Prediction model construction method and device, prediction method and device, and electronic equipment

Abstract

The present disclosure relates to a prediction model construction method and apparatus, a prediction method and apparatus, and electronic equipment. The prediction model construction method comprises: obtaining a first image feature of a region of interest (ROI) of each cerebral perfusion image in a cerebral perfusion image set, wherein the ROI includes an abnormal region; selecting a second image feature that satisfies a preset condition from the first image feature; obtaining a survival feature based on the second image feature; and training a model based on the second image feature and the survival feature to obtain a prediction model for assessing prognosis. Embodiments of the present disclosure can improve the accuracy of prognosis prediction.

Description

Technical Field

[0001] The present disclosure relates to the technical field of medical image processing, and in particular to a prediction model construction method and device, a prediction method and device, and an electronic device. Background Art

[0002] Stroke has become the second leading cause of death worldwide, with extremely high mortality and disability rates. Even patients who survive often suffer from varying degrees of neurological impairment, resulting in a compromised quality of life, which places a heavy burden on patients, their families, and society.

[0003] Currently, the main treatments for ischemic stroke include intravenous or arterial TPA and interventional surgery. Appropriate treatment strategies can ensure optimal treatment and recovery for patients, but these inevitably involve varying degrees of risk. Therefore, individual risk and patient benefit are important considerations when choosing a treatment plan. Therefore, accurate prognostic prediction can help physicians tailor treatment plans for patients, assist in arranging rehabilitation training programs, reduce adverse outcomes, and alleviate the burden on families and society.

[0004] Currently, before thrombolytic therapy for ischemic stroke, treatment decisions are typically assessed based on the National Institutes of Health Stroke Scale (NIHSS) score, clinical symptoms, and auxiliary examination results. This assessment method is simple and easy to use; however, it relies heavily on the physician's experience, cannot quantify the prognosis of treatment, and its accuracy needs to be improved. Some researchers have also conducted research based solely on medical imaging, using traditional imaging features to predict risk using artificial intelligence methods. Artificial intelligence methods such as the Vgg-16 twin network are used to process images and predict 90-day prognosis, but the prediction accuracy is not ideal. Therefore, there is an urgent need to develop a solution that can accurately assess and predict the prognosis of stroke patients. Summary of the Invention

[0005] The present disclosure proposes a prediction model construction method and device, a prediction method and device, and an electronic device.

[0006] According to one aspect of the present disclosure, a method for constructing a prognostic status prediction model is provided, comprising:

[0007] Acquiring a first image feature of a region of interest of each cerebral perfusion image in the cerebral perfusion image set, wherein the region of interest includes an abnormal region;

[0008] Filtering out a second image feature that meets a preset condition from the first image feature;

[0009] obtaining a survival feature based on the second image feature;

[0010] Based on the second image feature and the survival feature, a model is trained to obtain a prediction model for evaluating the prognosis status.

[0011] In some possible implementations, screening out the second image feature that meets a preset condition from the first image feature includes:

[0012] Determining a feature item of the second image feature based on a multi-level feature selection strategy;

[0013] The feature item is filtered out from the first image feature to obtain the second image feature.

[0014] In some possible implementations, the region of interest also includes a normal region, and the brain perfusion images in the brain perfusion image set include brain images at multiple moments;

[0015] The determining of the feature item of the second image feature based on the multi-level feature selection strategy includes:

[0016] extracting first image features of the region of interest based on the multiple moments respectively;

[0017] Based on a multi-level feature selection strategy, a second image feature for distinguishing the normal tissue from the abnormal tissue is screened out from the first image feature.

[0018] In some possible implementations, obtaining a survival feature based on the second image feature includes:

[0019] Predicting the prognostic risk of the second imaging feature using a survival model;

[0020] determining the survival signature based on the prognostic risk;

[0021] and / or

[0022] The obtaining of the survival feature based on the second image feature includes:

[0023] Acquiring clinical text information corresponding to the cerebral perfusion image set, wherein the clinical text information includes age;

[0024] using a survival model to predict the prognostic risk of the second imaging feature under the age condition;

[0025] The survival signature is determined based on the prognostic risk.

[0026] In some possible implementations, obtaining a first image feature of a region of interest of each cerebral perfusion image in the cerebral perfusion image set includes:

[0027] determining a region of interest of the cerebral perfusion image, wherein the region of interest includes abnormal tissue;

[0028] Based on brain images at multiple moments in the brain perfusion image, a first image feature of the region of interest is extracted.

[0029] In some possible implementations, the method further includes:

[0030] determining an optimal feature selection method when a second image feature that meets a preset condition is selected from the first image feature;

[0031] And the training of a model based on the second image feature and the survival feature to obtain a prediction model for evaluating the prognosis status includes:

[0032] selecting a prognostic feature related to the prognostic status from the second image features based on the optimal feature selection method;

[0033] Based on the combined features of the prognostic features and the survival features, a model is trained to obtain a prediction model for evaluating the prognostic status; or

[0034] The step of training a model based on the second image feature and the survival feature to obtain a prediction model for evaluating the prognosis status includes:

[0035] selecting a prognostic feature related to the prognostic status from the second image feature and the clinical text information based on the optimal feature selection method;

[0036] Based on the combined features of the prognostic features and the survival features, a model is trained to obtain a prediction model for evaluating the prognostic status; or

[0037] The step of training a model based on the second image feature and the survival feature to obtain a prediction model for evaluating the prognosis status includes:

[0038] selecting a prognostic feature related to the prognostic status from the second image features based on the optimal feature selection method;

[0039] Based on the combined features of the prognostic features, clinical text features and the survival features, a model is trained to obtain a prediction model for evaluating the prognostic status.

[0040] According to a second aspect of the present disclosure, a method for predicting risk of a prognostic state is provided, comprising:

[0041] obtaining a first image feature of a region of interest in a cerebral perfusion image;

[0042] Filtering out a second image feature that meets a preset condition from the first image feature;

[0043] obtaining a survival feature based on the second image feature;

[0044] A prediction model is used to predict the prognosis status within a preset time based on the second imaging feature and the survival feature. The prediction model is obtained by the method for constructing a prognosis status prediction model described in any one of the first aspects.

[0045] According to a third aspect of the present disclosure, there is provided a device for constructing a prognostic status prediction model, comprising:

[0046] A first acquisition module is used to acquire a first image feature of a region of interest of each cerebral perfusion image in the cerebral perfusion image set;

[0047] a first screening module, configured to screen out second image features that meet preset conditions from the first image features;

[0048] a first survival module, configured to obtain a survival feature based on the second image feature;

[0049] A training module is used to train a model based on the second image feature and the survival feature, and obtain a prediction model for evaluating the prognosis status.

[0050] According to a fourth aspect of the present disclosure, there is provided a prognostic status prediction device, comprising:

[0051] A second acquisition module is used to acquire a first image feature of a region of interest in a cerebral perfusion image;

[0052] a second screening module, configured to screen out second image features that meet preset conditions from the first image features;

[0053] a second survival module, configured to obtain a survival feature based on the second image feature;

[0054] A prediction module is used to predict the prognosis status within a preset time based on the second image feature and the survival feature using a prediction model.

[0055] According to a fifth aspect of the present disclosure, there is provided an electronic device, comprising:

[0056] processor;

[0057] a memory for storing processor-executable instructions;

[0058] The processor is configured to call the instructions stored in the memory to execute the method described in any one of the first aspects, or to execute the method described in the second aspect.

