRS < 2) and a second target indicating poor prognosis (90daymRS > 2). Group the dynamic image features of the cerebral perfusion images corresponding to the first target and the dynamic image features of the cerebral perfusion images corresponding to the second target in the cerebral perfusion image set, calculate the significance value p-value (assumed value) of the same feature item between the two groups of features. When the p-value is less than the significance threshold, determine that this feature is a significant feature. The embodiments of the present disclosure can also calculate the correlation coefficient between the two groups of features. When the correlation coefficient of the feature is higher than the coefficient threshold and the p-value is less than the significance threshold, determine that this feature is a significant feature. The coefficient threshold can be a value greater than 0.6.

[0114] When significant features are obtained, a first selected feature matching the status target can be selected using a preset feature selection method, and a first dimension-reduced feature can be obtained by performing feature dimension reduction on the significant features using a feature dimension reduction method.

[0115] Among them, the preset feature selection method can be specific information set in advance, or it can also be a feature selection method determined by comparing multiple selection methods. In the embodiments of the present disclosure, determining the preset feature selection method is also included, which includes: based on at least two feature selection methods, screening out the first features that meet the selection conditions of the feature selection method from the significant features of the dynamic image features; using the classification model to evaluate the first features obtained by each of the feature selection methods, and determining the feature selection method corresponding to the first feature with the highest classification score as the preset feature selection method.

[0116] Embodiments of the present disclosure can select the preset feature selection method from a variety of feature selection methods, and the selection principles of these various selection methods are different. In one example, the feature selection methods can include at least two of the methods based on information theory, methods based on similar features, methods based on statistical features, and methods based on sparse features and flow features. The methods based on information theory can include the maximum mutual information method (MIM), the conditional mutual information maximization method (CMIM), the conditional mutual information maximization method (MRMR), the best individual feature (BIF), the mutual information selection (MIFS), the joint mutual information (JMI), etc., and the methods based on similar features can include the distance separability measure (Fisher score algorithm), the Laplacian score (Lap score algorithm), the feature weight algorithm (ReliefF), the methods based on statistical features can include the Tscore algorithm and the Fscore algorithm, and the methods based on sparse features and flow features can include the multi-cluster feature selection algorithm (MCFS), the least absolute shrinkage and selection operator (Lasso), and the Alpha algorithm.

[0117] Embodiments of the present disclosure can use at least two of the above-mentioned feature selection methods to perform feature selection on the significant features of state targets with different target values. The selection conditions for feature selection methods other than the Lass algorithm can include: the maximum number of features is less than the feature quantity threshold, and the feature score is greater than the score threshold, where the feature quantity threshold is greater than 10, such as 20 set in the present disclosure, and the score threshold can be greater than 0.6, such as 0.8 set in the present disclosure. The selection condition for the Lasso algorithm is to select the feature terms with non-zero feature coefficients. The above is only an exemplary illustration and does not constitute a specific limitation of the present disclosure.

[0118] Based on the above configurations, each feature selection method can correspondingly select a set of first image features from the significant features. For example, if there are n feature selection methods, then n sets of first image features are generated. In the case of obtaining the first image features, at least one classification model can be further used to evaluate each set of first image features. Specifically, embodiments of the present disclosure can use the classification model to obtain the importance of each third image feature, and obtain the second image features according to the importance ranking. Among them, using the classification model to obtain the classification scores of each first image feature includes: separately inputting each set of first image features into the classification model, using the classification model to perform ten-fold cross-validation, and obtaining the metrics of the classification model. The metrics include at least two of AUC (area under the ROC curve), precision, accuracy, Recall, and F1, and using the average value of each metric as the classification score of this set of first image features. In the case of including multiple classification models, the classification scores corresponding to each classification model can be averaged to obtain the final classification score. In the case of obtaining the classification scores of each first image feature, each first image feature can be ranked from high to low. Among them, the feature selection method corresponding to the first image feature with the highest score can be determined as the preset feature selection method. In the case of determining the preset feature selection method with the highest score, the identifier of this method can be stored for subsequent feature processing.

