Processing method and apparatus for predicting aspects on basis of CT images

Abstract

A processing method and apparatus for predicting the ASPECTS on the basis of CT images. The method comprises: constructing four segmentation models, a classification model and a scoring model; after training is completed, receiving two CT images to form an image sequence; using the four segmentation models to separately perform image segmentation on the image sequence, so as to obtain first / second / third / fourth semantic map groups, and using the classification model to classify the image sequence, so as to obtain a predicted laterality group; performing comprehensive assessment on laterality information; performing left and right brain regional image extraction to obtain a regional image set; performing median filtering on the regional image set; on the basis of the first / second semantic map groups, eliminating sulcal / old infarction regions from the regional image set; extracting left and right brain regional image pairs; using the scoring model to score a regional image pair sequence; and on the basis of assessed laterality and a regional score pair sequence, performing overall score / affected-side summarization to obtain a summary report. By means of the present method, the prediction accuracy can be improved.

Description

Processing method and device for predicting ASPECTS score based on CT image

[0001] The present application claims priority to the Chinese patent application No. 202410871817.X, filed on July 1, 2024, and entitled "Processing method and device for predicting ASPECTS score based on CT image". TECHNICAL FIELD

[0002] The present application relates to the technical field of data processing, and particularly relates to a processing method and device for predicting ASPECTS score based on CT image. BACKGROUND

[0003] Cerebral ischemic stroke (CIS), also known as cerebral infarction (CI), is a kind of cerebrovascular event. Two main signs of cerebral ischemic stroke are: 1) high-density arterial sign of brain blood vessels: thrombus formed in the arterial blood vessels slows down or even stops the blood flow and causes ischemic injury to the blood vessels, and the thrombus in the blood vessels forms a high-density arterial sign on the non-contrast CT (NCCT) image; 2) cytotoxic edema of brain tissue: after ischemic injury, serum proteins and water in the blood begin to penetrate into the nearby brain tissue, causing cytotoxic edema of the brain tissue, and the cytotoxic edema forms a local low-density image area on the non-contrast CT image.

[0004] Alberta Stroke Program Early CT Score (ASPECTS) is a method of evaluating early ischemic change (EIC) of ischemic stroke based on plain CT images. The ASPECTS scoring method divides the brain into 7 pairs of left / right brain regions (left / right brain C region, left / right brain L region, left / right brain IC region, left / right brain I region, left / right brain M1 region, left / right brain M2 region, left / right brain M3 region) on the axial plain CT images of the basal ganglia nuclear layer of the brain, and divides the brain into 3 pairs of left / right brain regions (left / right brain M4 region, left / right brain M5 region, left / right brain M6 region) on the axial plain CT images above the basal ganglia nuclear layer of the brain. Based on the 10 pairs of left / right brain regions (also known as ASPECTS regions), the score (also known as ASPECTS score) is obtained, each region is scored as 0 or 1, 0 indicates that the current region has a low probability of ischemic stroke, and 1 indicates that the current region has a high probability of ischemic stroke. The scores of the 10 pairs of left / right brain regions are located on the main affected side, that is, the left or right side of the brain where ischemic stroke occurs. Based on the location of the affected side (such as left, right, or bilateral), the scores of the 10 pairs of left / right brain regions are screened and summed to obtain the corresponding total score. Finally, the ASPECTS score total score (usually fixed at 10) is subtracted from the total score to obtain the final ASPECTS overall score. The higher the score, the lower the probability of ischemic stroke, and the lower the score, the higher the probability of ischemic stroke.

[0005] Currently, the processing mechanism for ASPECTS scoring of plain CT images based on ASPECTS scoring rules is mostly achieved by manual scoring, that is, doctors rely on their own experience and ASPECTS scoring rules to score the plain CT images of the subjects. This manual scoring method is affected by human factors, and it is difficult to achieve high scoring efficiency and stable scoring quality.

[0006] To solve this problem, some researchers have proposed an intelligent scoring method based on an image detection model + a genomic feature extraction tool + a regression calculation model. The intelligent scoring method first uses the image detection model to detect the ASPECTS partition of the plain CT image to obtain 10 pairs of left and right brain partition images; then uses the genomic feature extraction tool to identify the left / right brain partition image genomic features of each pair of left and right brain partition images; then uses the regression calculation model to compare the features of each pair of left / right brain partition image genomic features (usually using a difference method) to obtain the corresponding comparison features (or difference features), and performs regression calculation based on the left / right brain partition image genomic features and comparison features of the 10 pairs of left / right brain partition to obtain the corresponding partition scores, and performs main affected side positioning based on the scores of the 10 pairs of left / right brain partition, and performs screening and summing of the scores of the 10 pairs of left / right brain partition based on the located affected side to obtain the corresponding sum total score, and finally subtracts the sum total score of the 10 pairs of left / right brain partition from the ASPECTS score total score to obtain the final ASPECTS overall score.