[0059] According to a sixth aspect of the present disclosure, a computer-readable storage medium is provided, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the method of the first aspect or the second aspect is implemented.

[0060] In the disclosed embodiments, a second imaging feature that satisfies a preset condition is selected from a set of cerebral perfusion image features, and a survival feature that reflects survival information is further obtained. The survival feature and the second imaging feature are then used to train a prediction model, resulting in a prediction model that accurately assesses prognosis. The introduction of the survival feature in the disclosed embodiments effectively improves the prediction accuracy of the model.

[0061] It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the disclosure.

[0062] Further features and aspects of the present disclosure will become apparent from the following detailed description of exemplary embodiments with reference to the attached drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] The accompanying drawings herein are incorporated into 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.

[0064] Figure 1 A flowchart illustrating a method for constructing a prognostic status prediction model according to an embodiment of the present disclosure;

[0065] Figure 2 A flowchart of obtaining a first image feature according to an embodiment of the present disclosure is shown;

[0066] Figure 3 A flow chart showing a method for screening a second image feature according to an embodiment of the present disclosure is shown;

[0067] Figure 4 A flowchart showing a method for predicting risk of a prognostic state according to an embodiment of the present disclosure is shown;

[0068] Figure 5 A block diagram illustrating a device for constructing a prognostic status prediction model according to an embodiment of the present disclosure;

[0069] Figure 6 A block diagram showing a prognostic state prediction device according to an embodiment of the present disclosure is shown;

[0070] Figure 7 A block diagram of an electronic device 800 according to an embodiment of the present disclosure is shown;

[0071] Figure 8 A block diagram of an electronic device 1900 according to an embodiment of the present disclosure is shown. DETAILED DESCRIPTION

[0072] 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 accompanying drawings represent elements with the same or similar functions. Although various aspects of the embodiments are shown in the accompanying drawings, the drawings are not necessarily drawn to scale unless otherwise indicated.

[0073] The word “exemplary” is used exclusively herein to mean “serving as an example, example, or illustration.” Any embodiment described herein as “exemplary” is not necessarily to be construed as preferred or advantageous over other embodiments.

[0074] The term "and / or" herein simply describes an association relationship between associated objects, indicating that three relationships can exist. For example, "A and / or B" can represent the existence of three situations: A alone, A and B simultaneously, and B alone. Furthermore, the term "at least one" herein refers to any combination of at least two of any one or more of a plurality of items. For example, "at least one of A, B, and C" can represent any one or more elements selected from the set consisting of A, B, and C.

[0075] In addition, numerous specific details are provided in the following detailed description to better illustrate the present disclosure. Those skilled in the art will appreciate that the present disclosure can be practiced without certain specific details. In some instances, methods, means, components, and circuits well known to those skilled in the art are not described in detail in order to highlight the main points of the present disclosure.

[0076] The method for constructing a prognostic state prediction model or the execution subject of the prognostic state prediction method provided in the embodiments of the present disclosure may be an image processing device. For example, the above method may be executed by a terminal device, a server, or other processing device, wherein 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, an in-vehicle device, a wearable device, etc. In some possible implementations, the method may be implemented by a processor calling computer-readable instructions stored in a memory.

[0077] It can be understood that the above-mentioned various 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 on them.

[0078] Figure 1 A flowchart showing a method for constructing a prognostic status prediction model according to an embodiment of the present disclosure is shown in FIG. Figure 1As shown, the method for constructing the prognostic status prediction model includes:

[0079] S10: obtaining a first image feature of a region of interest of each cerebral perfusion image in the cerebral perfusion image set;

[0080] In some possible embodiments, the brain perfusion image set may include multiple brain perfusion images, wherein 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). In addition, the brain perfusion image includes multiple groups of brain images, each group of brain images may be a brain image scanned over a time range, and the multiple groups of brain images in the brain perfusion image may be brain images scanned over a continuous time (as multiple moments). In addition, the brain perfusion images in the embodiments of the present disclosure may be images of patients with brain diseases, such as perfusion images of patients with ischemic stroke, or images of patients with glioma. The disclosure does not specifically limit the type, but the disease types of the brain perfusion images in the perfusion image set are the same, such as the perfusion image set of patients with cerebral ischemia.

[0081] In some possible implementations, the region of interest of each brain image in the brain perfusion image can be obtained, and the region of interest may include an abnormal region; the abnormal region is a brain tissue lesion region, such as a region corresponding to ischemic tissue, and in other embodiments may also be a tumor region or a hemorrhage region, which is not specifically limited in the present disclosure.

[0082] In some possible implementations, image features of the region of interest may be extracted for each brain image at each moment, and the first image features of the abnormal region may be formed.

[0083] S20: Filtering out second image features that meet preset conditions from the first image features;

[0084] In some possible implementations, multi-level feature selection may be performed using different feature selection methods to screen out second image features that can accurately distinguish normal areas from abnormal areas from the first image features.

[0085] S30: Obtaining a survival feature based on the second image feature;

[0086] In some possible implementations, a survival model may be used to obtain a survival feature corresponding to the second image feature, and the survival feature may represent the survival probability of the patient.

[0087] S40: Based on the second image feature and the survival feature, a model is trained to obtain a prediction model for evaluating the prognosis status.

[0088] In some possible implementations, the second image features of each cerebral perfusion image and the obtained survival features can be used to train a model to obtain a prediction model that can perform prognostic status assessment. In the disclosed embodiment, each cerebral perfusion image in the cerebral perfusion image set corresponds to a functional recovery score (prognostic score) of the patient within 90 days after treatment. The score is obtained according to the modified Rankin Stroke Scale, i.e., the mRS scoring standard. The mRS score includes seven levels from 0 to 6, and the higher the score, the worse the functional recovery. Among them, 0 indicates complete recovery, 6 indicates death, when greater than 2, it indicates the presence of varying degrees of disability, and 1 and 2 indicate good recovery. In some possible implementations, the prognostic score can be used as a label, and the second image feature and the survival feature can be input into the model for training to obtain a prediction model for assessing the prognostic status.

[0089] Based on the above configuration, the present disclosure can further obtain survival features that can reflect survival information by screening out second image features that meet preset conditions, and use the survival features and second image features to train a prediction model, thereby obtaining a prediction model capable of assessing prognosis. The introduction of the survival features can effectively improve the prediction accuracy of the model.

[0090] The following is a detailed description of the embodiments of the present disclosure in conjunction with the accompanying drawings. First, the embodiments of the present disclosure can obtain a brain perfusion image set, wherein the method of obtaining the brain perfusion image set can include at least one of the following methods:

[0091] A) Directly using a medical image acquisition device to acquire brain perfusion images; in the embodiment of the present disclosure, the medical image acquisition device may be a nuclear magnetic resonance device, but this is not a specific limitation of the present disclosure.

[0092] B) transmitting and receiving a brain perfusion image set including a plurality of brain perfusion images through an electronic device; the embodiment of the present disclosure can receive brain perfusion images transmitted by other electronic devices through communication, and the communication method can include wired communication and / or wireless communication, which is not specifically limited in the present disclosure.

[0093] C) Reading the brain perfusion images stored in the database; the embodiment of the present disclosure can read the brain perfusion images stored locally or in the server according to the received data reading instruction, so as to obtain a brain perfusion image set, which is not specifically limited in the present disclosure.