[0119] In addition, for feature dimensionality reduction methods, embodiments of the present disclosure can include various feature dimensionality reduction methods, including linear dimensionality reduction methods and non-linear dimensionality reduction methods. Among them, the linear dimensionality reduction methods include principal component analysis algorithm (PCA), independent component analysis (ICA); the non-linear dimensionality reduction methods include t-distributed neighborhood embedding algorithm (T-SNE), isometric feature mapping (ISOMAP), and uniform manifold approximation and projection (UMAP). Using the above dimensionality reduction methods to perform dimensionality reduction processing on the obtained significant features respectively, the corresponding first reduced-dimensional features are obtained. In embodiments of the present disclosure, the number of features after dimensionality reduction is set to 10, and in other embodiments, it can also be set to other values, and the present disclosure does not make specific limitations on this.

[0120] In the case of obtaining the first selected features and the first reduced-dimensional features, multiple feature combination strategies can be used to perform feature combination to obtain the corresponding combined features. The specific methods can be at least one of the following methods:

[0121] B1) In some possible embodiments, the first selection feature may be combined with each of the first dimensionality reduction features respectively to obtain a plurality of combined features. For example, the first selection feature may be represented as F-select, and each of the first dimensionality reduction features may be represented as a set {F1, F2, …, Fk}, where k represents the number of feature dimensionality reduction algorithms. The generated combined features are concat(F-select, F1), concat(F-select, F2), …, concat(F-select, Fk), and concat represents combination.

[0122] B2) Perform feature selection on the first dimensionality reduction features using the preset selection method to obtain corresponding second dimensionality reduction features, and combine the first selection feature and the second dimensionality reduction features to obtain a plurality of combined features. Embodiments of the present disclosure may further perform feature selection processing on the first dimensionality reduction features obtained by each dimensionality reduction algorithm to obtain second dimensionality reduction features that match the state target in the dimensionality reduction features. Among them, the preset selection method may be used to perform this feature selection operation and obtain corresponding second dimensionality reduction features. Among them, the set of second dimensionality reduction features formed by each dimensionality reduction method may be represented as {F ’ 1, F ’ 2, …, F ’ k}, where k represents the number of feature dimensionality reduction algorithms. The generated combined features are concat(F-select, F ’ 1), concat(F-select, F ’ 2), …, concat(F-select, F ’ k), and concat represents combination. In this way, the feature information in the dimensionality reduction algorithm can be effectively extracted, the number of features can be reduced at the same time, and the operation accuracy and speed can be improved.

[0123] B3) Use the classification model to evaluate the classification scores of the first dimension-reduced features for the state target, and obtain multiple combined features based on the combination of the preset number of first dimension-reduced features with the highest scores and the first selected features. Similarly, in the embodiments of the present disclosure, the first dimension-reduced features obtained by evaluating each dimension reduction algorithm using the classification model can also be processed. In one example, when the state target is whether a person has suffered a stroke, the state target may include a first target indicating suffering from a stroke and a second target indicating not suffering from a stroke. The classification module is used to classify the state target based on the first dimension-reduced features, and the classification model is used to perform ten-fold cross-validation to obtain the metrics of the classification model. The metrics include at least two of AUC (area under the ROC curve), precision, accuracy, Recall, and F1, and the average value of each metric is used as the classification score of this group of first dimension-reduced features. In the case of including multiple classification models, the classification scores corresponding to each classification model can be averaged to obtain the final classification score. When the classification scores of each first dimension-reduced feature are obtained, the groups of first dimension-reduced features can be ranked from high to low. Among them, the preset number of first dimension-reduced features with the highest scores can be combined with the first selected features to obtain combined features. Wherein, the preset number can be 1 or a value greater than 1, and the present disclosure does not make specific limitations in this regard. Additionally, the state target can also be a first target indicating neurological impairment (NIHSS score greater than 0) and a second target indicating no neurological impairment (NIHSS score equal to 0), as well as a first target indicating good prognosis (90daymRS < 2) and a second target indicating poor prognosis (90daymRS > 2).