[0007] We found in practical application that although the above intelligent scoring method can improve scoring efficiency and output stable scoring quality, the prediction accuracy of this type of scoring method is not ideal. The main reasons are as follows: 1) the rough partition detection boundary output by the image detection model results in insufficient boundary accuracy of the partition image, thereby reducing the image feature recognition accuracy, reducing the affected side positioning accuracy, and reducing the overall prediction accuracy; 2) there are many interference factors in the partition image, in addition to noise, common interference factors include genomic feature interference caused by asymmetric sulcus structure, genomic feature interference caused by old infarction area of brain artery, and the presence of these interference factors also reduces the image feature recognition accuracy, reduces the affected side positioning accuracy, and reduces the overall prediction accuracy. SUMMARY

[0008] The present application aims at the defects of the prior art, and provides a processing method and device for predicting ASPECTS score based on CT images, electronic equipment and computer readable storage medium. The present application constructs four image segmentation models (a first segmentation model for sulci region segmentation, a second segmentation model for cerebral arterial old infarction region segmentation, a third segmentation model for cerebral vascular high density arterial sign region segmentation, and a fourth segmentation model for left and right brain ASPECTS partition and ventricle region segmentation), a first classification model for classifying brain tissue cytotoxic edema brain side information, and a first scoring model for ASPECTS scoring of ten sets of left and right brain partition CT images. After training of all models is completed, an axial position brain plain CT image located at a basal ganglia nuclear group level of the brain and an axial position brain plain CT image located above the basal ganglia nuclear group form a CT image sequence. The first / second / third / fourth segmentation model respectively performs image segmentation on the CT image sequence to obtain corresponding first / second / third / fourth semantic graph sets, and the first classification model classifies and identifies the CT image sequence to obtain a corresponding predicted side group. Based on the third / fourth semantic graph set, the predicted side group and the CT image sequence, side information comprehensive evaluation is performed to obtain a corresponding evaluation side. Left and right brain ASPECTS partition images are extracted from the CT image sequence to obtain a corresponding partition image set. Each partition image in the partition image set is filtered based on a median filtering method, and after the filtering is completed, the first / second semantic graph set is used to eliminate interference regions (sulci region and cerebral arterial old infarction region) of each partition image. The first scoring model performs ASPECTS scoring on the partition image sequence to obtain a corresponding partition score pair sequence. Finally, based on the evaluation side and the partition score pair sequence, overall score summary and affected side summary are performed to obtain a corresponding summary report. As can be seen, the fourth segmentation model is used to improve the boundary accuracy of the partition image, the median filtering is used to reduce the noise of the partition image, the first / second segmentation model is used to eliminate two types of interference factor images (sulci region image and cerebral arterial old infarction region image) in the partition image, the third segmentation model + the first classification model and a side information comprehensive evaluation mechanism are used to improve the positioning accuracy of the affected side, and the batch processing mechanism of the first scoring model is used to improve the partition scoring efficiency. Through the present application, the overall prediction efficiency can be further improved, the partition image boundary accuracy can be improved, the noise and interference of the partition image can be reduced, the positioning accuracy of the affected side can be improved, and the prediction accuracy can be improved.

[0009] To achieve the above object, an embodiment of the present application provides a processing method for predicting ASPECTS score based on CT images, which comprises the following steps.

[0010] constructing a first segmentation model for semantic segmentation of sulci regions in brain CT images, denoted as a corresponding first segmentation model; constructing a second segmentation model for semantic segmentation of brain arterial old infarction regions in brain CT images, denoted as a corresponding second segmentation model; constructing a third segmentation model for semantic segmentation of brain vascular high-density arterial sign regions in brain CT images, denoted as a corresponding third segmentation model; constructing a fourth segmentation model for semantic segmentation of left and right brain ASPECTS partitions and brain ventricle regions in brain CT image sequences, denoted as a corresponding fourth segmentation model; constructing a first classification model for classifying and identifying brain side information of brain tissue cytotoxic edema in brain CT images; constructing a first scoring model for ASPECTS scoring of ten sets of left and right brain CT image pairs in ASPECTS partitions; and performing collective model training on the first, second, third and fourth segmentation models, the first classification model and the first scoring model based on a pre-set plain CT image library;

[0011] After the collective model training is completed, an axial brain plain CT image located at a brain basal ganglia nuclear layer and an axial brain plain CT image located above the brain basal ganglia nuclear layer are received, denoted as a first CT image and a second CT image; and a first CT image sequence is formed by the first and second CT images;

[0012] The first, second, third and fourth segmentation models are respectively used to perform image segmentation processing on the first CT image sequence to obtain a first semantic graph group, a second semantic graph group, a third semantic graph group and a fourth semantic graph group; and the first classification model is used to perform classification and identification on the first CT image sequence to obtain a first predicted side group; the first predicted side group includes a first image predicted side and a second image predicted side; the first and second image predicted sides both include empty, left side, right side and bilateral sides;

