Scoring method, device and equipment

Abstract

The embodiments of this specification disclose a scoring method, device and equipment, which belong to the field of medical imaging and computer technology. The method includes: obtaining image data to be processed; preprocessing the image data to be processed to obtain preprocessed image data; inputting the preprocessed image data into a segmentation model to obtain a second segmentation result and a second classification result of the preprocessed image data; based on the second segmentation result and the second classification result, scoring the image data to be processed to obtain a scoring result. The method provided in the embodiments of this specification is used to eliminate or reduce human factors such as technical level, operation method, image quality, human eye resolution ability, fatigue level, different cognitive experience, etc., as well as diagnostic differences caused by imaging differences of imaging equipment, and reduce the time required for human observation, thinking and scoring.

Description

Technical Field

[0001] The present invention relates to the field of medical imaging and computer technology, and in particular to a scoring method, device and equipment. Background Art

[0002] ASPECTS (Alberta Stroke Program Early CT Score) is an important imaging index for evaluating core infarction, which can be used to quantitatively evaluate the severity of posterior circulation brain tissue damage in patients with ischemic stroke, with a total score of 10. The pc-ASPECTS score (The posterior circulation Acute Stroke Prognosis Early CTScore) is an early CT score for evaluating the prognosis of AIS in the posterior circulation of patients with ischemic stroke, with a total score of 10.

[0003] In the prior art, ASPECTS scoring or pc-ASPECTS scoring is mainly based on the doctor's experience and the naked eye to judge the infarction, so as to perform ASPECTS scoring or pc-ASPECTS scoring. Due to the existence of different imaging equipment, different technicians, different patient conditions and other factors, the consistency of cranial imaging data cannot be guaranteed. The judgment of cerebral infarction by naked eye will cause large differences due to subjectivity. At the same time, it is easy to make misjudgments when judging cerebral infarction. Therefore, this "naked eye judgment method" is not very practical.

[0004] Therefore, there is a need for a scoring method that can eliminate or reduce human factors such as technical level, operating methods, image quality, human eye resolution, fatigue level, different cognitive experiences, and diagnostic differences caused by differences in imaging equipment, and reduce the time required for human observation, thinking and scoring. Summary of the invention

[0005] The embodiments of this specification provide a scoring method, device and equipment for solving the following technical problems: eliminating or reducing human factors such as technical level, operation method, image quality, human eye resolution ability, fatigue level, different cognitive experience, and diagnostic differences caused by imaging differences of imaging equipment, and reducing the time required for human observation, thinking and scoring.

[0006] To solve the above technical problems, the embodiments of this specification are implemented as follows:

[0007] A scoring method provided in the embodiments of this specification includes:

[0008] Acquire image data to be processed;

[0009] Preprocessing the image data to be processed to obtain preprocessed image data;

[0010] Input the preprocessed image data into a segmentation model to obtain a second segmentation result and a second classification result of the preprocessed image data, wherein the segmentation model includes a neural network model and a post-processing model, wherein the neural network model includes an encoder, a classifier, and a decoder, wherein the second classification result is a target layer to which a target frame for scoring in the image data to be processed belongs, and the second segmentation result is a probability that each pixel in the target frame for scoring in the image data to be processed belongs to a target area, and the second segmentation result and the second classification result are results obtained simultaneously from the segmentation model;

[0011] Based on the second segmentation result and the second classification result, the image data to be processed is scored to obtain a scoring result.

[0012] Furthermore, if the image data to be processed is used for the anterior circulation, the category of the target layer is 3, and the number of the target areas is 20, wherein the category of the target layer includes the nucleus layer, the upper nucleus layer and the background layer, and the target areas include: the caudate nucleus head, the lenticular nucleus, the posterior limb of the internal capsule, the insular zone and the M1-M6 areas of the left and right brains.

[0013] Furthermore, if the image data to be processed is used for the posterior circulation, the category of the target layer is 4, the number of the target areas is 6, the category of the target layer includes the cerebellum layer, the midbrain layer, the thalamus layer and the background layer, and the target areas include: pons, bilateral cerebellum, midbrain, bilateral thalamus and posterior cerebral artery blood supply area.

[0014] Furthermore, the step of inputting the preprocessed image data into a segmentation model to obtain a second segmentation result and a second classification result of the preprocessed image data specifically includes:

[0015] Inputting the preprocessed image data into the encoder to obtain image features of the preprocessed image data;

[0016] Inputting the image feature into the decoder to obtain a first segmentation result of the preprocessed image data, wherein the first segmentation result is a segmentation result corresponding to all frames included in the preprocessed image data;

[0017] and / or

[0018] Inputting the image feature into the classifier to obtain a first classification result of the preprocessed image data, wherein the first classification result is a classification result corresponding to all frames included in the preprocessed image data;

[0019] The first classification result and the first segmentation result are processed by the post-processing model to obtain a second segmentation result and the second classification result of the image data to be processed.

[0020] Furthermore, the encoder includes a first convolution module, a global feature fusion module and a second convolution module.

[0021] Furthermore, the decoder performs deconvolution and concatenation operations based on the image features output by the encoder to obtain a first segmentation result of the preprocessed image data.

[0022] The present specification also provides a scoring device, including:

[0023] An acquisition module, for acquiring image data to be processed;

[0024] A preprocessing module, preprocessing the image data to be processed to obtain preprocessed image data;

[0025] a segmentation module, inputting the preprocessed image data into a segmentation model, obtaining a second segmentation result and a second classification result of the preprocessed image data, wherein the segmentation model includes a neural network model and a post-processing model, wherein the neural network model includes an encoder, a classifier, and a decoder, wherein the second classification result is a target layer to which a target frame for scoring in the image data to be processed belongs, and the segmentation result is a probability that each pixel in the target frame for scoring in the image data to be processed belongs to a target area, and the second segmentation result and the second classification result are results obtained simultaneously from the segmentation model;

[0026] A scoring module is used to score the image data to be processed based on the second segmentation result and the second classification result to obtain a scoring result.

[0027] Furthermore, if the image data to be processed is used for the anterior circulation, the category of the target layer is 3, and the number of the target areas is 20, wherein the category of the target layer includes the nucleus layer, the upper nucleus layer and the background layer, and the target areas include: the caudate nucleus head, the lenticular nucleus, the posterior limb of the internal capsule, the insular zone and the M1-M6 areas of the left and right brains.

[0028] Furthermore, if the image data to be processed is used for the posterior circulation, the category of the target layer is 4, the number of the target areas is 6, the category of the target layer includes the cerebellum layer, the midbrain layer, the thalamus layer and the background layer, and the target areas include: pons, bilateral cerebellum, midbrain, bilateral thalamus and posterior cerebral artery blood supply area.