[0124] In the case of obtaining combined features corresponding to different feature combination strategies, a classification model can be used to evaluate each combined feature. The combined features in the embodiments of the present disclosure can be expressed as concat1, concat2, …, concatm. Where m is the number of combined features. Similarly, the classification model is used to perform ten-fold cross-validation on each combined feature to obtain the indicators of the classification model. The indicators include at least two of AUC (area under the ROC curve), precision, accuracy, Recall, and F1, and the average value of each indicator is used as the classification score of this group of first image features. In the case of including multiple classification models, the classification scores corresponding to each classification model can be averaged to obtain the final classification score. In the case of obtaining the classification scores of each combined feature, each group of combined features can be ranked from high to low. Among them, the combined feature with the highest classification score is determined as the optimal combined feature, and the corresponding feature combination strategy is the optimal feature combination strategy. For example, the optimal combination strategy obtained in the embodiments of the present disclosure can be Lasso+PCA_Lasso. It means that Lasso is used as the preset feature selection method, the dimensionality reduction algorithm is the PCA algorithm, the first selected features are obtained by using the preset feature selection method Lasso, the first dimensionality reduction features are obtained by the PCA algorithm, and the second dimensionality reduction features PCA_Lasso are further selected from the first dimensionality reduction features obtained by the PCA algorithm by using the preset feature selection method. Then, the first selected features and the second dimensionality reduction features are combined to obtain the optimal combined feature, and the corresponding Lasso+PCA_Lasso algorithm is expressed as the optimal combination strategy.

[0125] In some embodiments, the classification model in the embodiments of the present disclosure may include machine learning models based on different classification strategies. For example, it may include one or more of a support vector machine model (SVM) based on a non-linear relationship, a decision tree model, a random forest model, an Adaboost model, a neural network model, a k-nearest neighbor model (KNN), a logistic regression model (LR), a linear discriminant analysis model (DA), a gradient boosting classification model (GBDT), and a Gaussian naive Bayes model (NB).

[0126] Based on the above configuration, the embodiments of the present disclosure use a feature selection method and a dimensionality reduction method to process the dynamic features extracted from the cerebral perfusion images to obtain the first selected features and the first dimensionality reduction features, and use different combination strategies to perform feature combination on the first selected features and the first dimensionality reduction features obtained. Each combined feature is evaluated by a set classification model, and the combined feature that best matches the state target is selected, and then the optimal feature combination strategy corresponding to the state target is selected. The embodiments of the present disclosure can adaptively select the optimal combination strategy of dimensionality reduction features and selected features according to different state targets for the study of stroke diseases, improve the accuracy of analysis, and provide better support for clinical analysis.

[0127] In addition, an embodiment of the present disclosure further provides a state detection method for detecting the state corresponding to a cerebral perfusion image, such as whether there is a state of suffering from stroke, whether there is a state of neurological impairment, and a prognostic state.

[0128] Figure 4 The flowchart showing the state detection method according to an embodiment of the present disclosure, wherein the state detection method includes:

[0129] S100: Obtain a cerebral perfusion image;

[0130] S200: Extract the dynamic features of the cerebral perfusion image;

[0131] S300: Perform feature processing on the dynamic features according to a preset feature combination strategy to obtain combined features, and the feature combination strategy is determined according to the feature combination strategy selection method described in the above embodiment;

[0132] S300: Determine the state corresponding to the cerebral perfusion image by using the combined features.

[0133] In some possible implementation manners, the cerebral perfusion image may be at least one of magnetic resonance perfusion weighted imaging (PWI), computed tomography perfusion imaging (CTP), and arterial spin labeling perfusion imaging (ASL-MRI).

[0134] In addition, the manner of obtaining a cerebral perfusion image the same as that described in the above embodiment may include at least one of the following manners: A) directly collect a cerebral perfusion image by using a medical image acquisition device; in an embodiment of the present disclosure, the medical image acquisition device may be a nuclear magnetic resonance device, but it is not specifically limited in the present disclosure. B) Transmit and receive a cerebral perfusion image through an electronic device; in an embodiment of the present disclosure, a cerebral perfusion image transmitted by other electronic devices may be received through a communication manner, and the communication manner may include wired communication and / or wireless communication, which is not specifically limited in the present disclosure. C) Read the cerebral perfusion image stored in a database; in an embodiment of the present disclosure, a cerebral perfusion image stored locally or on a server may be read according to a received data reading instruction, and the present disclosure does not make a specific limitation thereto.