[0013] Based on the third semantic graph group, the fourth semantic graph group, the first predicted side group and the first CT image sequence, side information comprehensive evaluation is performed to obtain a first evaluation side; the first evaluation side includes empty, left side, right side and bilateral sides;

[0014] performing left and right brain ASPECTS partition image extraction processing on the first CT image sequence to obtain a corresponding first partition image set; performing filtering on each partition image in the first partition image set based on a median filtering method; and after the filtering ends, performing sulcal region and cerebral arterial old infarction region elimination processing on each partition image in the first partition image set based on the first and second semantic graph groups to obtain a corresponding second partition image set; and performing ASPECTS partition left and right brain CT image pair extraction and sorting based on the second partition image set to obtain a corresponding first partition image pair sequence;

[0015] performing ASPECTS scoring on the first partition image pair sequence by the first scoring model to obtain a corresponding first partition score pair sequence;

[0016] performing overall score aggregation and affected side aggregation based on the first evaluation side and the first partition score pair sequence to obtain a corresponding first aggregation report; the first aggregation report includes a first total score and a first affected side type; the first total score is an integer with a value between 0 and 10; and the first affected side type includes left side, right side, and bilateral.

[0017] Preferably, the cerebral vascular hyperdensity arterial sign and the brain tissue cytotoxic edema are each a typical sign of a type of ischemic stroke.

[0018] The ASPECTS partition is composed of ten types of brain tissue partitions, namely C partition, L partition, IC partition, I partition, M1 partition, M2 partition, M3 partition, M4 partition, M5 partition, and M6 partition; the C partition is the caudate nucleus region of the brain; the L partition is the lenticular nucleus region of the brain; the IC partition is the internal capsule region of the brain; the I partition is the insular cortex region of the brain; the M1 partition is the anterior cortical region of the middle cerebral artery; the M2 partition is the lateral cortical region of the middle cerebral artery outside the insula; the M3 partition is the posterior cortical region of the middle cerebral artery; the M4 partition is the middle cerebral artery cortical region above the M1 partition; the M5 partition is the middle cerebral artery cortical region above the M2 partition; the M6 partition is the middle cerebral artery cortical region above the M3 partition; and each type of brain tissue partition of the ASPECTS partition is symmetrical between the left and right brains.

[0019] Preferably, the first segmentation model is based on a U-Net image segmentation network; the first segmentation model is used for sulci region segmentation processing of a current CT image input by the model to obtain a corresponding first semantic map; the current CT image is an axial brain plain CT image located at a basal ganglia nuclear layer or an axial brain plain CT image located above the basal ganglia nuclear layer; the first semantic map includes a plurality of first pixel points; each first pixel point corresponds to a first semantic type; the first semantic type includes sulci region semantics and background semantics;

[0020] The second segmentation model is based on a U-Net image segmentation network; the second segmentation model is used for sulci region segmentation processing of a current CT image input by the model to obtain a corresponding first semantic map; the current CT image is an axial brain plain CT image located at a basal ganglia nuclear layer or an axial brain plain CT image located above the basal ganglia nuclear layer; the first semantic map includes a plurality of first pixel points; each first pixel point corresponds to a first semantic type; the first semantic type includes sulci region semantics and background semantics;

[0021] The third segmentation model is based on a U-Net image segmentation network; the third segmentation model is used for sulci region segmentation processing of a current CT image input by the model to obtain a corresponding first semantic map; the current CT image is an axial brain plain CT image located at a basal ganglia nuclear layer or an axial brain plain CT image located above the basal ganglia nuclear layer; the first semantic map includes a plurality of first pixel points; each first pixel point corresponds to a first semantic type; the first semantic type includes sulci region semantics and background semantics;

[0022] The fourth segmentation model is implemented based on two U-Net image segmentation networks, denoted as a corresponding first U-Net network and a second U-Net network; the fourth segmentation model is used for left and right brain ASPECTS partition and ventricle region segmentation processing of a current CT image sequence input by the model to obtain a corresponding fourth semantic image sequence, specifically: the fourth segmentation model is used for taking a first CT image in the current CT image sequence input by the model as a corresponding first network input image, and taking a second CT image in the current CT image sequence as a corresponding second network input image; and the first U-Net network is used for fourteen-class left and right brain ASPECTS partition and ventricle region segmentation processing of the first network input image to obtain a corresponding fourth one semantic image; and the second U-Net network is used for six-class left and right brain ASPECTS partition and ventricle region segmentation processing of the second network input image to obtain a corresponding fourth two semantic image; and the obtained fourth one and fourth two semantic images are sequentially sorted to form the corresponding fourth semantic image sequence output;

[0023] The current CT image sequence is sorted by two CT images, wherein the first CT image is an axial brain plain CT image located at the basal ganglia nuclear layer of the brain, and the second CT image is an axial brain plain CT image located above the basal ganglia nuclear layer of the brain;