[0094] It should be noted that the brain perfusion images in the embodiments of the present disclosure can be perfusion images acquired by the same device or different devices. Personnel in the relevant technical field can select the corresponding device according to their needs, and no specific limitation is made here.

[0095] In the embodiment of the present disclosure, on the premise of obtaining a set of brain perfusion images, the brain perfusion images can be pre-processed, or a region of interest in the brain perfusion images can be directly determined for subsequent processing.

[0096] The preprocessing of the cerebral perfusion image includes at least one of the following methods: performing bone removal processing on the cerebral perfusion image; performing time series registration processing on the cerebral perfusion image; and performing time series smoothing processing on the cerebral perfusion image.

[0097] FSL software can be used to perform bone removal on brain perfusion images, thereby preventing skull pixels from interfering with subsequent image processing and improving the accuracy of region of interest and feature extraction. Furthermore, time series registration of brain perfusion images involves rigid registration of multiple brain images within the perfusion images to eliminate motion deviations caused by patient movement during image acquisition. The registration algorithm is implemented using ITK Elastix and is not specifically limited in this disclosure.

[0098] Smoothing the temporal sequence of cerebral perfusion images includes: obtaining the grayscale values of the same pixel in brain tissue at multiple moments to form a temporal grayscale sequence; and performing smoothing on the temporal grayscale sequence. The smoothing process may include three moving averages. In the disclosed embodiment, a 1x3 moving window may be used to smooth the temporal grayscale sequence, but this disclosure does not specifically limit this. Smoothing can reduce noise in cerebral perfusion images and improve image quality.

[0099] After preprocessing or obtaining a set of cerebral perfusion images, a first image feature of a region of interest in the cerebral perfusion images may be further obtained, wherein the region of interest may include an abnormal region. The abnormal region may include a cerebral ischemic region, or in other embodiments, may also be at least one of a cerebral hemorrhage region, a brain tumor region, a cerebral infarction region, and the like.

[0100] Figure 2 A flow chart of obtaining a first image feature according to an embodiment of the present disclosure is shown. The process of obtaining a first image feature of a region of interest of each cerebral perfusion image in a cerebral perfusion image set includes:

[0101] S101: determining a region of interest of the brain perfusion image, where the region of interest includes abnormal tissue;

[0102] S102: Extracting a first image feature of the region of interest based on brain images at multiple moments in the brain perfusion image.

[0103] In one example, the embodiment of the present disclosure can directly segment abnormal areas, such as ischemic areas, from cerebral perfusion images through a deep learning model, or the embodiment of the present disclosure can further obtain corresponding kinetic parameters based on the cerebral perfusion images, and determine abnormal areas based on the kinetic parameters. The kinetic parameters may include cerebral blood flow (CBF), cerebral blood volume (CBV), mean transit time (MTT), time to peak (TTP), maximum residual function time (TMAX), etc. For example, the cerebral perfusion image can be post-processed by perfusion post-processing software Rapid, etc. to obtain the Tmax parameter, and the abnormal area can be determined using pixels whose Tmax parameter is greater than a threshold. The threshold value can be 6S, or other values, such as 4S or 10S, and the present disclosure does not make specific limitations on this.

[0104] When a brain perfusion image is obtained, the first image feature of the region of interest can be extracted. For each brain perfusion image, feature extraction processing can be performed on the region of interest in the brain image at different times, and then the features at each time are fused to obtain the first image feature.

[0105] Among them, the extracting of the first image feature of the region of interest based on the brain images at multiple moments in the brain perfusion image includes: based on the multiple moments, performing feature extraction processing on the region of interest at each moment to obtain the moment image feature at each moment; for the brain perfusion image, combining the moment image features of the region of interest at each moment to obtain the first image feature of the region of interest.

[0106] In the embodiment of the present disclosure, the brain perfusion image may include t groups of brain images, each of which corresponds to t moments. For example, in the embodiment of the present disclosure, t may be an integer greater than 1 and less than or equal to 50, but this is not a specific limitation of the present disclosure. In the embodiment of the present disclosure, feature extraction processing may be performed on each region of interest in the brain image at moment t.

[0107] In one example, feature extraction processing may include: performing at least one image transformation on the region of interest, obtaining an augmented set of the region of interest based on the region of interest and its image transformation results; and extracting at least one of first-order gradient features, shape features, and texture features from any image in the augmented set. The image transformation includes at least one of Fourier transform, Gabor transform, Laplace transform, wavelet transform, square root filtering, and exponential function filtering. The disclosed embodiments may utilize the original region of interest and the image transformation results to form an augmented set, and perform feature extraction on each region of interest in the augmented set to obtain richer image features. The extracted first-order gradient features may include features describing a single pixel or single voxel, such as the grayscale mean, maximum grayscale value, minimum grayscale value, variance, and percentiles (14 and 15) of the region of interest; 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 skewness, below the mean) or to the right (positive skewness, above the mean); while kurtosis reflects the tailing of the data distribution due to outliers relative to the Gaussian distribution. Shape features can include surface and volume-based features, such as compactness and sphericity features. Texture features can include absolute gradient (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).

[0108] In some possible implementations, the feature extraction process described above can be performed using extractive imaging omics to obtain the moment-by-moment image features corresponding to the region of interest at each moment. By combining the moment-by-moment image features at each moment, the first image features of the region of interest are obtained. The disclosed embodiment can calculate 65,800 first image features (3D brain images at 50 moments × 1,316 moment-by-moment image features). These first 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 = 1200), (4) gray-level run length matrix GLRLM (16 × 50 = 800), (5) gray-level size region matrix GLSZM (16 × 50 = 800), (6) neighboring gray tone difference matrix NGTDM (5 × 50 = 250), (7) gray-level correlation matrix GLDM (14 × 50 = 700), (8) Laplace transform (465 × 50 = 23250), (9) wavelet transform (744 × 50 = 37200). In the embodiment of the present disclosure, each moment image feature can be defined as a 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 time image feature “Log-sigma-1-0-mm-3d_firstorder_skewness” at the 17th time in the DSC-PWI cerebral perfusion image.

[0109] In addition, when performing feature extraction processing on a region of interest, the disclosed embodiment can also optimize the time sequence, reduce the time values, and improve computational efficiency. Specifically, the disclosed embodiment can divide the time t of the brain perfusion image into three groups, such as the first group being the preparation stage, the second group being the reaction stage, and the third group being the recovery stage. The preparation stage is the stage during the perfusion imaging process when the brain image is not affected by the contrast agent, the reaction stage is the stage when the contrast agent flows through the blood vessels, causing the grayscale value of the pixel to change, and the recovery stage is the process in which the grayscale value of the pixel returns to its initial state after the contrast agent leaves the blood vessels. In the disclosed embodiment, t is 50 moments, of which the first group is moments 1-10, the second group is moments 11-30, and the third group is moments 31-50. The above is merely an example of the disclosed embodiment and is not intended to be a specific limitation. When three groups of moments are obtained, mean processing can be performed on the brain images corresponding to the first group of moments, and mean processing can be performed on the brain images corresponding to the third group of moments. The brain images obtained by the mean processing of the first group, the brain images corresponding to the second group of moments, and the brain images obtained by the mean processing of the third group are used as new brain perfusion images for feature extraction processing. This reduces the amount of computation and improves feature extraction efficiency while ensuring comprehensive information.