[0135] In addition, the process of extracting dynamic features in an embodiment of the present disclosure is the same as that described in the above embodiment, and will not be repeated here.

[0136] In the case of obtaining the optimal feature combination strategy using the above-described embodiments, the optimal feature combination strategy can be used to perform feature processing on the dynamic features. For example, the first selected features can be selected from the dynamic features using a preset feature selection method, and the feature dimensionality reduction algorithm in the optimal feature combination strategy can be used for feature dimensionality reduction, or feature extraction can be further performed on the dimensionality-reduced features (specifically determined according to the optimal feature combination strategy) to obtain the corresponding combined features.

[0137] And the combined features are used to perform the detection of the state target, such as the detection of whether a stroke is suffered, the detection of nerve function damage, and the detection of the prognosis state.

[0138] Based on the above embodiments, the detection of the state corresponding to the cerebral perfusion image can be realized, and the state detection accuracy can be improved by processing the features according to the optimal feature combination strategy.

[0139] It can be understood that the above-mentioned method embodiments mentioned in the present disclosure can be combined with each other to form combined embodiments without violating the principle logic. Due to space limitations, the present disclosure will not elaborate further.

[0140] In addition, the present disclosure also provides a feature combination strategy selection device, a state detection device, an electronic device, a computer-readable storage medium, and a program, all of which can be used to implement any one of the image processing methods provided by the present disclosure. The corresponding technical solutions and descriptions are referred to the corresponding records in the method part and will not be elaborated further.

[0141] Those skilled in the art can understand that in the above method of the specific embodiment, the writing order of each step does not mean a strict execution order and does not constitute any limitation on the implementation process. The specific execution order of each step should be determined according to its function and possible internal logic.

[0142] Figure 5 The block diagram of the feature combination strategy selection device according to an embodiment of the present disclosure is shown, such as Figure 5 shown, the feature combination strategy selection device includes:

[0143] A first acquisition module 10, which acquires the dynamic image features and state targets of the cerebral perfusion images in the cerebral perfusion image set;

[0144] A feature processing module 20, configured to perform feature processing on the dynamic image features based on a preset feature selection method and at least two feature dimensionality reduction methods to obtain first selected features and at least two groups of first dimensionality-reduced features;

[0145] A first combination module 30, configured to perform feature combination processing on the first selected features and the first dimensionality-reduced features based on multiple combination strategies to obtain multiple combined features;

[0146] The first determination module 40 is configured to evaluate the combined features by using a classification model and the state target, and determine the combined strategy corresponding to the combined feature with the highest classification score as the optimal feature combination strategy for evaluating the state target.

[0147] Figure 6 The block diagram of a state detection device according to an embodiment of the present disclosure is shown, where the state detection device may include:

[0148] The second acquisition module 100 is configured to acquire cerebral perfusion images;

[0149] The second feature processing module 200 extracts the dynamic features of the cerebral perfusion images;

[0150] The second combination module 300 is configured to perform feature processing on the dynamic features according to a preset feature combination strategy to obtain combined features, where the feature combination strategy is determined by the feature combination strategy selection method described in the above embodiments;

[0151] The second determination module 400 is configured to determine the state corresponding to the cerebral perfusion image by using the combined features. In some embodiments, the functions or modules included in the device provided by the embodiments of the present disclosure may be used to execute the methods described in the method embodiments above, and the specific implementation may refer to the description of the method embodiments above. For the sake of brevity, it will not be repeated here.

[0152] The embodiments of the present disclosure also propose a computer-readable storage medium, on which computer program instructions are stored, and when the computer program instructions are executed by a processor, the above methods are implemented. The computer-readable storage medium may be a non-volatile computer-readable storage medium.

[0153] The embodiments of the present disclosure also propose an electronic device, including: a processor; a memory for storing processor-executable instructions; wherein, the processor is configured to perform the above methods.

[0154] The electronic device may be provided as a terminal, a server or other forms of devices.

[0155] Figure 7 The block diagram of an electronic device 800 according to an embodiment of the present disclosure is shown. For example, the electronic device 800 may be a terminal such as a mobile phone, a computer, a digital broadcast terminal, a messaging device, a game console, a tablet device, a medical device, a fitness device, a personal digital assistant, etc.