[0024] The fourth one semantic image includes a plurality of fourth one pixel points; each fourth one pixel point corresponds to a fourth one semantic type; the fourth one semantic type includes fourteen-class left and right brain ASPECTS partition semantics, ventricle region semantics, and background semantics; the fourteen-class left and right brain ASPECTS partition semantics include left / right brain C partition semantics, left / right brain L partition semantics, left / right brain IC partition semantics, left / right brain I partition semantics, left / right brain M1 partition semantics, left / right brain M2 partition semantics, and left / right brain M3 partition semantics;

[0025] The fourth two semantic image includes a plurality of fourth two pixel points; each fourth two pixel point corresponds to a fourth two semantic type; the fourth two semantic type includes six-class left and right brain ASPECTS partition semantics, ventricle region semantics, and background semantics; the six-class left and right brain ASPECTS partition semantics include left / right brain M4 region semantics, left / right brain M5 region semantics, and left / right brain M6 region semantics;

[0026] The first classification model comprises a first feature extraction module and a first classification module; the first feature extraction module is realized based on a ResNet network; the first classification module is realized based on an MLP network; the first classification model is used for classifying and identifying brain side information of brain tissue cytotoxic edema on a current CT image input by the model to obtain a corresponding predicted side, specifically: the first classification model is used for inputting the current CT image input by the model into the first feature extraction module for feature extraction processing to obtain a corresponding feature tensor; and inputting the feature tensor into the first classification module for classification prediction processing to obtain a corresponding predicted type; the predicted type includes empty, left side, right side and bilateral;

[0027] The first scoring model is composed of ten parallel partition scoring models and a partition score merging module; each partition scoring model is composed of a left and right brain image sorting module, a left brain feature extraction module, a right brain feature extraction module, a left and right brain feature difference module, a left brain feature fusion module, a right brain feature fusion module, a left brain score prediction module, a right brain score prediction module and a left and right brain score merging module; the left and right brain score prediction modules are realized based on a binary classification prediction model;

[0028] The first scoring model is used for ASPECTS scoring of a partition image pair sequence input by the model to obtain a corresponding partition score pair sequence, specifically: the first scoring model is used for inputting each partition image pair in the partition image pair sequence input by the model into the corresponding partition scoring model, sending the corresponding partition score pair obtained by each partition scoring model according to the corresponding partition image pair to the partition score merging module by left and right brain ASPECTS scoring, and sequentially sorting all the partition score pairs obtained by the partition score merging module to form the corresponding partition score pair sequence;

[0029] The corresponding partition score pair obtained by each partition score model according to the corresponding partition image pair is sent to the partition score merging module, specifically: each partition score model inputs the corresponding partition image pair into the left and right brain image sorting module, extracts the corresponding left brain partition image and right brain partition image from the partition image pair by the left and right brain image sorting module, and sends the corresponding left brain partition image and right brain partition image to the left brain feature extraction module and the right brain feature extraction module; and the left brain feature extraction module extracts radiomics features from the left brain partition image to obtain corresponding left brain radiomics features, which are sent to the left and right brain feature difference module and the left brain feature fusion module; and the right brain feature extraction module extracts radiomics features from the right brain partition image to obtain corresponding right brain radiomics features, which are sent to the left and right brain feature difference module and the right brain feature fusion module; and the left and right brain feature difference module subtracts the difference features of the left brain radiomics features from the right brain radiomics features, and the difference features of the right brain radiomics features from the left brain radiomics features, and records the difference features as corresponding left-right brain radiomics difference features and right-left brain radiomics difference features, and sends the left-right brain radiomics difference features to the corresponding left brain feature fusion module and the right-left brain radiomics difference features to the right brain feature fusion module; and the left brain feature fusion module performs feature splicing processing on the left brain radiomics features and the left-right brain radiomics difference features to obtain corresponding left brain splicing features, which are sent to the left brain score prediction module; and the right brain feature fusion module performs feature splicing processing on the right brain radiomics features and the right-left brain radiomics difference features to obtain corresponding right brain splicing features, which are sent to the right brain score prediction module; and the left brain score prediction module performs binary ASPECTS score prediction processing on the left brain splicing features to obtain corresponding left brain partition score values, which are sent to the left and right brain score merging module; and the right brain score prediction module performs binary ASPECTS score prediction processing on the right brain splicing features to obtain corresponding right brain partition score values, which are sent to the left and right brain score merging module; and the left and right brain score merging module combines the left brain partition score values and the right brain partition score values to obtain the corresponding partition score pair, which is sent to the partition score merging module;

[0030] The partition image pair sequence is sequentially sorted by ten partition image pairs; each partition image pair corresponds to a type of partition in the ASPECTS partition; the partition image pair corresponds to the partition score model in the first score model; each partition image is composed of the corresponding left brain partition image and right brain partition image;

[0031] The partition score pairs are sequentially ordered by ten of the partition score pairs; the partition score pairs correspond to the partition image pairs one by one; each of the partition score pairs is composed of the corresponding left brain partition score and the right brain partition score; the left and right brain partition scores are valued as 0 or 1.