[0029] Furthermore, the step of inputting the preprocessed image data into a segmentation model to obtain a second segmentation result and a second classification result of the preprocessed image data specifically includes:

[0030] Inputting the preprocessed image data into the encoder to obtain image features of the preprocessed image data;

[0031] Inputting the image feature into the decoder to obtain a first segmentation result of the preprocessed image data, wherein the first segmentation result is a segmentation result corresponding to all frames included in the preprocessed image data;

[0032] and / or

[0033] Inputting the image feature into the classifier to obtain a first classification result of the preprocessed image data, wherein the first classification result is a classification result corresponding to all frames included in the preprocessed image data;

[0034] The first classification result and the first segmentation result are processed by the post-processing model to obtain a second segmentation result and the second classification result of the image data to be processed.

[0035] Furthermore, the encoder includes a first convolution module, a global feature fusion module and a second convolution module.

[0036] Furthermore, the decoder performs deconvolution and concatenation operations based on the image features output by the encoder to obtain a first segmentation result of the preprocessed image data.

[0037] The embodiment of the present specification also provides an electronic device, including:

[0038] at least one processor; and,

[0039] a memory communicatively connected to the at least one processor; wherein,

[0040] The memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to:

[0041] Acquire image data to be processed;

[0042] Preprocessing the image data to be processed to obtain preprocessed image data;

[0043] Input the preprocessed image data into a segmentation model to obtain a second segmentation result and a second classification result of the preprocessed image data, wherein the segmentation model includes a neural network model and a post-processing model, wherein the neural network model includes an encoder, a classifier, and a decoder, wherein the second classification result is a target layer to which a target frame for scoring in the image data to be processed belongs, and the second segmentation result is a probability that each pixel in the target frame for scoring in the image data to be processed belongs to a target area, and the second segmentation result and the second classification result are results obtained simultaneously from the segmentation model;

[0044] Based on the second segmentation result and the second classification result, the image data to be processed is scored to obtain a scoring result.

[0045] The embodiment of the present specification obtains image data to be processed; pre-processes the image data to be processed to obtain pre-processed image data; inputs the pre-processed image data into a segmentation model to obtain a second segmentation result and a second classification result of the pre-processed image data, wherein the segmentation model includes a neural network model and a post-processing model, and the neural network model includes an encoder, a classifier and a decoder, the classification result is a target layer to which a target frame used for scoring in the image data to be processed belongs, the second segmentation result is the probability that each pixel point in the image data to be processed belongs to a target area in the target frame used for scoring, and the second segmentation result and the second classification result are results obtained from the segmentation model at the same time; based on the second segmentation result and the second classification result, the image data to be processed is scored to obtain a scoring result, eliminating or reducing human factors such as technical level, operation method, image quality, human eye resolution ability, fatigue level, different cognitive experience, etc., as well as diagnostic differences caused by imaging differences of imaging equipment, reducing the time required for human observation, thinking and scoring, and having better robustness and stronger generalization ability. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the embodiments of this specification or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this specification. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0047] Figure 1 A schematic diagram of a scoring method provided in an embodiment of this specification;

[0048] Figure 2 A schematic diagram of the ASPECTS scoring area provided in the embodiments of this specification;

[0049] Figure 3 A schematic diagram of the pc-ASPECTS scoring area provided in the embodiments of this specification;

[0050] Figure 4 A schematic diagram of a neural network model of a segmentation model provided in an embodiment of this specification;

[0051] Figure 5 A schematic diagram of a scoring method provided in Example 1 of this specification;

[0052] Figure 6 A schematic diagram of another scoring method provided in Example 2 of this specification;

[0053] Figure 7 A schematic diagram of another scoring method provided in Example 3 of this specification;

[0054] Figure 8 A schematic diagram of the training process of the neural network model of the segmentation model provided in the embodiment of this specification;

[0055] Fig. 9 A schematic diagram of a scoring device provided in an embodiment of this specification.

[0056] illustrate

[0057] C caudate nucleus head;

[0058] L lentiform nucleus;

[0059] IC internal capsule posterior limb;

[0060] I island belt;

[0061] M1 middle cerebral artery anterior cortical area;

[0062] M2 Middle cerebral artery lateral insular cortex;

[0063] M3 middle cerebral artery posterior cortical area;

[0064] M4 Middle cerebral artery cortex above M1;

[0065] M5 Middle cerebral artery cortex above M2;

[0066] M6 Middle cerebral artery cortex above M3;

[0067] Figure 3 a shows the pons and bilateral cerebellum from top to bottom;

[0068] Figure 3 b shows a schematic diagram of the midbrain region;

[0069] Figure 3c shows the blood supply areas of the bilateral thalamus and posterior cerebral arteries from top to bottom;

[0070] Figure 3 The numbers 1 and 2 in the table represent the weights of each region. DETAILED DESCRIPTION

[0071] In the prior art, imaging methods (MRI or CT, etc.) can be used to check whether the brain tissue has ischemic lesions. For patients with middle cerebral artery (MCA) occlusion, scoring can be performed on noncontrast CT (NCCT) to obtain ASPECTS results; for patients with vertebral basilar artery occlusion, scoring can be performed on diffusion-weighted MRI (DWI) to obtain pc-ASPECTS results.

[0072] Whether it is the anterior circulation ASPECTS score or the posterior circulation pc-ASPECTS score, the total score is 10. From the total score, the score corresponding to the infarct area is deducted to obtain the score result. In the anterior circulation ASPECTS score, the weight of the caudate nucleus head, lenticular nucleus, posterior limb of the internal capsule, insular zone and M1-M6 area of ​​the left and right brain used for scoring is 1; in the posterior circulation pc-ASPECTS score, the weight of each target area is also different. Specifically, the weight of the pons and midbrain is 2, and the weight of other areas is 1.

[0073] Based on NCCT image data or MR image data, rapid scoring is achieved, which has important clinical significance. Based on this, the embodiment of this specification provides a scoring method, which can achieve rapid scoring of NCCT image data or MR image data.

[0074] In order to enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below in conjunction with the drawings in the embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this application, not all of them. Based on the embodiments of this specification, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of this application.

[0075] Figure 1 A schematic diagram of a scoring method provided in an embodiment of this specification, the scoring method includes:

[0076] Step S101: Obtain image data to be processed.

[0077] In the embodiments of the present specification, the image data to be processed is cranial image data, specifically NCCT (non contrast CT) image data or MR image data. In a specific implementation, the MR image data is preferably DWI image data (diffusion-weighted MRI, DWI).

[0078] In one embodiment of the present specification, the image data to be processed is NCCT image data or DWI image data.

[0079] Step S103: pre-processing the image data to be processed to obtain pre-processed image data.

[0080] In the embodiment of the present specification, preprocessing the image data to be processed to obtain the preprocessed image data specifically includes:

[0081] After removing the skull from the image data to be processed, normalization is performed to obtain the preprocessed image data.