[0156] Refer to Figure 7, the electronic device 800 may include one or more of the following components: a processing component 802, a memory 804, a power component 806, a multimedia component 808, an audio component 810, an input / output (I / O) interface 812, a sensor component 814, and a communication component 816.

[0157] The processing component 802 generally controls the overall operation of the electronic device 800, such as operations associated with display, telephone calls, data communications, camera operations, and recording operations. The processing component 802 may include one or more processors 820 to execute instructions to complete all or part of the steps of the above methods. In addition, the processing component 802 may include one or more modules to facilitate the interaction between the processing component 802 and other components. For example, the processing component 802 may include a multimedia module to facilitate the interaction between the multimedia component 808 and the processing component 802.

[0158] The memory 804 is configured to store various types of data to support the operation of the electronic device 800. Examples of these data include instructions for any application or method operating on the electronic device 800, contact data, phone book data, messages, pictures, videos, etc. The memory 804 may be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, a magnetic disk, or an optical disk.

[0159] The power component 806 provides power to various components of the electronic device 800. The power component 806 may include a power management system, one or more power sources, and other components associated with generating, managing, and distributing power for the electronic device 800.

[0160] The multimedia component 808 includes a screen that provides an output interface between the electronic device 800 and the user. In some embodiments, the screen may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen can be implemented as a touch screen to receive input signals from the user. The touch panel includes one or more touch sensors to sense touches, swipes, and gestures on the touch panel. The touch sensors can sense not only the boundaries of touch or swipe actions but also detect the duration and pressure associated with the touch or swipe operation. In some embodiments, the multimedia component 808 includes a front camera and / or a rear camera. When the electronic device 800 is in an operation mode, such as a shooting mode or a video mode, the front camera and / or the rear camera can receive external multimedia data. Each of the front camera and the rear camera can be a fixed optical lens system or have a focal length and optical zoom capabilities.

[0161] The audio component 810 is configured to output and / or input audio signals. For example, the audio component 810 includes a microphone (MIC) that is configured to receive external audio signals when the electronic device 800 is in an operation mode, such as a call mode, a recording mode, and a voice recognition mode. The received audio signals can be further stored in the memory 804 or transmitted via the communication component 816. In some embodiments, the audio component 810 further includes a speaker for outputting audio signals.

[0162] The I / O interface 812 provides an interface between the processing component 802 and a peripheral interface module, and the peripheral interface module can be a keyboard, a click wheel, buttons, etc. These buttons can include but are not limited to: a home button, a volume button, a power button, and a lock button.

[0163] The sensor component 814 includes one or more sensors for providing status assessments of various aspects of the electronic device 800. For example, the sensor component 814 can detect the on / off state of the electronic device 800, the relative positioning of components, such as the display and the keypad of the electronic device 800. The sensor component 814 can also detect a change in the position of the electronic device 800 or a component of the electronic device 800, the presence or absence of user contact with the electronic device 800, the orientation or acceleration / deceleration of the electronic device 800, and the temperature change of the electronic device 800. The sensor component 814 can include a proximity sensor configured to detect the presence of nearby objects without any physical contact. The sensor component 814 can also include a light sensor, such as a CMOS or a CCD image sensor, for use in imaging applications. In some embodiments, the sensor component 814 can further include an acceleration sensor, a gyroscope sensor, a magnetic sensor, a pressure sensor, or a temperature sensor.

[0164] The communication component 816 is configured to facilitate communication, either wired or wirelessly, between the electronic device 800 and other devices. The electronic device 800 can access a wireless network based on communication standards, such as WiFi, 2G, or 3G, or a combination thereof. In an exemplary embodiment, the communication component 816 receives a broadcast signal or broadcast-related information from an external broadcast management system via a broadcast channel. In an exemplary embodiment, the communication component 816 further includes a Near Field Communication (NFC) module to facilitate short-range communication. For example, the NFC module can be implemented based on Radio Frequency Identification (RFID) technology, Infrared Data Association (IrDA) technology, Ultra Wideband (UWB) technology, Bluetooth (BT) technology, and other technologies.