[0032] Preferably, the plain CT image library comprises a plurality of first plain CT image groups; each of the first plain CT image groups corresponds to two plain CT acquisition images of an image acquisition object; the first plain CT image group comprises a first acquisition object type, a first CT acquisition image and a second CT acquisition image; the first acquisition object type is used to mark the feature type of the corresponding image acquisition object, at least including a healthy type without brain disease, an old infarction history type with brain arterial vessel old infarction history but without ischemic stroke signs, an A type ischemic stroke type without brain arterial vessel old infarction history but with the ischemic stroke signs, and a B type ischemic stroke type with brain arterial vessel old infarction history and with the ischemic stroke signs; the ischemic stroke signs include one or all of the brain vessel high-density arterial sign and the brain tissue cytotoxic edema; the first CT acquisition image is an axial brain plain CT image of the corresponding image acquisition object located at the level of basal ganglia nuclear mass; the second CT acquisition image is an axial brain plain CT image of the corresponding image acquisition object located above the level of basal ganglia nuclear mass; the first and second CT acquisition images are acquired in the following manner: once axial brain CT scanning is performed on the image acquisition object to obtain a corresponding current plain CT image sequence, and a CT image located at the level of basal ganglia nuclear mass is extracted from the current plain CT image sequence as the corresponding first CT acquisition image, and a CT image located above the level of basal ganglia nuclear mass is extracted from the current plain CT image sequence as the corresponding second CT acquisition image;

[0033] The first semantic graph group comprises a first image semantic graph A 11 and a second image semantic graph A 12 ; the first image semantic graph A 11 comprises a plurality of first semantic image pixels a 11 , each of the first semantic image pixels a 11 corresponds to one of the first semantic types; the second image semantic graph A 12 comprises a plurality of second semantic image pixels a 12 , each of the second semantic image pixels a 12 corresponds to one of the first semantic types;

[0034] The second semantic graph group comprises a first image semantic graph A 21and a second image semantic graph A 22 ; the first image semantic graph A 21 comprises a plurality of first semantic graph pixels a 21 , each of the first semantic graph pixels a 21 corresponds to one of the second semantic types; the second image semantic graph A 22 comprises a plurality of second semantic graph pixels a 22 , each of the second semantic graph pixels a 22 corresponds to one of the second semantic types;

[0035] The third semantic graph set comprises a first image semantic graph A 31 and a second image semantic graph A 32 ; the first image semantic graph A 31 comprises a plurality of first semantic graph pixels a 31 , each of the first semantic graph pixels a 31 corresponds to one of the third semantic types; the second image semantic graph A 32 comprises a plurality of second semantic graph pixels a 32 , each of the second semantic graph pixels a 32 corresponds to one of the third semantic types;

[0036] The fourth semantic graph set comprises a first image semantic graph A 41 and a second image semantic graph A 42 ; the first image semantic graph A 41 comprises a plurality of first semantic graph pixels a 41 , each of the first semantic graph pixels a 41 corresponds to one of the fourth semantic types; the second image semantic graph A 42 comprises a plurality of second semantic graph pixels a 42 , each of the second semantic graph pixels a 42 corresponds to one of the fourth semantic types;

[0037] The first partition image set comprises twenty first partition images, the first to the twentieth first partition images are respectively corresponding left / right brain C area images, left / right brain L area images, left / right brain IC area images, left / right brain I area images, left / right brain M1 area images, left / right brain M2 area images, left / right brain M3 area images, left / right brain M4 area images, left / right brain M5 area images and left / right brain M6 area images;

[0038] The second partition image set includes twenty second partition images, and the first to the twentieth second partition images are corresponding left / right brain C region images, left / right brain L region images, left / right brain IC region images, left / right brain I region images, left / right brain M1 region images, left / right brain M2 region images, left / right brain M3 region images, left / right brain M4 region images, left / right brain M5 region images and left / right brain M6 region images respectively.

[0039] The first partition image pair sequence is sequentially ordered by ten first partition image pairs, and the first to the tenth first partition image pairs are corresponding left / right brain C region image pairs, left / right brain L region image pairs, left / right brain IC region image pairs, left / right brain I region image pairs, left / right brain M1 region image pairs, left / right brain M2 region image pairs, left / right brain M3 region image pairs, left / right brain M4 region image pairs, left / right brain M5 region image pairs and left / right brain M6 region image pairs respectively.