[0110] The above-mentioned embodiment can extract image features from the three-dimensional brain image at each moment, so as to obtain dynamic image features at different moments from the perspective of three-dimensional images at multiple moments.

[0111] In addition, in some other embodiments of the present disclosure, the features of each layer of brain image in the cerebral perfusion image at time t can also be comprehensively analyzed to obtain a first image feature. The extracting the first image feature of the region of interest based on brain images at multiple moments in the cerebral perfusion image can also include: generating first brain images based on brain images of the same layer at different moments, the number of the first brain images being the same as the number of layers of the brain image, and the number of layers of the first brain image being the same as the number of moments; performing feature extraction processing on the region of interest in the first brain image to obtain layer image features; and obtaining the first image feature of the region of interest based on a combination of the layer image features of the region of interest in the cerebral perfusion image.

[0112] The brain images in the brain perfusion images of the present disclosure are 3D images, each of which has the same dimensions. These images may include multiple layers of brain images. The 3D images may include brain images in three directions, such as coronal, sagittal, and transverse planes. Each layer of image in each direction may be used as a feature extraction target in the present disclosure.

[0113] In one example, the dimension of a brain perfusion image can be expressed as t*C*W*H, where t represents the number of moments, C represents the number of layers of the brain image, and W and H represent the width and height of the brain image, respectively. In the embodiment of the present disclosure, brain images at t moments are extracted in the order of the first layer to the Cth layer, and a first brain image is formed. Each first brain image corresponds to the number of layers of the brain image, the number of 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 moments. Therefore, the dimension of the first brain image obtained for each brain perfusion image is t*W*H, and the number is C.

[0114] When a first brain image is obtained, feature extraction processing can be performed on the region of interest in the first brain image to obtain layer image features. The feature extraction processing is the same as the configuration of the above embodiment, including: performing at least one image transformation on the region of interest, obtaining an augmented set of the region of interest based on the region of interest and its image transformation results; and extracting at least one of the first-order gradient features, shape features, and texture features of any image in the augmented set. After obtaining the first brain image for each layer and performing feature extraction processing on the region of interest, the corresponding layer image features can be obtained, and the layer image features of each layer can be combined to obtain the first image features. When the number of layers of brain images is 20, the number of first image features can be 26320 (1316*20), but this is not a specific limitation of the present disclosure.

[0115] Similarly, the embodiment of the present disclosure may optimize the time value before extracting the layer image features to reduce the time value. The specific method is described in the above embodiment.

[0116] Based on the above registration, embodiments of the present disclosure can construct a three-dimensional image from a moment-by-moment perspective, extract layer image features, and further enrich the extracted dynamic image features. In other embodiments, the first image features obtained by embodiments of the present disclosure can be a combination of the two aforementioned types of first image features, further enriching the image features, but this is not a specific limitation of the present disclosure.

[0117] When a rich set of first image features is obtained, feature screening can be further performed to select second image features that can accurately distinguish normal areas from abnormal areas. The disclosed embodiment adopts a multi-level selection strategy to optimize selection accuracy.

[0118] Figure 3 A flow chart of a method for screening a second image feature according to an embodiment of the present disclosure is shown. Screening a second image feature that meets a preset condition from the first image feature includes:

[0119] S301: Determine feature items of the second image feature based on a multi-level feature selection strategy;

[0120] S302: Filter the feature items from the first image features to obtain the second image features.

[0121] In some possible embodiments, determining feature items of the second image features based on a multi-level feature selection strategy includes: extracting first image features of the region of interest based on the multiple time instants; and selecting, based on the multi-level feature selection strategy, second image features for distinguishing normal tissue from abnormal tissue from the first image features. Specifically, feature items capable of accurately distinguishing normal tissue from abnormal tissue can be determined in advance using the multi-level feature selection strategy, and these feature items can be extracted from the first image features for subsequent prediction of prognosis.

[0122] Specifically, the region of interest in the embodiment of the present disclosure may also include a normal area. In the process of determining the feature item of the second image feature, the first image features of the normal area and the abnormal area may be extracted respectively first. The process of extracting the first image feature of the normal area is the same as the process of extracting the first image feature of the abnormal area, and will not be repeated here. For details, please refer to the description of the above embodiment. In the case of determining the abnormal area, the brain tissue area outside the abnormal area may be determined as the normal area, or the non-lesion area in the symmetrical area of the abnormal area may be determined as the normal area. In the embodiment of the present disclosure, by determining the normal tissue on the symmetrical side of the abnormal area as the normal area, on the one hand, the parameter amount of the normal area can be reduced and the computational efficiency can be improved; on the other hand, since the brain has the characteristic of symmetry, determining the normal area in a symmetrical manner can further highlight the difference between the normal area and the abnormal area, which is conducive to the subsequent extraction of the distinguishing features between the two.

[0123] After determining a normal region and extracting a first image feature of the normal region, a feature item of a second image feature for distinguishing the normal region from the abnormal region can be screened from the first image feature based on a multi-level feature selection strategy. This process may include: selecting a salient feature that satisfies a saliency requirement from the first image feature; screening a third image feature that satisfies a selection condition of the feature selection method from the salient feature based on at least two feature selection methods; and selecting a second image feature that satisfies a classification condition from the third image feature using at least one classification model.

[0124] In some possible implementations, the obtained first image feature can be first standardized to reduce the influence of the numerical span of the feature itself. Each row of features of the first image feature obtained in the embodiment 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 is performed on each column of features of the first image feature respectively. For example, the standardization in the embodiment of the present disclosure can be mean-variance standardization, so that the feature after standardization has a mean of 0 and a variance of 1. In other implementations, the ratio of each column of features to the maximum value of the column of features can also be used as the standardized feature value. Feature screening can then be performed using the first image feature after standardization.

[0125] In some possible implementations, the embodiments of the present disclosure perform feature screening from multiple angles, on the one hand to achieve dimensionality reduction processing of high-dimensional features, and on the other hand to improve the accuracy of selected features. First, the embodiments of the present disclosure can select significant features from the first image features, perform significance analysis on each first image feature of the normal area and the abnormal area, calculate the p-value (assumed value) between the two groups of features, and determine that the feature is a significant feature when the p-value is less than the significance threshold. The significance threshold is 0.05, and the p-value calculation method includes the T test. The above is only an exemplary explanation and is not a specific limitation of the present disclosure. In addition, the embodiments of the present disclosure can also calculate the correlation coefficient between the first image features of the normal area and the abnormal area. When the correlation coefficient of the feature is higher than the coefficient threshold and the p-value is less than the significance threshold, the feature is determined to be a significant feature. The coefficient threshold can be a value greater than 0.6, such as 0.9.

[0126] Secondly, the embodiment of the present disclosure adopts a variety of feature selection methods to perform feature selection, and the selection principles of the various selection methods are different. In one example, the feature selection method may include at least two of a method based on information theory, a method based on similar features, a method based on statistical features, and a method based on sparse features and flow features. The information theory-based method may 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), mutual information selection (MIFS), the joint mutual information (JMI), etc., and the similarity feature-based method may include a distance separability measure (Fisher score algorithm), a Laplacian score (Lap score algorithm), a feature weight algorithm (ReliefF), a statistical feature-based method may include a T score algorithm and an F score algorithm, and a sparse feature and flow feature-based method may include a multi-cluster feature selection algorithm (MCFS), a minimum absolute shrinkage selection operator (Lasso), and an Alpha algorithm.