[0082] Since there are irrelevant tissues such as skulls in the image data to be processed, in order to ensure the accuracy of subsequent scoring, it is necessary to remove the skull in the image data to be processed. In the specific implementation process, the skull in the image data to be processed is removed, specifically including: by threshold segmentation, according to the first threshold, the skull is extracted from the first image, a skull mask image is obtained, and the brain is divided into two parts, the inside of the skull and the outside of the skull. In practical applications, the threshold for extracting the skull is >100. Furthermore, the pixels below the second threshold belong to the inside of the skull, and the skull is taken out from the skull mask image to obtain a tissue mask image after removing the skull. In the specific implementation process, the second threshold can be 80. To remove the skull in the image data to be processed, other methods of removing the skull can also be adopted, and the specific method of skull removal does not constitute a limitation of the present application.

[0083] In the embodiments of this specification, the normalization process includes: one or more of coordinate centering, x-shearing normalization, scaling normalization or rotation normalization. The normalization process may also adopt other methods, and the specific method of the normalization process does not constitute a limitation to this application.

[0084] Step S105: Input the preprocessed image data into a segmentation model to obtain a second segmentation result and a second classification result of the preprocessed image data, wherein the segmentation model includes a neural network model and a post-processing model, wherein the neural network model includes an encoder, a classifier and a decoder, and the second classification result is a target layer to which a target frame used for scoring in the image data to be processed belongs. The second segmentation result is a probability that each pixel point in the target frame used for scoring in the image data to be processed belongs to a target area, and the second segmentation result and the second classification result are results obtained simultaneously from the segmentation model.

[0085] In the embodiment of the present specification, the image data to be processed is used for the anterior circulation, then the category of the target layer is 3, and the number of the target areas is 20, wherein the category of the target layer includes the nuclear layer, the upper nuclear layer and the background layer, and the target areas include: the caudate nucleus head, the lenticular nucleus, the posterior limb of the internal capsule, the insular band and the M1-M6 areas of the left and right brains. Specifically, when the image data to be processed is used for ASPECT scoring of the anterior circulation, the target layer includes the nuclear layer and the upper nuclear layer, as well as the background layer as the background, wherein the upper nuclear layer is located 2 cm above the nuclear layer. The target areas located in the nuclear layer and the upper nuclear layer mainly include: M1, M2, M3, insula, lenticular nucleus, caudate nucleus and posterior limb of the internal capsule of the left and right brains located in the nuclear layer, a total of 14 areas, and the target areas located in the upper nuclear layer include M4, M5, M6 areas of the left and right brains, a total of 6 areas. Figure 2 A schematic diagram of the ASPECTS scoring area provided in the embodiment of this specification is provided to further understand the locations of the 20 target areas when the image data to be processed in the embodiment of this specification is used for the anterior circulation.

[0086] In the embodiment of the present specification, the image data to be processed is used for the posterior circulation, the category of the target level is 4, the number of the target areas is 6, the category of the target level includes the cerebellum level, the midbrain level, the thalamus level and the background level, and the target area includes: pons, bilateral cerebellum, midbrain, bilateral thalamus and posterior cerebral artery blood supply area. Specifically, when the image data to be processed is used for pc-ASPECT scoring of the posterior circulation, the target levels include the cerebellum level, the midbrain level, the thalamus level and the background level as the background. Figure 3 A schematic diagram of the pc-ASPECTS scoring area provided in the embodiments of this specification, wherein: Figure 3 a is a schematic diagram of the pons and bilateral cerebellum regions of the pc-ASPECTS score provided in the examples of this specification, Figure 3 b is a schematic diagram of the midbrain region of the pc-ASPECTS score provided in the examples of this specification, Figure 3c is a schematic diagram of the bilateral thalamus and posterior cerebral artery blood supply area of ​​the pc-ASPECTS score provided in the embodiment of this specification, so as to further understand the locations of the six target areas when the image data to be processed in the embodiment of this specification is used for the posterior circulation.

[0087] In the embodiment of the present specification, the inputting the preprocessed image data into the segmentation model to obtain the second segmentation result and the second classification result of the preprocessed image data specifically includes:

[0088] Inputting the preprocessed image data into the encoder to obtain image features of the preprocessed image data;

[0089] Inputting the image feature into the decoder to obtain a first segmentation result of the preprocessed image data, wherein the first segmentation result is a segmentation result corresponding to all frames included in the preprocessed image data;

[0090] and / or

[0091] Inputting the image feature into the classifier to obtain a second classification result of the preprocessed image data, wherein the first classification result is a classification result corresponding to all frames included in the preprocessed image data;

[0092] The first classification result and the first segmentation result are processed by the post-processing model to obtain a second segmentation result and the second classification result of the image data to be processed.

[0093] It should be noted that after the image data to be processed is preprocessed, the preprocessed image data is obtained; the preprocessed image data is input into the neural network model, and after being processed by the neural network model, the neural network model simultaneously outputs the first classification result and the first segmentation result, and the first classification result and the first segmentation result are obtained at the same time. The first classification result and the first segmentation result, after being processed by the post-processing model, generate the second classification result and the second segmentation result. Therefore, the second segmentation result and the second classification result are the results obtained from the segmentation model at the same time, that is, after the preprocessed image data is input into the segmentation model, the second segmentation result and the second classification result corresponding to the target frame used for scoring in the preprocessed image data or the image data to be processed will be obtained at the same time. The method provided in the embodiment of this specification is used to obtain the segmentation result and the classification result at the same time. Compared with the method of obtaining the segmentation result and the classification result separately, the method has a faster processing speed, better processing results, and the obtained segmentation result and classification result have higher accuracy, better robustness, and stronger generalization ability.

[0094] In the embodiment of the present specification, the encoder includes a first convolution module, a global feature fusion module and a second convolution module.

[0095] In the embodiment of the present specification, the decoder performs deconvolution and connection operations based on the image features output by the encoder to obtain the first segmentation result of the preprocessed image data. It should be noted that when the image data to be processed is used for the ASPECTS scoring of the front cycle, the number of target areas is 20. In addition, the decoder will also output non-target areas, or background areas, so the number of images of the first segmentation result output by the decoder is 21. When the image data to be processed is used for the pc-ASPECTS scoring of the back cycle, the number of target areas is 6. In addition, the decoder will also output non-target areas, or background areas, so the number of images of the first segmentation result output by the decoder is 7.