[0040] The first partition score pair sequence is sequentially ordered by ten first partition score pairs, and the first to the tenth first partition score pairs are corresponding left / right brain C region score pairs, left / right brain L region score pairs, left / right brain IC region score pairs, left / right brain I region score pairs, left / right brain M1 region score pairs, left / right brain M2 region score pairs, left / right brain M3 region score pairs, left / right brain M4 region score pairs, left / right brain M5 region score pairs and left / right brain M6 region score pairs respectively.

[0041] Preferably, the side information comprehensive evaluation based on the third semantic graph group, the fourth semantic graph group, the first prediction side group and the first CT image sequence obtains a corresponding first evaluation side, and specifically includes:

[0042] In step 501, the first image semantic graph A of the third semantic graph group is obtained by performing semantic segmentation on the first CT image sequence. 31 In the first image semantic graph A, each first semantic pixel point a corresponding to a high-density arterial sign region semantic is obtained by performing semantic segmentation on the first CT image sequence. 31the first CT image sequence, and a plurality of first type point sets are obtained by clustering all the first type points based on a preset point clustering algorithm; and a corresponding first marked region of a high-density arterial sign area is marked on the first CT image of the first CT image sequence based on each of the first type point sets;

[0043] The point clustering algorithm at least includes a K-means clustering algorithm, a DBSCAN clustering algorithm, and an OPTICS clustering algorithm.

[0044] Step 502, and the second image semantic graph A 32 of the third semantic graph group is marked as a corresponding second type point; and a plurality of second type point sets are obtained by clustering all the second type points based on the point clustering algorithm; and a corresponding second marked region of a high-density arterial sign area is marked on the second CT image of the first CT image sequence based on each of the second type point sets. 32 of the third semantic graph group is marked as a corresponding second type point; and a plurality of second type point sets are obtained by clustering all the second type points based on the point clustering algorithm; and a corresponding second marked region of a high-density arterial sign area is marked on the second CT image of the first CT image sequence based on each of the second type point sets.

[0045] Step 503, and the first image semantic graph A 41 of the fourth semantic graph group is marked as a corresponding third type point; and fourteen third type point sets are obtained by clustering all the first semantic graph pixels a 41 corresponding to the fourteen types of left and right brain ASPECTS partition semantics into a corresponding third type point set; and fourteen first marked partitions are marked on the first CT image based on the fourteen third type point sets; and a corresponding first left brain partition set is formed by seven left brain partitions in the fourteen first marked partitions, and a corresponding first right brain partition set is formed by seven right brain partitions.

[0046] Step 504, and the second image semantic graph A 42 of the fourth semantic graph group is marked as a corresponding fourth type point; and six fourth type point sets are obtained by clustering all the second semantic graph pixels a 42 corresponding to the six types of left and right brain ASPECTS partition semantics into a corresponding fourth type point set; and six second marked partitions are marked on the second CT image based on the six fourth type point sets; and a corresponding second left brain partition set is formed by three left brain partitions in the six second marked partitions, and a corresponding second right brain partition set is formed by three right brain partitions; and a corresponding third left brain partition set is obtained by merging the first and second left brain partition sets, and a corresponding third right brain partition set is obtained by merging the first and second right brain partition sets.

[0047] Step 505, and set a corresponding first side evaluation result based on the area intersection relationship between the third left brain partition set and the third right brain partition set and all the first and second marker regions; and set the third average precision obtained by the third segmentation model in the last model training as a first confidence corresponding to the first side evaluation result;

[0048] The first side evaluation result includes null, left side, right side, and bilateral.

[0049] Step 506, identify the first image prediction side and the second image prediction side of the first prediction side group; if both the first and second image prediction sides are null, set a corresponding second side evaluation result as null; if one of the first and second image prediction sides is left side and the other is left side or null, set a corresponding second side evaluation result as left side; if one of the first and second image prediction sides is right side and the other is right side or null, set a corresponding second side evaluation result as right side; if one of the first and second image prediction sides is left side and the other is right side, set a corresponding second side evaluation result as bilateral; and set the fifth precision obtained by the first classification model in the last model training as a second confidence corresponding to the second side evaluation result.

[0050] The second side evaluation result includes null, left side, right side, and bilateral.

[0051] Step 507, identify the switch state of a preset manual side evaluation switch; if the switch state of the manual side evaluation switch is an open state, set a corresponding third side evaluation result as null and set a third confidence corresponding to the third side evaluation result as 0; if the switch state of the manual side evaluation switch is a closed state, send the first CT image sequence to a preset first manual side evaluation interface and receive the third side evaluation result and the corresponding third confidence sent back by the first manual side evaluation interface.

[0052] The switch state of the manual side evaluation switch includes an open state and a closed state.

[0053] Step 508, identify whether the first, second, and third side evaluation results are all the same; if so, take any one of the first, second, and third side evaluation results as a corresponding current evaluation side and go to step 516; if not, go to step 509.