[0165] In an exemplary embodiment, the electronic device 800 can be implemented by one or more Application Specific Integrated Circuits (ASICs), Digital Signal Processors (DSPs), Digital Signal Processing Devices (DSPDs), Programmable Logic Devices (PLDs), Field Programmable Gate Arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components for performing the above-described method.

[0166] In an exemplary embodiment, a non-transitory computer-readable storage medium is also provided, such as a memory 804 including computer program instructions, which can be executed by a processor 820 of the electronic device 800 to complete the above-described method.

[0167] Figure 8 A block diagram of another electronic device 1900 according to an embodiment of the present disclosure is shown. For example, the electronic device 1900 can be provided as a server. Referring to Figure 8 , the electronic device 1900 includes a processing component 1922, which further includes one or more processors, and memory resources represented by a memory 1932 for storing instructions executable by the processing component 1922, such as application programs. The application programs stored in the memory 1932 can include one or more modules each corresponding to a set of instructions. In addition, the processing component 1922 is configured to execute instructions to perform the above-described method.

[0168] The electronic device 1900 may further include a power component 1926 configured to perform power management of the electronic device 1900, a wired or wireless network interface 1950 configured to connect the electronic device 1900 to a network, and an input / output (I / O) interface 1958. The electronic device 1900 can operate based on an operating system stored in the memory 1932, such as Windows ServerTM, Mac OS XTM, UnixTM, LinuxTM, FreeBSDTM, or the like.

[0169] In an exemplary embodiment, a non-volatile computer-readable storage medium is also provided, such as a memory 1932 including computer program instructions, and the computer program instructions can be executed by a processing component 1922 of the electronic device 1900 to complete the above method.

[0170] The present disclosure may be a system, a method, and / or a computer program product. The computer program product may include a computer-readable storage medium having thereon computer-readable program instructions for causing a processor to implement various aspects of the present disclosure.

[0171] A computer-readable storage medium may be a tangible device that can retain and store instructions for use by an instruction execution device. The computer-readable storage medium may be, for example, but not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing. More specific examples (a non-exhaustive list) of the computer-readable storage medium include: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a static random access memory (SRAM), a portable compact disc read-only memory (CD-ROM), a digital versatile disc (DVD), a memory stick, a floppy disk, a mechanically encoded device such as a punch card or raised structures in a groove having instructions stored thereon, and any suitable combination of the foregoing. The computer-readable storage medium used herein is not construed as being a transient signal per se, such as a radio wave or other freely propagating electromagnetic wave, an electromagnetic wave propagated through a waveguide or other transmission medium (e.g., an optical pulse through an optical fiber cable), or an electrical signal transmitted through a wire.

[0172] The computer-readable program instructions described herein can be downloaded from a computer-readable storage medium to respective computing / processing devices, or downloaded to an external computer or an external storage device through a network, such as the Internet, a local area network, a wide area network, and / or a wireless network. The network may include a copper transmission cable, an optical fiber transmission, a wireless transmission, a router, a firewall, a switch, a gateway computer, and / or an edge server. A network adapter or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards the computer-readable program instructions for storage in a computer-readable storage medium in each computing / processing device.

[0173] The computer program instructions for performing the operations of the present disclosure may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine - related instructions, microcode, firmware instructions, state - setting data, or source code or object code written in any combination of one or more programming languages, including object - oriented programming languages such as Smalltalk, C++, etc., and conventional procedural programming languages such as the "C" language or similar programming languages. The computer - readable program instructions may be executed entirely on the user's computer, partially on the user's computer, executed as a stand - alone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., through the Internet using an Internet service provider). In some embodiments, by using the state information of the computer - readable program instructions to customize an electronic circuit, such as a programmable logic circuit, a field - programmable gate array (FPGA), or a programmable logic array (PLA), the electronic circuit can execute the computer - readable program instructions to implement various aspects of the present disclosure.

[0174] Aspects of the present disclosure are described herein with reference to the flowcharts and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the present disclosure. It should be understood that each block of the flowcharts and / or block diagrams, and combinations of blocks in the flowcharts and / or block diagrams, can be implemented by computer - readable program instructions.