[0096] In the embodiment of the present specification, the second classification result is the probability that the target frame used for scoring in the image data to be processed belongs to the target area, which is expressed in the form of probability. The second segmentation result is the probability that each pixel point in the target frame used for scoring in the image data to be processed belongs to the target layer, which is expressed in the form of probability. In one embodiment of the present specification, for n frames of NCCT cranial image data, the output of the first classification result of the segmentation model is: P1, P2, ..., P n , where P i =[P i1 ,P i2 ,P i3 ],i=1,2,...,n,P i1 represents the probability that the i-th frame image belongs to other layers, P i2 represents the probability that the i-th frame image belongs to the nucleus layer, P i3 represents the probability that the i-th frame image belongs to the upper layer of the nucleus. In another embodiment of the present specification, for n frames of DWI cranial brain image data, the output of the first classification result of the segmentation model is: P1, P2, ..., P n , where P i =[P i1 ,P i2 ,P i3 , P i4 ],i=1,2,...,n,P i1 represents the probability that the i-th frame image belongs to other layers, P i2 represents the probability that the i-th frame image belongs to the cerebellum level, P i3 represents the probability that the i-th frame image belongs to the midbrain level, P i4 Represents the probability that the i-th frame image belongs to the thalamus level.

[0097] In addition, it should be further explained that the segmentation model in the embodiment of this specification includes a post-processing model in addition to the neural network model. The first classification result and the first segmentation result output by the neural network model are processed by the post-processing model to output the second classification result and the second segmentation result, which are used as the final output results of the segmentation model for subsequent scoring.

[0098] In the embodiment of the present specification, the preprocessed image data is input into the segmentation model to obtain the second segmentation result and the second classification result of the preprocessed image data, specifically including:

[0099] Inputting the preprocessed image data into the encoder to obtain image features of the preprocessed image data;

[0100] Inputting the image feature into the decoder to obtain a first segmentation result of the preprocessed image data, wherein the first segmentation result is a segmentation result corresponding to all frames included in the preprocessed image data;

[0101] and / or

[0102] The image features are input into the classifier to obtain a first classification result of the preprocessed image data, where the first classification result is the classification result corresponding to all frames contained in the preprocessed image data; the first classification result and the first segmentation result are processed by the post-processing model of the segmentation model to obtain a second segmentation result and the second classification result of the image data to be processed.

[0103] In the embodiments of the present specification, after the image data to be processed or the preprocessed image data is input into the segmentation model, it is necessary to determine the classification results and segmentation results corresponding to all frames contained in the preprocessed image data, that is, the first classification result and the first segmentation result; further, the target layer to which the target frame used for scoring in the image data to be processed belongs and the probability that each pixel point in the target frame used for scoring in the image data to be processed belongs to the target area are determined from the first classification result, which is also referred to as the second classification result and the second segmentation result, that is, the second classification result is determined from the first classification result, and the second segmentation result is determined from the first segmentation result and the second segmentation result.

[0104] In order to further understand the determination of the second classification result from the first classification result provided by the embodiment of this specification, the following will be described in conjunction with a specific embodiment. Specifically, the output results of the target level corresponding to the first classification result are sorted in order from large to small or from small to large, and the frames corresponding to the largest 2 or 3 values ​​are selected from the sorting results as the target level for scoring, that is, the second classification results corresponding to the frames corresponding to the largest 2 or 3 values ​​are selected from the sorting results as the second classification results. For the ASPECTS score of the front cycle, based on the first classification result, the frames corresponding to the nucleus layer and the frames corresponding to the upper layer of the nucleus for ASPECTS scoring are selected. It should be particularly noted that in the front cycle, if the partial frames of the upper layer of the nucleus selected for ASPECTS scoring are also frames belonging to the nucleus layer, the frames belonging to the nucleus layer should be removed from the upper layer of the nucleus for ASPECTS scoring. For the pc-ASPECTS score of the back cycle, based on the first classification result, the frames corresponding to the cerebellum level, the midbrain level, and the thalamus level for pc-ASPECTS scoring are selected. It should be noted that in the posterior loop, if some of the frames at the midbrain level selected for pc-ASPECTS scoring are also frames at the cerebellum level, the frames at the cerebellum level should be removed from the midbrain level used for pc-ASPECTS scoring; if some of the frames at the thalamus level selected for pc-ASPECTS scoring are also frames at the midbrain level, the frames at the midbrain level should be removed from the thalamus level used for pc-ASPECTS scoring.

[0105] Continuing with the previous example, for the ASPECTS score of the anterior circulation, based on the first classification result, it is necessary to further determine the target level involved in the ASPECTS score, that is, to determine the second classification result from the first classification result. Specifically, the selection of the nuclear layer for ASPECTS scoring: 12 ,P 22 ,...,P n2 Sorting is performed, and the frames corresponding to the largest two or three values ​​are selected from the sorting results as the nucleus layer. 13 ,P 23 ,...,P n3 Sorting is performed, and the frames corresponding to the largest two or three values ​​are selected from the sorting results as the upper layer of the nucleus.

[0106] Similarly, for the pc-ASPECTS score of the posterior circulation, based on the first classification result, it is necessary to further determine the target layer involved in the pc-ASPECTS score, that is, to determine the second classification result from the first classification result. 12 ,P22 ,...,P n2 Sorting is performed, and the frames corresponding to the largest two or three values ​​are selected from the sorting results as the nucleus layer. 13 ,P 23 ,...,P n3 Sorting is performed, and the frames corresponding to the largest two or three values ​​are selected from the sorting results as the midbrain level. 14, P 24 ,...,P n4 Sorting is performed, and the frames corresponding to the largest two or three values ​​are selected from the sorting results as the thalamus level.

[0107] In the embodiment of the present specification, after determining the second classification result based on the first classification result, further, based on the first segmentation result, it is necessary to determine the second segmentation result according to the second classification result. Specifically, after determining the first classification result, the first segmentation result corresponding to the target frame in the first classification result is selected from the first segmentation result as the second segmentation result.

[0108] In order to further understand the segmentation model provided in the embodiments of this specification, a detailed description will be given below using a structural diagram as an example. Figure 4 A neural network model schematic diagram of a segmentation model provided in an embodiment of this specification. Figure 4 As shown, the preprocessed image data is input into the encoder, the image features of the preprocessed image data are extracted, the image features corresponding to the preprocessed image data are obtained, and the encoder outputs the image features of the preprocessed image data. The image features of the preprocessed image data are input into the classifier, and the first classification result of the image data to be processed is output. At the same time, the image features of the preprocessed image data are input into the decoder, and the first segmentation result of the image data to be processed is output. It should be particularly noted that the image features corresponding to the preprocessed image data are one or more of the size of the image corresponding to the preprocessed image data and the grayscale of the image. In order to further understand the specific application of the segmentation model provided in the embodiments of this specification, NCCT or MR image data is taken as an example and described in detail with specific embodiments.