[0054] Step 509, identify whether the third confidence is 0; if the third confidence is 0, go to step 510; if the third confidence is not 0, go to step 511;

[0055] Step 510, identify whether there is one of the first and second side evaluation results being empty; if yes, take the other side evaluation result which is not empty as the corresponding current evaluation side and go to step 516; if no, take the side evaluation result with the maximum confidence in the first and second side evaluation results as the corresponding current evaluation side and go to step 516;

[0056] Step 511, identify the preset first evaluation mode; if the first evaluation mode is the first mode, go to step 512; if the first evaluation mode is the second mode, go to step 513; if the first evaluation mode is the third mode, go to step 514; if the first evaluation mode is the fourth mode, go to step 515;

[0057] The first evaluation mode includes the first mode, the second mode, the third mode and the fourth mode.

[0058] Step 512, take the side evaluation result with the maximum confidence in the first, second and third side evaluation results as the corresponding current evaluation side and go to step 516;

[0059] Step 513, set four initialized empty vote box sets for four side type categories of empty, left side, right side and double side, and record the four vote box sets as the corresponding first, second, third and fourth vote box sets; when the first, second or third side evaluation result matches one of the four side type categories, add the current side evaluation result to the first, second, third or fourth vote box set corresponding to the current side type category; take the total number of the added side evaluation results in the finally obtained first, second, third and fourth vote box sets as the corresponding first, second, third and fourth vote box votes; take the maximum value in the first, second, third and fourth vote box votes as the corresponding maximum vote; identify whether the number of the vote box set corresponding to the maximum vote is unique; if the number of the vote box set corresponding to the maximum vote is unique, take the side type category corresponding to the maximum vote among the four side type categories of empty, left side, right side and double side as the corresponding current evaluation side and go to step 516; if the number of the vote box set corresponding to the maximum vote is not unique, take the side evaluation result with the maximum confidence in all the side evaluation results in all the vote box sets corresponding to the maximum vote as the corresponding current evaluation side and go to step 516;

[0060] Step 514, subtracting the difference between the first, second or third confidence and a preset baseline confidence threshold as the corresponding first, second or third differential confidence; setting four differential confidence sets initialized as empty for the four side type categories of empty, left, right and bilateral, denoted as the corresponding first, second, third and fourth differential confidence sets; adding the first, second or third differential confidence corresponding to the current side type evaluation result to the first, second, third or fourth differential confidence set corresponding to the current side type category when the first, second or third side type evaluation result matches one of the four side type categories; summing all differential confidences in the final first, second, third or fourth differential confidence set to obtain the corresponding first, second, third or fourth set sum; taking the maximum value of the first, second, third and fourth set sums as the corresponding maximum sum; identifying whether the number of differential confidence sets corresponding to the maximum sum is unique; if the number of differential confidence sets corresponding to the maximum sum is unique, taking the side type category corresponding to the maximum sum among the four side type categories of empty, left, right and bilateral as the corresponding current evaluation side type and proceeding to step 516; if the number of differential confidence sets corresponding to the maximum sum is not unique, taking the side type evaluation result with the maximum confidence among all side type evaluation results in all differential confidence sets corresponding to the maximum sum as the corresponding current evaluation side type and proceeding to step 516;

[0061] Step 515, subtract the difference between the first, second or third confidence and the baseline confidence threshold from the corresponding fourth, fifth or sixth difference confidence; and set four difference confidence sets initialized as empty for the four side type categories of empty, left, right and bilateral, denoted as the fifth, sixth, seventh and eighth difference confidence sets; and when the first, second or third side evaluation result matches one of the four side type categories, add the fourth, fifth or sixth difference confidence corresponding to the current side evaluation result to the fifth, sixth, seventh or eighth difference confidence set corresponding to the current side type category; and calculate the mean value of all difference confidences in the final fifth, sixth, seventh or eighth difference confidence set to obtain the corresponding first, second, third or fourth set mean value; and take the maximum value of the first, second, third and fourth set mean values as the corresponding maximum mean value; and identify whether the number of difference confidence sets corresponding to the maximum mean value is unique; if the number of difference confidence sets corresponding to the maximum mean value is unique, take the side type category corresponding to the maximum mean value among the four side type categories of empty, left, right and bilateral as the corresponding current evaluation side and go to step 516; if the number of difference confidence sets corresponding to the maximum mean value is not unique, take the side evaluation result with the maximum confidence among all side evaluation results in all difference confidence sets corresponding to the maximum mean value as the corresponding current evaluation side and go to step 516;

[0062] Step 516, output the final current evaluation side as the corresponding first evaluation side.