[0175] These computer - readable program instructions can be provided to a processor of a general - purpose computer, a special - purpose computer, or other programmable data - processing apparatus to produce a machine such that the instructions, when executed by the processor of the computer or other programmable data - processing apparatus, create a means for implementing the functions / acts specified in one or more blocks of the flowcharts and / or block diagrams. These computer - readable program instructions can also be stored in a computer - readable storage medium, which causes a computer, a programmable data - processing apparatus, and / or other devices to operate in a particular manner, so that the computer - readable medium storing the instructions includes a manufacture comprising instructions for implementing various aspects of the functions / acts specified in one or more blocks of the flowcharts and / or block diagrams.

[0176] Computer-readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device, causing a series of operational steps to be performed on the computer, other programmable data processing apparatus, or other device to produce a computer-implemented process such that the instructions executed on the computer, other programmable data processing apparatus, or other device implement the functions / acts specified in one or more boxes of the flowchart and / or block diagram.

[0177] The flowcharts and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present disclosure. In this regard, each block in the flowchart or block diagram may represent a module, a segment of a program, or a portion of an instruction, which contains one or more executable instructions for implementing the specified logical function. In some alternative implementations, the functions noted in the blocks may occur out of the order noted in the figures. For example, two consecutive blocks may in fact be executed substantially in parallel, or they may sometimes be executed in the reverse order, depending on the functions involved. It should also be noted that each block of the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented by a dedicated hardware-based system that performs the specified functions or acts, or by a combination of dedicated hardware and computer instructions.

[0178] The embodiments of the present disclosure have been described above. The above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and variations will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The choice of terms used herein is intended to best explain the principles of the embodiments, the practical application, or the technical improvement of the technology in the market, or to enable other ordinary skilled persons in the art to understand the embodiments disclosed herein.

Claims

1. A method for selecting a feature combination strategy, characterized in that Including: Obtaining the dynamic image features and state targets of the brain perfusion images in the brain perfusion image set; Performing feature processing on the dynamic image features based on a preset feature selection method and at least two feature dimensionality reduction methods to obtain first selected features and at least two groups of first dimensionality-reduced features; wherein, performing feature processing on the dynamic image features based on a preset feature selection method and at least two feature dimensionality reduction methods to obtain first selected features and at least two groups of first dimensionality-reduced features includes: obtaining the significant features in the dynamic image features; performing feature processing on the significant features based on a preset feature selection method and at least two feature dimensionality reduction methods to obtain first selected features and at least two groups of first dimensionality-reduced features; determining the preset feature selection method includes: based on at least two feature selection methods, screening out the first features that meet the selection conditions of the feature selection methods from the significant features of the dynamic image features; using a classification model to evaluate the first features obtained by each of the feature selection methods, and determining the feature selection method corresponding to the first feature with the highest classification score as the preset feature selection method; Performing feature combination processing on the first selected features and the first dimensionality-reduced features based on multiple combination strategies to obtain multiple combined features; wherein, performing feature combination processing on the first selected features and the first dimensionality-reduced features based on multiple combination strategies to obtain multiple combined features includes at least one of the following ways: combining the first selected features with each of the first dimensionality-reduced features respectively to obtain multiple combined features; performing feature selection on the first dimensionality-reduced features using a preset feature selection method to obtain corresponding second dimensionality-reduced features, and combining the first selected features and the second dimensionality-reduced features to obtain multiple combined features; using the classification model to evaluate the classification scores of the first dimensionality-reduced features for the state target, and obtaining multiple combined features based on the combination of the preset number of first dimensionality-reduced features with the highest scores and the first selected features; Evaluating the combined features using a classification model and the state target, and determining the combination strategy corresponding to the combined feature with the highest classification score as the optimal feature combination strategy for evaluating the state target; wherein, evaluating the combined features using a classification model and the state target, and determining the combination strategy corresponding to the combined feature with the highest classification score as the optimal feature combination strategy includes: obtaining the scores of multiple classification metrics of the combined features using the classification model; determining the classification score based on the scores of the multiple classification metrics.

2. The method for selecting a feature combination strategy according to claim 1, wherein In the case where there are multiple classification models, determining the final classification score based on the mean of the classification scores corresponding to each classification model.