[0109] Example 1

[0110] The image data to be processed is normalized and resampled, and then input into the neural network model of the segmentation model as a 512*512*1 grayscale image, that is, the input is a 512*512 grayscale image with 1 channel. Figure 5A schematic diagram of a scoring method provided in Example 1 of this specification. In the example of this specification, the encoder (in the dotted box) of the neural network model is composed of 4 convolution blocks, and each convolution block has a convolution operation and a pooling operation. Each convolution block of the encoder uses a 3*3 convolution kernel, and after the convolution operation, a dropout (random inactivation) operation, a normalization (normalization) operation, an activation operation, etc. can be performed. The pooling operation can be a max pool (maximum pool) or an average pool (average pool), etc. In the current example 1, the step size of the pooling operation is 2, and each time a pooling operation is performed, the size of the feature map is reduced by half. The first convolution block uses 32 convolution kernels to obtain 32 feature maps. After the pooling operation, the image is reduced from 512*512 to 256*256; the second convolution block uses 64 convolution kernels and obtains 64 feature maps. After the pooling operation, the image is reduced from 256*256 to 128*128; the third convolution block uses 128 convolution kernels and obtains 128 feature maps. After the pooling operation, the image is reduced from 256*256 to 64*64; the size of the feature map output by the last convolution block (the fourth convolution block) is 32*32, with a total of 256 feature maps, which are output to the classification module. The first layer of the classification module can be a global average pooling layer, which takes the average of the feature maps of each channel. Through the Global Average Pooling layer, the feature map output by the encoder is converted into a vector of length 256. Through the full connection operation and the softmax activation function, the first classification result is output.

[0111] In Example 1 of this specification, the decoder (in the dotted box on the right) is mainly composed of four deconvolution blocks. For the output of the fourth convolution block in the encoder, in addition to being used as the input of the classification module, it is also used as the input of the first deconvolution block of the decoder. The step size of the deconvolution operation is 2. After each deconvolution operation, the size of the feature map doubles from 32*32 to 64*64. The first deconvolution block uses 128 convolution kernels to obtain 128 feature maps. The output of the first deconvolution block is connected to the output of the third convolution block to obtain 256 feature maps of 64*64 size, which are input to the second deconvolution block. The second deconvolution block uses 64 convolution kernels to obtain 64 feature maps. The output of the second deconvolution block is connected to the output of the second convolution block to obtain 128 feature maps of 128*128 size, which are input to the third deconvolution block. The third deconvolution block uses 32 convolution kernels to obtain 32 feature maps. The output of the third deconvolution block is connected to the output of the first convolution block to obtain 64 feature maps of size 256*256, which are input to the fourth deconvolution block. Through continuous deconvolution and connection, the size of the output feature map becomes larger and larger. In this embodiment, the fourth deconvolution block uses 64 convolution kernels, and the output is restored to the size of the input image. After the final convolution layer, the softmax activation function is used to finally output the first segmentation result.

[0112] Example 2

[0113] Figure 6 A schematic diagram of another scoring method provided for Example 2 of this specification. In Example 2 of this specification, the decoder and classification module are the same as those in Example 1, and the difference between Example 2 and Example 1 lies in the encoder part. In Example 2 of this specification, the step size of the first convolution block is 4. After the first deconvolution block, the size of the feature map is increased by 4 times, from 32*32 to 128*128, and its output is connected to the output of the second convolution block and input into the second convolution block. The subsequent deconvolution operation is the same as that in Example 1.

[0114] Example 3

[0115] Figure 7A schematic diagram of another scoring method provided for Example 3 of this specification. In Example 3 of this specification, the encoder consists of a first convolution module, a global feature fusion module and a second convolution module. The input image is a grayscale image of 512*512*1. The first encoder (in the dotted box) consists of three convolution blocks. Each convolution block has a convolution operation and a pooling operation. Generally, a 3*3 convolution kernel is used. After the convolution operation, a dropout operation, a normalization operation, an activation operation, etc. can be performed. The pooling operation can be a max pool or an average pool, etc. In the current Example 3, the step size of the pooling operation is 2. The first convolution block uses 32 convolution kernels to obtain 32 feature maps. After the pooling operation, the image is reduced from 512*512 to 256*256; the second convolution block uses 64 convolution kernels and obtains 64 feature maps. After the pooling operation, the image is reduced from 256*256 to 128*128; the third convolution block uses 128 convolution kernels and obtains 128 feature maps. After the pooling operation, the image is reduced from 256*256 to 64*64. Since most of the information extracted by the convolution operation is local feature information, a global feature fusion module is set after the first convolution module, and a second convolution module is set after the global feature fusion module.

[0116] In this third embodiment, the classification module is the same as that in the first embodiment.

[0117] Since the encoder in this embodiment 3 is provided with a global feature fusion module, correspondingly, a global feature fusion module is also provided after the first deconvolution block of the decoder. After the output of the global feature fusion module is connected with the output of the third convolution block of the encoder, it is input into the next deconvolution block, and the subsequent deconvolution operation is the same as that in embodiment 1.

[0118] In the embodiments of this specification, the neural network model of the segmentation model is a model obtained by neural network training. Embodiment 1, Embodiment 2 and Embodiment 3 are only schematic illustrations of this specification. In the specific implementation process, Embodiment 1 is preferred. The first classification result and the first segmentation result output by Embodiment 1, Embodiment 2 and Embodiment 3 are only the classification result and the segmentation result corresponding to the image data to be processed. The first classification result and the first segmentation result need to be further processed by the post-processing model to obtain the segmentation result corresponding to the target frame and the segmentation result corresponding to the target frame, that is, the second classification result and the second segmentation result.

[0119] Figure 8 Schematic diagram of the training process of the neural network model of the segmentation model provided in the embodiment of this specification. Figure 8 As shown, the training of the neural network model of the segmentation model includes:

[0120] Step S801: Divide the acquired learning sample set data into training set data, tuning set data and test set data.

[0121] In the embodiment of the present specification, the learning sample set data is cranial brain imaging data, which can be specifically NCCT imaging data or MR imaging data. The learning sample set data is labeled. When labeling, classification labels and segmentation labels need to be labeled. Among them, the classification labels of NCCT imaging data are used to label the nucleus layer, the upper layer of the nucleus and the background layer, and the segmentation labels of NCCT imaging data are used to label the caudate nucleus head, lenticular nucleus, posterior limb of the internal capsule, insular band and M1-M6 areas of the left and right brains; the classification labels of MR imaging data are used to label the cerebellum layer, midbrain layer, thalamus layer and background layer, and the segmentation labels of MR imaging data are used to label the pons, bilateral cerebellum, midbrain, bilateral thalamus and posterior cerebral artery blood supply area. In a specific embodiment, for NCCT image data, it is necessary to mark: the frame where the nucleus layer is located and the target area to which each pixel point on the frame belongs, the frame where the upper layer of the nucleus is located and the target area to which each pixel point on the frame belongs; for MR image data, it is necessary to mark: the frame where the cerebellum layer is located and the target area to which each pixel point on the frame belongs, the frame where the midbrain layer is located and the target area to which each pixel point on the frame belongs, the frame where the thalamus layer is located and the target area to which each pixel point on the frame belongs. The labeled learning sample set data is randomly divided into training set data, tuning set data and test set data in a ratio of 5:2:3.