[0127] In the embodiments of the present disclosure, at least two of the above-mentioned feature selection methods can be used to select significant features of normal and abnormal areas, wherein the selection conditions of feature selection methods other than the Lass algorithm may include: the maximum number of features is less than the feature number threshold, and the feature score is greater than the score threshold, wherein the feature number threshold is greater than 10, such as 20 in the present disclosure, and the score threshold can be greater than 0.6, such as 0.8 in the present disclosure. The selection condition of the Lasso algorithm is to select feature items with non-zero feature coefficients. The above is only an exemplary description and does not constitute a specific limitation of the present disclosure.

[0128] Based on the above configuration, each feature selection method can correspondingly select a set of third image features from the salient features. For example, if n feature selection methods are used, n sets of third image features can be generated. Once the third image features are obtained, at least one classification model can be further used to select second image features from the third image features that meet the classification criteria. The disclosed embodiments can implement the above process in two ways.

[0129] In some possible implementations, the embodiments of the present disclosure can combine the third image features obtained by each feature selection method to obtain all third image features, and use at least one classification model to perform classification of normal areas and abnormal areas based on all third image features, and determine the third image features that meet the classification conditions as second image features.

[0130] Specifically, embodiments of the present disclosure can utilize a classification model to obtain the importance of each third image feature, and rank them according to the importance to obtain the second image feature. The process of utilizing the classification model to obtain the importance of each third image feature can include: individually inputting each third image feature into the classification model, performing ten-fold cross-validation using the classification model, obtaining classification model indicators, the indicators including at least two of AUC (area under the ROC curve), precision, accuracy, Reall, and F1, and using the average of each indicator as the importance of the feature. In the case of including multiple classification models, the importance corresponding to each classification model can be averaged to obtain the final importance. Once the importance of each third image feature is obtained, the third image features can be ranked from high to low according to their importance, wherein a preset number of third image features with the highest importance can be used as the second image feature, or third image features with an importance above an importance threshold can be used as the second image feature. The preset number can be a value greater than 5, and the importance threshold can be a value greater than 0.6, but this is not a specific limitation of the present disclosure.

[0131] In other possible implementations, the disclosed embodiments may perform feature selection on each group of third image features obtained by each feature selection method, and select the best performing group of third image features as the second image features. Specifically, at least one classification model may be used to perform classification of normal areas and abnormal areas based on each group of third image features, and one or more groups of third image features that meet the classification criteria may be determined as the second image features. Based on the performance of each group of third image features on the classification model, a score for each group of third image features may be calculated; based on the score, the third image features that meet the classification criteria may be determined as the second image features. A group of third image features may be input into the classification model separately, and a ten-fold cross validation may be performed using the classification model to obtain an index of the classification model, the index including at least two of AUC (area under the ROC curve), precision, accuracy, Reall, and F1, and the average value of each index may be used as the score of the group of third image features. In the case of including multiple classification models, the scores corresponding to each classification model may be averaged to obtain the final score. After obtaining the scores for each group of third image features, the groups of third image features can be ranked from high to low according to the scores. The group of third image features with the highest scores can be used as the second image feature, or the third image features with scores above a scoring threshold can be used as the second image feature. The scoring threshold can be a value greater than 0.6, but is not a specific limitation of the present disclosure.

[0132] The classification model of the embodiment of the present disclosure may include machine learning models based on different classification strategies, such as a support vector machine model (SVM) based on nonlinear relationships, a decision tree model, a random forest model, an Adaboost model, a neural network model, a 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) One or more.

[0133] In addition, the disclosed embodiments can use the scores of the third image features obtained by each feature selection method on each classification model to obtain the score of the feature selection method, and select the feature selection method with the highest score as the optimal feature selection method. Specifically, the score of each group of third image features can be determined as the score of the feature selection method corresponding to the group of third image features, or the average importance of the third image features within each group can be used to determine the score of the feature selection method corresponding to the group of third image features.

[0134] Based on the above configuration, the embodiment of the present disclosure can utilize a multi-level feature selection strategy, integrate selection methods with different selection principles, and screen out second image features that can highly distinguish normal areas from abnormal areas, thereby improving the feature selection accuracy.

[0135] When the feature items of the second image feature are determined, these feature items can be selected from the first image features of the abnormal area, thereby obtaining the image feature that can accurately express the abnormal area, that is, the second image feature.

[0136] Furthermore, embodiments of the present disclosure can utilize the second imaging feature of the abnormal region to further derive a survival feature. Determining the survival feature based on the second imaging feature includes: predicting the prognostic risk of the second imaging feature using a survival model; and determining the survival feature based on the prognostic risk.

[0137] The disclosed embodiment can use the second image feature to generate the input feature to the survival model. The input feature can only include the second image feature, or in some examples, each patient object in the perfusion image set is also associated with corresponding clinical text information, that is, each brain perfusion image corresponds to matching clinical text information. In this case, the input feature can be the clinical text information and the second image feature.

[0138] Among them, the clinical text information includes at least one of the patient's age, history of hypertension, history of diabetes, history of atrial fibrillation, limb flexibility, whether coma, unclear speech, time of onset of symptoms, whether thrombolytic treatment has been performed and the method of thrombolysis, the volume of the abnormal area and the location of the abnormal area. In addition, the clinical text information can also include the patient's prognostic score (90day mRS) after 90 days. The embodiment of the present disclosure uses the prognostic score as the prediction target. In the process of obtaining survival characteristics, the prognostic score is used as the outcome state variable, the input feature is used as the feature variable, and the age is used as the time variable for survival prediction. The survival risk value output by the model is extracted and used as the survival feature. In other words, the embodiment of the present disclosure can obtain the clinical text information corresponding to the cerebral perfusion image set, and can use the survival model to predict the prognostic risk corresponding to the second image feature under the condition of age, and can determine the survival feature based on the prognostic risk.

[0139] The survival model of the embodiment of the present disclosure can be a Cox model, which uses the Cox model to establish a linear relationship between the input features and the prognostic score, and regresses to obtain the survival characteristics. Alternatively, in other real-time methods, the survival model can also be a DeepSurv model, which uses the DeepSurv model to obtain a nonlinear relationship between the input features and the prognostic score, and predicts the survival characteristics. In some embodiments of the present disclosure, the survival characteristics output by the Cox model and the DeepSurv model are also weighted and summed to obtain the final survival characteristics, where the weights are all 0.5, or can also be other values, which are not specifically limited in the present disclosure.

[0140] Furthermore, the embodiment of the present disclosure may also perform a standardization operation on the obtained survival features to further optimize the feature distribution.

[0141] In the disclosed embodiment, when a survival feature is obtained, a model can be trained using the second image feature and the survival feature to obtain a prediction model. In the disclosed embodiment, the second image feature can be combined with the obtained survival feature to train the prediction model, or clinical text information, the second image feature, and the survival feature can be combined to train the prediction model.

[0142] Alternatively, the training of a model based on the second image feature and the survival feature to obtain a prediction model for evaluating the prognostic status may also include: selecting a prognostic feature related to the prognostic status from the second image feature based on an optimal feature selection method; and training a model based on a combination of the prognostic feature and the survival feature to obtain a prediction model for evaluating the prognostic status.