3. The method for selecting a feature combination strategy according to any one of claims 1 or 2, characterized in that The brain perfusion images include brain images at multiple moments, and obtaining the dynamic image features of the brain perfusion images in the brain perfusion image set includes: Determining the brain tissue region of the brain perfusion images; Based on the multiple moments, respectively performing feature extraction processing on the brain tissue region at each moment to obtain the moment image features at each moment; For the cerebral perfusion image, the temporal image features of the brain tissue region at each moment are respectively combined to obtain the dynamic image features of the cerebral perfusion image.

4. The method for selecting a feature combination strategy according to claim 1, wherein The state target includes at least one of a first target indicating whether there is an ischemic stroke lesion, a second target indicating the degree of nerve function injury, and a third target indicating the prognosis state.

5. A feature combination strategy selection device, characterized in that It includes: A first acquisition module that acquires the dynamic image features and state targets of the cerebral perfusion images in the cerebral perfusion image set; A feature processing module for performing feature processing on the dynamic image features based on a preset feature selection method and at least two feature dimensionality reduction methods to obtain a first selected feature and at least two sets of first dimensionality-reduced features; wherein, performing feature processing on the dynamic image features based on a preset feature selection method and at least two feature dimensionality reduction methods to obtain a first selected feature and at least two sets of first dimensionality-reduced features includes: obtaining the significant features in the dynamic image features; performing feature processing on the significant features based on a preset feature selection method and at least two feature dimensionality reduction methods to obtain a first selected feature and at least two sets of first dimensionality-reduced features; determining the preset feature selection method includes: based on at least two feature selection methods, screening out the first features that meet the selection conditions of the feature selection method from the significant features of the dynamic image features; using a classification model to evaluate the first features obtained by each feature selection method, and determining the feature selection method corresponding to the first feature with the highest classification score as the preset feature selection method; A first combination module for performing feature combination processing on the first selected feature and the first dimensionality-reduced features based on multiple combination strategies to obtain multiple combined features; wherein, performing feature combination processing on the first selected feature and the first dimensionality-reduced features based on multiple combination strategies to obtain multiple combined features includes at least one of the following methods: combining the first selected feature with each of the first dimensionality-reduced features to obtain multiple combined features; performing feature selection on the first dimensionality-reduced features using a preset feature selection method to obtain corresponding second dimensionality-reduced features, and combining the first selected feature and the second dimensionality-reduced features to obtain multiple combined features; using the classification model to evaluate the classification scores of the first dimensionality-reduced features for the state target, and obtaining multiple combined features based on the combination of the preset number of first dimensionality-reduced features with the highest scores and the first selected feature; A first determination module for evaluating the combined features using a classification model and the state target, and determining the combination strategy corresponding to the combined feature with the highest classification score as the optimal feature combination strategy for evaluating the state target; wherein, evaluating the combined features using a classification model and the state target, and determining the combination strategy corresponding to the combined feature with the highest classification score as the optimal feature combination strategy includes: obtaining the scores of multiple classification metrics of the combined features using the classification model; determining the classification score based on the scores of the multiple classification metrics.

6. A state detection method, characterized in that, It includes: Obtain a cerebral perfusion image; Extract the dynamic features of the cerebral perfusion image; Performing feature processing on the dynamic features according to a preset feature combination strategy to obtain combined features, where the feature combination strategy is determined according to the feature combination strategy selection method described in any one of claims 1-4; Determining the state corresponding to the cerebral perfusion image by using the combined features.

7. A state detection device, characterized in that, Comprising: A second acquisition module, configured to acquire a cerebral perfusion image; A second extraction module, extracting the dynamic features of the cerebral perfusion image; A second combination module, configured to obtain the combined features of the cerebral perfusion image according to a preset feature combination strategy, where the feature combination strategy is determined according to the feature combination strategy selection method described in any one of claims 1-4; A second determination module, configured to determine the state corresponding to the cerebral perfusion image by using the combined features.

8. An electronic device, characterized in that, Comprising: A processor; A memory for storing instructions executable by the processor; Wherein, the processor is configured to call the instructions stored in the memory to execute the feature combination strategy selection method described in any one of claims 1-4.

9. An electronic device, characterized in that, Comprising: A processor; A memory for storing instructions executable by the processor; Wherein, the processor is configured to call the instructions stored in the memory to execute the state detection method described in claim 6.

Images