[0122] Step S803: normalize the learning sample set data to obtain normalized learning sample set data.

[0123] The learning sample set data is normalized and resampled to obtain a 512*512 matrix, which is associated with the corresponding classification labels and segmentation labels of the scores.

[0124] Step S805: training the normalized learning sample set data to obtain a neural network model of the segmentation model.

[0125] Based on the constructed neural network, select the optimal hyperparameters. Input the image matrix in the training set data and the data corresponding to the classification label and the data corresponding to the segmentation label into the constructed neural network. First, use a low learning rate to warm up the model, and then gradually increase the learning rate. During the training process, the cross entropy cost function can be used as the loss function to train the model. During the training process, if the loss function on the tuning data no longer decreases, stop training to prevent overfitting. For each model structure, different hyperparameters are used. Multiple models are trained under the same initialization conditions. The average value of the loss function of the model on the tuning set data is taken as the evaluation indicator of the group of hyperparameters. The hyperparameter with the smallest average loss function is taken as the optimal hyperparameter for each model structure.

[0126] Subsequently, the segmentation model corresponding to the above optimal hyperparameters is tested using the test set data, and the neural network model of the optimal segmentation model is selected for subsequent scoring.

[0127] It should be noted that when the test set data is insufficient, cross-validation can also be used to select the neural network model of the optimal segmentation model.

[0128] It should be noted that the segmentation model provided in the embodiments of this specification is composed of a neural network model and a post-processing model connected in series. In addition to being a separate model, the various modules included in the segmentation model can also be used as separate models to form a segmentation model in an integrated manner, or in other words, the segmentation model is composed of multiple separate models.

[0129] Step S107: Based on the second segmentation result and the second classification result, the image data to be processed is scored to obtain a scoring result.

[0130] In the embodiment of the present specification, after determining the target layer for scoring based on the first classification result, the image data to be processed is scored based on the second segmentation result of each pixel in the target layer for scoring. That is, after obtaining the second classification result based on the first classification result, the image data to be processed is scored according to the second segmentation result corresponding to the second classification result.

[0131] In the embodiments of the present specification, the ASPECTS score of the anterior circulation or the pc-ASPECTS score of the posterior circulation adopts the following steps: based on the second segmentation result corresponding to the frame of the target layer used for scoring, an infarction judgment is performed to obtain an infarction judgment result; based on the infarction judgment result, according to the weight of each target area, the score of the infarction area is deducted from 10 points to obtain the final score.

[0132] In one embodiment of the present specification, for the ASPECTS score of the anterior circulation, based on the second segmentation result corresponding to the frame of the target layer for scoring, the average grayscale value of each frame of the target layer for scoring is determined as the grayscale value of each frame of the target layer for scoring. The difference method, the ratio method, the combination of the ratio and the difference method, and other methods are further used to determine infarction. In one embodiment of the present specification, an infarction determination model of the anterior circulation is used to determine infarction, wherein the infarction determination model of the anterior circulation is a pre-trained network model. It should be particularly noted that the infarction determination method of the anterior circulation does not constitute a limitation of the present application.

[0133] In the embodiments of the present specification, the pc-ASPECTS score for the posterior circulation is input into the infarction judgment model of the posterior circulation based on the second segmentation result corresponding to the frame of the target layer for scoring, and infarction judgment is performed, wherein the infarction judgment model of the posterior circulation is a pre-trained network model. It should be noted that the infarction judgment method of the posterior circulation does not constitute a limitation of the present application.

[0134] The scoring method provided in the embodiments of this specification can eliminate or reduce human factors such as technical level, operating method, image quality, human eye resolution ability, fatigue level, different cognitive experience, and diagnostic differences caused by imaging differences of imaging equipment, reduce the time required for human observation, thinking and scoring, and have better robustness and stronger generalization ability.

[0135] The above content describes in detail a scoring method. Correspondingly, this specification also provides a scoring device, such as Fig. 9 shown. Fig. 9 A schematic diagram of a scoring device provided in an embodiment of this specification, the device comprising:

[0136] An acquisition module 901 is used to acquire image data to be processed;

[0137] A preprocessing module 903 preprocesses the image data to be processed to obtain preprocessed image data;

[0138] The segmentation module 905 inputs the preprocessed image data into a segmentation model to obtain a second segmentation result and a second classification result of the preprocessed image data, wherein the segmentation model includes a neural network model and a post-processing model, wherein the neural network model includes an encoder, a classifier, and a decoder, wherein the second classification result is a target layer to which a target frame for scoring in the image data to be processed belongs, and the second segmentation result is a probability that each pixel in the target frame for scoring in the image data to be processed belongs to a target area, and the second segmentation result and the second classification result are results obtained simultaneously from the segmentation model;

[0139] The scoring module 907 scores the image data to be processed based on the second segmentation result and the second classification result to obtain a scoring result.

[0140] In the embodiment of the present specification, the image data to be processed is used for the anterior circulation, the category of the target layer is 3, and the number of the target areas is 20, wherein the category of the target layer includes the nuclear layer, the upper nuclear layer and the background layer, and the target areas include: the caudate nucleus head, the lenticular nucleus, the posterior limb of the internal capsule, the insular zone and the M1-M6 areas of the left and right brain.

[0141] In the embodiment of the present specification, the image data to be processed is used for the posterior circulation, the category of the target layer is 4, the number of the target areas is 6, the category of the target layer includes the cerebellum layer, the midbrain layer, the thalamus layer and the background layer, and the target areas include: pons, bilateral cerebellum, midbrain, bilateral thalamus and posterior cerebral artery blood supply area.

[0142] In the embodiment of the present specification, the inputting the preprocessed image data into the segmentation model to obtain the second segmentation result and the second classification result of the preprocessed image data specifically includes:

[0143] Inputting the preprocessed image data into the encoder to obtain image features of the preprocessed image data;

[0144] Inputting the image feature into the decoder to obtain a first segmentation result of the preprocessed image data, wherein the first segmentation result is a segmentation result corresponding to all frames included in the preprocessed image data;

[0145] and / or

[0146] Inputting the image feature into the classifier to obtain a first classification result of the preprocessed image data, wherein the first classification result is a classification result corresponding to all frames included in the preprocessed image data;

[0147] The first classification result and the first segmentation result are processed by the post-processing model to obtain a second segmentation result and the second classification result of the image data to be processed.

[0148] In the embodiment of the present specification, the encoder includes a first convolution module, a global feature fusion module and a second convolution module.

[0149] In an embodiment of the present specification, the decoder performs deconvolution and concatenation operations based on the image features output by the encoder to obtain a first segmentation result of the preprocessed image data.