[0143] As in the above embodiment, in the process of executing the feature selection method, the scores of each feature selection method can be obtained, and the feature selection method with the highest score can be determined as the optimal feature selection method. Before executing the training operation of the prediction model, the prognostic feature related to the prognostic state can be selected from the obtained second image features. The embodiment of the present disclosure can use the optimal feature selection method to perform the selection operation of the prognostic feature, thereby selecting the prognostic feature that is closely related to the prognostic state in the feature sequence of the abnormal expression area. When the prognostic feature is obtained, the prognostic feature and the survival feature can be used to train the prediction model.

[0144] In another embodiment, a prognostic feature can be obtained by combining the second image feature with the clinical text information. Alternatively, a first prognostic feature can be selected from the second image feature and a second prognostic feature can be selected from the clinical text information, and the first and second prognostic features can be combined to obtain a prognostic feature, thereby integrating text and image features to improve prediction accuracy. When a prognostic feature is obtained, the prognostic feature and the survival feature can be used to train a prediction model.

[0145] In another embodiment, the prognostic feature may be selected only from the second image feature, and the survival feature, the prognostic feature and the clinical text information may be combined to train the prediction model.

[0146] Among them, the prediction model of the embodiment of the present disclosure may also include a machine learning model based on different classification strategies, such as a support vector machine model (SVM) based on nonlinear relationships, a decision tree model, a random forest model, an Adaboost model, a neural network model, a 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). Among them, the 90-day mRS can be used as a learning target, and the corresponding prognostic features can be feature information in the training set. When the prediction accuracy of the prediction model is greater than the accuracy threshold, it can be determined that the training requirements are met, thereby obtaining a corresponding prediction model, wherein the accuracy threshold is a value greater than 0.6.

[0147] In other embodiments of the present disclosure, multiple prediction models can be used to obtain an integrated prediction model, which is obtained by integrating prediction models that meet training requirements. The integrated prediction model can perform voting processing through the prediction models that meet training requirements to obtain a predicted value of the final prognosis status.

[0148] Based on the above configuration, the disclosed embodiment selects a second image feature that meets a preset condition from a set of cerebral perfusion image features, further extracts a survival feature that reflects survival information, and uses the survival feature and the second image feature to train a prediction model, resulting in a prediction model that accurately assesses prognosis. The introduction of the survival feature in the disclosed embodiment effectively improves the prediction accuracy of the model.

[0149] In addition, the embodiments of the present disclosure also provide a risk prediction method for prognostic status to achieve accurate prediction of prognostic status.

[0150] Figure 4 A flowchart showing a method for predicting the risk of a prognostic state according to an embodiment of the present disclosure is shown, wherein the method for predicting the risk of a prognostic state includes:

[0151] S100: Acquire a first image feature of a region of interest in a cerebral perfusion image;

[0152] S200: Filtering out second image features that meet preset conditions from the first image features;

[0153] S300: Obtaining a survival feature based on the second image feature;

[0154] S400: Predicting the prognosis within a preset time period based on the second imaging feature and the survival feature using a prediction model, wherein the prediction model is obtained by the above-mentioned method for constructing a prognosis prediction model.

[0155] Similarly, the embodiments of the present disclosure can perform an assessment and prediction of the prognosis status of a patient object. First, a first image feature can be extracted from the region of interest (such as an abnormal area) of the cerebral perfusion image of the patient object. The extraction process of the first image feature can refer to the description in the method for constructing the prediction model, which will not be repeated here. Furthermore, corresponding features are selected from the first image features based on the feature items of the second image feature to accurately express the features of the abnormal area. Survival features can then be obtained based on the second image features, where the process of obtaining the second image features and the survival features can also refer to the above-mentioned embodiment, and this disclosure will not repeat the description.

[0156] When survival features are obtained, prognostic features can be selected from the second image features and / or clinical text information, and the prognostic model can be used to predict the prognostic features and survival features to obtain a prognostic status score. The input features of the prognostic model are determined by the type of input features during the training process, and this disclosure does not specifically limit this.

[0157] Those skilled in the art will understand that in the above-mentioned method of the specific implementation method, 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 by its function and possible internal logic.

[0158] In addition, the present disclosure also provides a construction device and prediction device for a risk prediction model of a prognostic state, an electronic device, a computer-readable storage medium, and a program. The above can all be used to implement the construction method and prediction method of any prognostic state prediction model provided by the present disclosure. The corresponding technical solutions and descriptions are referred to the corresponding records in the method section and will not be repeated here.

[0159] Figure 5 A block diagram of a device for constructing a prognostic state prediction model according to an embodiment of the present disclosure is shown. Figure 5 As shown, the construction device of the prognosis status prediction model includes:

[0160] A first acquisition module 10 is used to acquire a first image feature of a region of interest of each brain perfusion image in a brain perfusion image set;

[0161] A first screening module 20 is configured to screen out second image features that meet a preset condition from the first image features;

[0162] A first survival module 30, configured to obtain a survival feature based on the second image feature;

[0163] The training module 40 is used to train a model based on the second image feature and the survival feature, and obtain a prediction model for evaluating the prognosis status.

[0164] Figure 6A block diagram of a prognosis status prediction device according to an embodiment of the present disclosure is shown, wherein the prognosis status prediction device includes: a second acquisition module 100, used to obtain a first image feature of a region of interest in a cerebral perfusion image; a second screening module 200, used to screen out a second image feature that meets a preset condition from the first image feature; a second survival module 300, used to obtain a survival feature based on the second image feature; and a prediction module 400, used to use a prediction model to predict the prognosis status within a preset time based on the second image feature and the survival feature.

[0165] In some embodiments, the functions or modules included in the device provided by the embodiments of the present disclosure can be used to execute the method described in the above method embodiments. The specific implementation can refer to the description of the above method embodiments. For the sake of brevity, it will not be repeated here.

[0166] The present disclosure also provides a computer-readable storage medium having computer program instructions stored thereon, wherein the computer program instructions implement the above method when executed by a processor. The computer-readable storage medium may be a non-volatile computer-readable storage medium.

[0167] An embodiment of the present disclosure further proposes an electronic device, comprising: a processor; and a memory for storing instructions executable by the processor; wherein the processor is configured to perform the above method.

[0168] The electronic device may be provided as a terminal, a server, or other forms of devices.

[0169] 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 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, or the like.

[0170] Reference 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 .

[0171] The processing component 802 generally controls the overall operation of the electronic device 800, such as operations associated with display, phone calls, data communications, camera operation, and recording operations. The processing component 802 may include one or more processors 820 to execute instructions to perform all or part of the steps of the above-described method. In addition, the processing component 802 may include one or more modules to facilitate interaction between the processing component 802 and other components. For example, the processing component 802 may include a multimedia module to facilitate interaction between the multimedia component 808 and the processing component 802.

[0172] The memory 804 is configured to store various types of data to support operations on the electronic device 800. Examples of such 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 can 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, magnetic disk, or optical disk.

[0173] The power supply component 806 provides power to the various components of the electronic device 800. The power supply component 806 may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power to the electronic device 800.

[0174] 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 touch, slide, and gestures on the touch panel. The touch sensor can not only sense the boundaries of the touch or slide action, but also detect the duration and pressure associated with the touch or slide 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 operating mode, such as a shooting mode or a video mode, the front camera and / or the rear camera can receive external multimedia data. Each front camera and rear camera can be a fixed optical lens system or have a focal length and optical zoom capability.