[0150] This specification also provides an electronic device, including:

[0151] at least one processor; and,

[0152] a memory communicatively connected to the at least one processor; wherein,

[0153] The memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to:

[0154] Acquire image data to be processed;

[0155] Preprocessing the image data to be processed to obtain preprocessed image data;

[0156] Input the preprocessed image data into a segmentation model to obtain a second segmentation result and a second classification result of the preprocessed image data, wherein the segmentation model includes a neural network model and a post-processing model, wherein the neural network model includes an encoder, a classifier, and a decoder, wherein the second classification result is a target layer to which a target frame for scoring in the image data to be processed belongs, and the second segmentation result is a probability that each pixel in the target frame for scoring in the image data to be processed belongs to a target area, and the second segmentation result and the second classification result are results obtained simultaneously from the segmentation model;

[0157] Based on the second segmentation result and the second classification result, the image data to be processed is scored to obtain a scoring result.

[0158] The above is a description of a specific embodiment of the specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0159] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device, electronic device, and non-volatile computer storage medium embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiments.

[0160] The apparatus, electronic device, non-volatile computer storage medium and method provided in the embodiments of this specification correspond to each other, and therefore, the apparatus, electronic device and non-volatile computer storage medium also have similar beneficial technical effects as the corresponding method. Since the beneficial technical effects of the method have been described in detail above, the beneficial technical effects of the corresponding apparatus, electronic device and non-volatile computer storage medium will not be repeated here.

[0161] In the 1990s, improvements to a technology could be clearly distinguished as hardware improvements (for example, improvements to the circuit structure of diodes, transistors, switches, etc.) or software improvements (improvements to the method flow). However, with the development of technology, many improvements to the method flow today can be regarded as direct improvements to the hardware circuit structure. Designers almost always obtain the corresponding hardware circuit structure by programming the improved method flow into the hardware circuit. Therefore, it cannot be said that an improvement in a method flow cannot be implemented using a hardware entity module. For example, a programmable logic device (PLD) (such as a field programmable gate array (FPGA)) is such an integrated circuit whose logical function is determined by the user's programming of the device. Designers can "integrate" a digital system on a PLD by programming it themselves, without having to ask a chip manufacturer to design and produce a dedicated integrated circuit chip. Moreover, nowadays, instead of manually making integrated circuit chips, this kind of programming is mostly implemented by "logic compiler" software, which is similar to the software compiler used when developing and writing programs, and the original code before compilation must also be written in a specific programming language, which is called hardware description language (HDL). There is not only one HDL, but many kinds, such as ABEL (Advanced Boolean Expression Language), AHDL (Altera Hardware Description Language), Confluence, CUPL (Cornell University Programming Language), HDCal, JHDL (Java Hardware Description Language), Lava, Lola, MyHDL, PALASM, RHDL (Ruby Hardware Description Language), etc. The most commonly used ones are VHDL (Very-High-Speed ​​Integrated Circuit Hardware Description Language) and Verilog. Those skilled in the art should also know that it is only necessary to program the method flow slightly in the above-mentioned hardware description languages ​​and program it into the integrated circuit, and then it is easy to obtain the hardware circuit that implements the logic method flow.

[0162] The controller can be implemented in any appropriate manner, for example, the controller can take the form of a microprocessor or processor and a computer-readable medium storing a computer-readable program code (such as software or firmware) that can be executed by the (micro)processor, a logic gate, a switch, an application-specific integrated circuit (ASIC), a programmable logic controller, and an embedded microcontroller. Examples of controllers include, but are not limited to, the following microcontrollers: ARC 625D, Atmel AT91SAM, Microchip PIC18F26K20, and Silicone Labs C8051F320. The memory controller can also be implemented as part of the control logic of the memory. Those skilled in the art also know that in addition to implementing the controller in a purely computer-readable program code manner, the controller can be implemented in the form of a logic gate, a switch, an application-specific integrated circuit, a programmable logic controller, and an embedded microcontroller by logically programming the method steps. Therefore, this controller can be considered as a hardware component, and the devices included therein for implementing various functions can also be regarded as structures within the hardware component. Or even, the devices for implementing various functions can be regarded as both software modules for implementing the method and structures within the hardware component.

[0163] The systems, devices, modules or units described in the above embodiments may be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, the computer may be, for example, a personal computer, a laptop computer, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

[0164] For the convenience of description, the above devices are described in terms of functions and are divided into various units. Of course, when implementing one or more embodiments of this specification, the functions of each unit can be implemented in the same or multiple software and / or hardware.

[0165] Those skilled in the art will appreciate that the embodiments of this specification may be provided as methods, systems, or computer program products. Therefore, the embodiments of this specification may be in the form of complete hardware embodiments, complete software embodiments, or embodiments in combination with software and hardware. Moreover, the embodiments of this specification may be in the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0166] This specification is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of this specification. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0167] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0168] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0169] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0170] The memory may include non-permanent storage in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0171] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0172] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0173] The specification may be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The specification may also be practiced in distributed computing environments where tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules may be located in local and remote computer storage media including storage devices.

[0174] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0175] The above is only an embodiment of this specification and is not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.

Claims

1. A scoring method, characterized in that: The method comprises: Acquire image data to be processed; wherein the image data to be processed is cranial brain image data; Preprocessing the image data to be processed to obtain preprocessed image data; Input the preprocessed image data into a segmentation model to obtain a second segmentation result and a second classification result of the preprocessed image data, wherein the segmentation model includes a neural network model and a post-processing model, wherein the neural network model includes an encoder, a classifier, and a decoder, wherein the second classification result is a target layer to which a target frame for scoring in the image data to be processed belongs, and the second segmentation result is a probability that each pixel in the target frame for scoring in the image data to be processed belongs to a target area, and the second segmentation result and the second classification result are results obtained simultaneously from the segmentation model; Wherein, when the image data to be processed is used for the anterior circulation, the categories of the target layers include the nucleus layer, the upper nucleus layer and the background layer, and the target areas include: the caudate nucleus head, the lenticular nucleus, the posterior limb of the internal capsule, the insular band and the M1-M6 area of ​​the left and right brains; when the image data to be processed is used for the posterior circulation, the categories of the target layers include the cerebellum layer, the midbrain layer, the thalamus layer and the background layer, and the target areas include: the pons, the bilateral cerebellum, the midbrain, the bilateral thalamus and the posterior cerebral artery blood supply area; The preprocessed image data is input into the segmentation model to obtain a second segmentation result and a second classification result of the preprocessed image data, specifically including: Inputting the preprocessed image data into the encoder to obtain image features of the preprocessed image data; Inputting the image feature into the decoder to obtain a first segmentation result of the preprocessed image data; The image features are input into the classifier to obtain a first classification result of the preprocessed image data; wherein the encoder includes a first convolution module, a global feature fusion module and a second convolution module, and the decoder is provided with a corresponding global feature fusion module and a plurality of deconvolution blocks, wherein the output of the second convolution module is respectively connected to the input of the classifier and the input of the first deconvolution block in the decoder, and the output of the first deconvolution block in the decoder is connected with the output of the first convolution module through the output of the global feature fusion module provided by the decoder, and then input to the next deconvolution block in the decoder, so as to perform deconvolution and connection operations based on the image features output by the encoder until the last deconvolution block, so as to obtain a first segmentation result of the preprocessed image data; The first classification result and the first segmentation result are processed by the post-processing model to obtain a second segmentation result and a second classification result of the image data to be processed; Based on the second segmentation result and the second classification result, the image data to be processed is scored to obtain a scoring result.