[0175] The audio component 810 is configured to output and / or input audio signals. For example, the audio component 810 includes a microphone (MIC), which is configured to receive external audio signals when the electronic device 800 is in an operating mode, such as a call mode, a recording mode, and a voice recognition mode. The received audio signal can be further stored in the memory 804 or transmitted via the communication component 816. In some embodiments, the audio component 810 also includes a speaker for outputting audio signals.

[0176] I / O interface 812 provides an interface between processing component 802 and peripheral interface modules, such as a keyboard, click wheel, buttons, etc. These buttons may include but are not limited to: a home button, volume buttons, a start button, and a lock button.

[0177] The sensor assembly 814 includes one or more sensors for providing various aspects of status assessment for the electronic device 800. For example, the sensor assembly 814 can detect the open / closed state of the electronic device 800, the relative positioning of components, such as the display and keypad of the electronic device 800. The sensor assembly 814 can also detect changes 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 temperature changes of the electronic device 800. The sensor assembly 814 may include a proximity sensor configured to detect the presence of nearby objects without any physical contact. The sensor assembly 814 may also include a light sensor, such as a CMOS or CCD image sensor, for use in imaging applications. In some embodiments, the sensor assembly 814 may also include an accelerometer, a gyroscope sensor, a magnetic sensor, a pressure sensor, or a temperature sensor.

[0178] The communication component 816 is configured to facilitate wired or wireless communication between the electronic device 800 and other devices. The electronic device 800 can access a wireless network based on a communication standard, 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 also 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.

[0179] In an exemplary embodiment, the electronic device 800 may 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 to perform the above methods.

[0180] In an exemplary embodiment, a non-volatile computer-readable storage medium is also provided, such as a memory 804 including computer program instructions. The computer program instructions can be executed by the processor 820 of the electronic device 800 to perform the above method.

[0181] Figure 8 FIG1 shows a block diagram of an electronic device 1900 according to an embodiment of the present disclosure. For example, the electronic device 1900 may be provided as a server. Figure 8 The electronic device 1900 includes a processing component 1922, which further includes one or more processors, and a memory resource represented by a memory 1932 for storing instructions executable by the processing component 1922, such as an application. The application stored in the memory 1932 may include one or more modules, each corresponding to a set of instructions. In addition, the processing component 1922 is configured to execute the instructions to perform the above-described method.

[0182] The electronic device 1900 may further include a power supply 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 may operate based on an operating system stored in the memory 1932, such as Windows Server™, Mac OS X™, Unix™, Linux™, FreeBSD™, or the like.

[0183] In an exemplary embodiment, a non-volatile computer-readable storage medium is also provided, such as a memory 1932 including computer program instructions that can be executed by the processing component 1922 of the electronic device 1900 to perform the above method.

[0184] The present disclosure may be a system, method and / or computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for causing a processor to implement various aspects of the present disclosure.

[0185] A computer-readable storage medium can be a tangible device that can hold and store instructions for use by an instruction execution device. A computer-readable storage medium can 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 thereof. More specific examples (a non-exhaustive list) of computer-readable storage media 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 disk (DVD), a memory stick, a floppy disk, a mechanical encoding device, such as a punch card or a raised structure in a groove on which instructions are stored, and any suitable combination thereof. As used herein, a computer-readable storage medium is not to be construed as a transient signal per se, such as a radio wave or other freely propagating electromagnetic wave, an electromagnetic wave propagating through a waveguide or other transmission medium (e.g., a light pulse through a fiber optic cable), or an electrical signal transmitted through an electrical wire.

[0186] The computer-readable program instructions described herein can be downloaded from a computer-readable storage medium to each computing / processing device, or downloaded to an external computer or external storage device via a network, such as the Internet, a local area network, a wide area network, and / or a wireless network. The network can include copper transmission cables, fiber optic transmission, wireless transmission, routers, firewalls, switches, gateway computers, and / or edge servers. The network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards the computer-readable program instructions to be stored in the computer-readable storage medium in each computing / processing device.

[0187] The computer program instructions for performing the operations of the present disclosure may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent 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++, and conventional procedural programming languages such as "C" language or similar programming languages. Computer-readable program instructions may be executed entirely on a user's computer, partially on a user's computer, as an independent software package, partially on a user's computer, 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 via 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., utilizing an Internet service provider to connect via the Internet). In some embodiments, an electronic circuit, such as a programmable logic circuit, a field programmable gate array (FPGA), or a programmable logic array (PLA), may be personalized by utilizing the state information of the computer-readable program instructions. The electronic circuit may execute the computer-readable program instructions, thereby realizing various aspects of the present disclosure.

[0188] Various aspects of the present disclosure are described herein with reference to 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.

[0189] 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 device, thereby producing a machine, so that when these instructions are executed by the processor of the computer or other programmable data processing device, a device is generated that implements the functions / actions specified in one or more blocks in the flowchart and / or block diagram. These computer-readable program instructions can also be stored in a computer-readable storage medium, where these instructions cause the computer, programmable data processing device, and / or other device to operate in a specific manner. Thus, the computer-readable medium storing the instructions comprises an article of manufacture that includes instructions for implementing various aspects of the functions / actions specified in one or more blocks in the flowchart and / or block diagram.

[0190] Computer-readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device so that a series of operational steps are performed on the computer, other programmable data processing apparatus, or other device to produce a computer-implemented process, thereby causing the instructions executed on the computer, other programmable data processing apparatus, or other device to implement the functions / actions specified in one or more blocks in the flowchart and / or block diagram.

[0191] The flow charts and block diagrams in the accompanying drawings show the possible architecture, functions and operations of the systems, methods and computer program products according to multiple embodiments of the present disclosure. In this regard, each box in the flow chart or block diagram can represent a part of a module, program segment or instruction, and the part of the module, program segment or instruction contains one or more executable instructions for realizing the prescribed logical function. In some alternative implementations, the functions marked in the box can also occur in a sequence different from that marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flow chart, and the combination of the boxes in the block diagram and / or flow chart can be implemented by a dedicated hardware-based system that performs the prescribed function or action, or can be implemented by a combination of dedicated hardware and computer instructions.

[0192] While various embodiments of the present disclosure have been described above, the foregoing description is intended to be illustrative, non-exhaustive, and not limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is selected to best explain the principles of the embodiments, their practical applications, or technical improvements to existing technologies, or to enable others skilled in the art to understand the embodiments disclosed herein.