2. The method according to claim 1, characterized in that The image data to be processed is used for the front cycle, then the category of the target layer is 3, and the number of the target areas is 20.

3. The method according to claim 1, characterized in that The image data to be processed is used for post-loop, then the category of the target layer is 4, and the number of the target areas is 6.

4. The method according to claim 1, characterized in that The first segmentation result is the segmentation result corresponding to all frames included in the preprocessed image data; and / or The first classification result is the classification result corresponding to all frames included in the preprocessed image data.

5. A scoring device, characterized in that: The device comprises: An acquisition module, for acquiring image data to be processed; wherein the image data to be processed is cranial brain image data; A preprocessing module, preprocessing the image data to be processed to obtain preprocessed image data; a segmentation module, inputting the preprocessed image data into a segmentation model, obtaining a second segmentation result and a second classification result of the preprocessed image data, wherein the segmentation model includes a neural network model and a post-processing model, wherein the neural network model includes an encoder, a classifier, and a decoder, wherein the second classification result is a target layer to which a target frame for scoring in the image data to be processed belongs, and the segmentation result is a probability that each pixel in the target frame for scoring in the image data to be processed belongs to a target area, and the second segmentation result and the second classification result are results obtained simultaneously from the segmentation model; Wherein, when the image data to be processed is used for the anterior circulation, the categories of the target layers include the nucleus layer, the upper nucleus layer and the background layer, and the target areas include: the caudate nucleus head, the lenticular nucleus, the posterior limb of the internal capsule, the insular band and the M1-M6 area of ​​the left and right brains; when the image data to be processed is used for the posterior circulation, the categories of the target layers include the cerebellum layer, the midbrain layer, the thalamus layer and the background layer, and the target areas include: the pons, the bilateral cerebellum, the midbrain, the bilateral thalamus and the posterior cerebral artery blood supply area; The preprocessed image data is input into the segmentation model to obtain a second segmentation result and a second classification result of the preprocessed image data, specifically including: Inputting the preprocessed image data into the encoder to obtain image features of the preprocessed image data; Inputting the image feature into the decoder to obtain a first segmentation result of the preprocessed image data; The image features are input into the classifier to obtain a first classification result of the preprocessed image data; wherein the encoder includes a first convolution module, a global feature fusion module and a second convolution module, and the decoder is provided with a corresponding global feature fusion module and a plurality of deconvolution blocks, wherein the output of the second convolution module is respectively connected to the input of the classifier and the input of the first deconvolution block in the decoder, and the output of the first deconvolution block in the decoder is connected with the output of the first convolution module through the output of the global feature fusion module provided by the decoder, and then input to the next deconvolution block in the decoder, so as to perform deconvolution and connection operations based on the image features output by the encoder until the last deconvolution block, so as to obtain a first segmentation result of the preprocessed image data; The first classification result and the first segmentation result are processed by the post-processing model to obtain a second segmentation result and a second classification result of the image data to be processed; A scoring module is used to score the image data to be processed based on the second segmentation result and the second classification result to obtain a scoring result.

6. The device according to claim 5, characterized in that The image data to be processed is used for the front cycle, then the category of the target layer is 3, and the number of the target areas is 20.

7. The device according to claim 5, characterized in that The image data to be processed is used for post-loop, then the category of the target layer is 4, and the number of the target areas is 6.

8. The device according to claim 5, characterized in that The first segmentation result is the segmentation result corresponding to all frames included in the preprocessed image data; and / or The first classification result is the classification result corresponding to all frames included in the preprocessed image data.

9. An electronic device, comprising: at least one processor; as well as, a memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to: Acquire image data to be processed; wherein the image data to be processed is cranial brain image data; Preprocessing the image data to be processed to obtain preprocessed image data; Input the preprocessed image data into a segmentation model to obtain a second segmentation result and a second classification result of the preprocessed image data, wherein the segmentation model includes a neural network model and a post-processing model, wherein the neural network model includes an encoder, a classifier, and a decoder, wherein the second classification result is a target layer to which a target frame for scoring in the image data to be processed belongs, and the second segmentation result is a probability that each pixel in the target frame for scoring in the image data to be processed belongs to a target area, and the second segmentation result and the second classification result are results obtained simultaneously from the segmentation model; Wherein, when the image data to be processed is used for the anterior circulation, the categories of the target layers include the nucleus layer, the upper nucleus layer and the background layer, and the target areas include: the caudate nucleus head, the lenticular nucleus, the posterior limb of the internal capsule, the insular band and the M1-M6 area of ​​the left and right brains; when the image data to be processed is used for the posterior circulation, the categories of the target layers include the cerebellum layer, the midbrain layer, the thalamus layer and the background layer, and the target areas include: the pons, the bilateral cerebellum, the midbrain, the bilateral thalamus and the posterior cerebral artery blood supply area; The preprocessed image data is input into the segmentation model to obtain a second segmentation result and a second classification result of the preprocessed image data, specifically including: Inputting the preprocessed image data into the encoder to obtain image features of the preprocessed image data; Inputting the image feature into the decoder to obtain a first segmentation result of the preprocessed image data; The image features are input into the classifier to obtain a first classification result of the preprocessed image data; wherein the encoder includes a first convolution module, a global feature fusion module and a second convolution module, and the decoder is provided with a corresponding global feature fusion module and a plurality of deconvolution blocks, wherein the output of the second convolution module is respectively connected to the input of the classifier and the input of the first deconvolution block in the decoder, and the output of the first deconvolution block in the decoder is connected with the output of the first convolution module through the output of the global feature fusion module provided by the decoder, and then input to the next deconvolution block in the decoder, so as to perform deconvolution and connection operations based on the image features output by the encoder until the last deconvolution block, so as to obtain a first segmentation result of the preprocessed image data; The first classification result and the first segmentation result are processed by the post-processing model to obtain a second segmentation result and a second classification result of the image data to be processed; Based on the second segmentation result and the second classification result, the image data to be processed is scored to obtain a scoring result.

Images