Claims

1. A method for constructing a prognostic status prediction model, characterized in that: include: Obtaining first image features corresponding to regions of interest of brain images at multiple moments corresponding to each brain perfusion image after bone removal and registration in a brain perfusion image set; wherein the region of interest includes: an abnormal region and a normal region; determining the abnormal region of the region of interest includes: segmenting the abnormal region of the region of interest from the brain perfusion image; or determining the abnormal region based on one or more corresponding dynamic parameters of cerebral blood flow, cerebral blood volume, mean transit time, time to peak, and maximum residual functional time corresponding to the region of interest in the brain perfusion image; When performing first image feature extraction processing corresponding to the regions of interest of the brain images at the multiple moments, optimizing the time sequence corresponding to the multiple moments, including: performing mean processing on the brain images corresponding to the stage where the brain images are not affected by the contrast agent during the perfusion imaging process and the process where the grayscale values of the pixels return to their initial state after the contrast agent leaves the pixel; using the brain perfusion images after the mean processing and the brain perfusion images at the stage where the contrast agent flows through the blood vessels and causes the grayscale values of the pixels to change as the brain perfusion images for feature extraction processing; Screening out a second image feature that satisfies a preset condition from the first image feature; wherein screening out the second image feature that satisfies the preset condition from the first image feature comprises: determining a feature item of the second image feature based on a multi-level feature selection strategy; screening out the feature item from the first image feature to obtain the second image feature; wherein determining the feature item of the second image feature based on the multi-level feature selection strategy comprises: extracting first image features of the region of interest based on the multiple moments; selecting a salient feature that satisfies significance from the first image features; screening out a third image feature that satisfies a selection condition of the feature selection method from the salient features based on at least two feature selection methods; and selecting a feature item of the second image feature that is used to distinguish normal tissue in a normal region of the region of interest from abnormal tissue in the abnormal region using at least one classification model from the third image feature; Obtaining a survival feature based on screening out the second image feature that meets a preset condition from the first image feature; Based on the second image feature, the survival feature and the functional recovery score corresponding to each cerebral perfusion image, a machine learning model is trained to obtain a prediction model for evaluating the functional recovery score corresponding to the prognostic status.

2. The construction method according to claim 1, characterized in that The obtaining of the survival feature based on screening out the second image feature that meets a preset condition from the first image feature includes: Predicting the prognostic risk of the second imaging feature using a survival model; The survival signature is determined based on the prognostic risk.

3. The construction method according to any one of claims 1 or 2, characterized in that The obtaining of the survival feature based on screening out the second image feature that meets a preset condition from the first image feature includes: Acquire clinical text information corresponding to the cerebral perfusion image set; wherein the clinical text information includes: age; Predicting the prognostic risk of the second imaging feature under the age corresponding to the clinical text information using a survival model; The survival signature is determined based on the prognostic risk.

4. The construction method according to claim 1, characterized in that The obtaining of the first image feature of the region of interest of each cerebral perfusion image in the cerebral perfusion image set includes: determining a region of interest of the cerebral perfusion image, wherein the region of interest includes abnormal tissue; Based on the brain images at multiple moments in the brain perfusion image, a first image feature of the region of interest is extracted.

5. The construction method according to claim 1, characterized in that Also includes: determining an optimal feature selection method when a second image feature that meets a preset condition is selected from the first image feature; as well as, The step of training a machine learning model based on the second image feature and the survival feature to obtain a prediction model for evaluating the prognosis status includes: selecting a prognostic feature related to the prognostic status from the second image features based on the optimal feature selection method; Based on the combined features of the prognostic features and the survival features, a machine learning model is trained to obtain a prediction model for evaluating the prognostic status.

6. The construction method according to claim 1, characterized in that Also includes: determining an optimal feature selection method when a second image feature that meets a preset condition is selected from the first image feature; as well as, The step of training a machine learning model based on the second image feature and the survival feature to obtain a prediction model for evaluating the prognosis status includes: selecting a prognostic feature related to the prognostic status from the second image feature and the clinical text information based on the optimal feature selection method; Based on the combined features of the prognostic features and the survival features, a machine learning model is trained to obtain a prediction model for evaluating the prognostic status.

7. The construction method according to claim 1, characterized in that Also includes: determining an optimal feature selection method when a second image feature that meets a preset condition is selected from the first image feature; as well as, The step of training a machine learning model based on the second image feature and the survival feature to obtain a prediction model for evaluating the prognosis status includes: selecting a prognostic feature related to the prognostic status from the second image features based on the optimal feature selection method; Based on the combined features of the prognostic features, clinical text features, and survival features, a machine learning model is trained to obtain a prediction model for evaluating the prognostic status.

8. A method for predicting the risk of prognosis, characterized in that: include: obtaining a first image feature of a region of interest in a cerebral perfusion image; Filtering out a second image feature that meets a preset condition from the first image feature; Obtaining a survival feature based on screening out the second image feature that meets a preset condition from the first image feature; A prediction model is used to predict the prognosis status within a preset time based on the second image feature and the survival feature, wherein the prediction model is obtained by the method for constructing a prognosis status prediction model according to any one of claims 1 to 7.

9. A device for constructing a prognosis status prediction model, characterized in that: include: A first acquisition module is configured to acquire first image features of a region of interest (ROI) of brain images corresponding to each cerebral perfusion image at multiple moments after bone removal and registration in a cerebral perfusion image set; wherein determining an abnormal region in the ROI includes: segmenting the abnormal region in the ROI from the cerebral perfusion image; or determining the abnormal region based on one or more corresponding dynamic parameters of cerebral blood flow, cerebral blood volume, mean transit time, time to peak, and maximum residual functional time corresponding to the ROI in the cerebral perfusion image; When performing first image feature extraction processing corresponding to the regions of interest of the brain images at the multiple moments, optimizing the time sequence corresponding to the multiple moments, including: performing mean processing on the brain images corresponding to the stage where the brain images are not affected by the contrast agent during the perfusion imaging process and the process where the grayscale values of the pixels return to their initial state after the contrast agent leaves the pixel; using the brain perfusion images after the mean processing and the brain perfusion images at the stage where the contrast agent flows through the blood vessels and causes the grayscale values of the pixels to change as the brain perfusion images for feature extraction processing; A first screening module is configured to screen out a second image feature that satisfies a preset condition from the first image feature; wherein screening out the second image feature that satisfies the preset condition from the first image feature comprises: determining a feature item of the second image feature based on a multi-level feature selection strategy; screening out the feature item from the first image feature to obtain the second image feature; wherein determining the feature item of the second image feature based on the multi-level feature selection strategy comprises: extracting first image features of the region of interest based on the multiple moments; selecting a significant feature that satisfies significance from the first image features; screening out a third image feature that satisfies selection conditions of the feature selection method from the significant features based on at least two feature selection methods; and selecting a feature item of the second image feature that is used to distinguish normal tissue in a normal region of the region of interest from abnormal tissue in the abnormal region using at least one classification model from the third image feature; a first survival module, configured to obtain a survival feature based on screening out the second image features that meet a preset condition from the first image features; A training module is used to train a machine learning model based on the second image feature, the survival feature and the functional recovery score corresponding to each cerebral perfusion image, and obtain a prediction model for evaluating the functional recovery score of the prognostic status.

10. A prognosis status prediction device, characterized in that: include: A second acquisition module is used to acquire a first image feature of a region of interest in a cerebral perfusion image; a second screening module, configured to screen out second image features that meet preset conditions from the first image features; a second survival module, configured to obtain a survival feature based on screening out the second image features that meet a preset condition from the first image features; A prediction module, configured to predict the prognosis status within a preset time period based on the second imaging feature and the survival feature using a prediction model; wherein the prediction model is obtained by the method for constructing a prognosis status prediction model according to any one of claims 1 to 7.

11. An electronic device, characterized in that: include: processor; a memory for storing processor-executable instructions; The processor is configured to call the instructions stored in the memory to execute the construction method described in any one of claims 1 to 7.

12. An electronic device, characterized in that: include: processor; a memory for storing processor-executable instructions; Wherein, the processor is configured to call the instructions stored in the memory to execute the risk prediction method described in claim 8.

Images