[0063] Further, the first side evaluation result is set based on the intersection relationship between the third left brain partition set and the third right brain partition set and all the first and second marker regions, and specifically includes:

[0064] If all left brain partitions in the third left brain partition set have no intersection with all the first and second marker regions, and all right brain partitions in the third right brain partition set have no intersection with all the first and second marker regions, the corresponding first side evaluation result is set as empty;

[0065] If at least one left brain partition in the third left brain partition set has an intersection with at least one of the first or second marker regions, and at least one right brain partition in the third right brain partition set has an intersection with at least one of the first or second marker regions, the corresponding first side evaluation result is set as bilateral;

[0066] If at least one left brain partition in the third left brain partition set intersects with at least one of the first or second marked areas, and all right brain partitions in the third right brain partition set do not intersect with all the first and second marked areas, the corresponding first side evaluation result is set to left side;

[0067] If at least one right brain partition in the third right brain partition set intersects with at least one of the first or second marked areas, and all left brain partitions in the third left brain partition set do not intersect with all the first and second marked areas, the corresponding first side evaluation result is set to right side.

[0068] Preferably, the first CT image sequence is subjected to left and right brain ASPECTS partition image extraction processing to obtain a corresponding first partition image set, specifically including:

[0069] The left brain C region image, left brain L region image, left brain IC region image, left brain I region image, left brain M1 region image, left brain M2 region image, left brain M3 region image, left brain M4 region image, left brain M5 region image and left brain M6 region image corresponding to the ten left brain partitions of the third left brain partition set on the first and second CT images of the first CT image sequence are extracted as the corresponding ten first partition images;

[0070] The right brain C region image, right brain L region image, right brain IC region image, right brain I region image, right brain M1 region image, right brain M2 region image, right brain M3 region image, right brain M4 region image, right brain M5 region image and right brain M6 region image corresponding to the ten right brain partitions of the third right brain partition set on the first and second CT images of the first CT image sequence are extracted as the corresponding ten first partition images;

[0071] The twenty first partition images obtained are used to form the corresponding first partition image set.

[0072] Preferably, after the filtering is completed, each partition image in the first partition image set is subjected to sulcal region and cerebral artery blood vessel old infarction region elimination processing based on the first and second semantic graph groups to obtain a corresponding second partition image set, specifically including:

[0073] Each first semantic graph pixel a 11 in the first image semantic graph A 11The points are denoted as the corresponding fifth-class points; and all the fifth-class points are clustered based on a preset point clustering algorithm to obtain multiple fifth-class point sets; and the corresponding brain sulcus regions are marked on the first CT image based on each fifth-class point set to obtain the corresponding third-marked regions; the point clustering algorithm includes at least K-means clustering algorithm, DBSCAN clustering algorithm, and OPTICS clustering algorithm;

[0074] The second image semantic map A of the first semantic map group 12 In the middle, each pixel a of the second semantic map corresponding to the semantics of the brain sulcus region 12 The points are denoted as the corresponding sixth-class points; and the sixth-class points are clustered based on the point clustering algorithm to obtain multiple sixth-class point sets; and the corresponding brain sulcus regions are marked on the second CT image based on each sixth-class point set to obtain the corresponding fourth-class marked regions;

[0075] The first image semantic map A of the second semantic map group 21 In the middle, each pixel a in the first semantic map corresponding to the semantics of the old infarct region 21 The points are denoted as the corresponding seventh-class points; and the point clustering algorithm is used to cluster all the seventh-class points to obtain multiple seventh-class point sets; and based on each seventh-class point set, the corresponding old infarct area is marked on the first CT image to obtain the corresponding fifth marked area;

[0076] The second image semantic map A of the second semantic map group 22 In the middle, each pixel a in the second semantic map corresponding to the semantics of the old infarct region 22 The corresponding eighth-class points are denoted as points; and based on the point clustering algorithm, all the eighth-class points are clustered to obtain multiple eighth-class point sets; and based on each eighth-class point set, the corresponding old infarct area is marked on the second CT image to obtain the corresponding sixth marked area;

[0077] Each of the first partition images in the first partition image set is sequentially taken as the corresponding current partition image; the intersection region of the current partition image with any of the third, fourth, fifth, or sixth marked regions is taken as the corresponding first intersection region; the total number of the first intersection regions is counted to obtain the corresponding first intersection total; the first intersection total is identified; if the first intersection total is 0, the current partition image is taken as a corresponding second partition image; if the first intersection total is greater than 0, the local area image covered by each of the first intersection regions in the current partition image is deleted, and the current partition image after deletion is taken as a corresponding second partition image.

[0078] And the corresponding second partition image set is composed of all the second partition images obtained.

[0079] Preferably, the first evaluation side and the first partition score are based on the overall score of the sequence and the affected side, and specifically include:

[0080] All left brain partition values of 1 in the first partition score pair sequence are aggregated into a corresponding first left brain partition value set; and all right brain partition values of 1 in the first partition score pair sequence are aggregated into a corresponding first right brain partition value set;

[0081] The first evaluation side is identified;

[0082] If the first evaluation side is empty, the sum of the values of the first left brain partition value set is calculated to obtain a corresponding first total score, and the sum of the values of the first right brain partition value set is calculated to obtain a corresponding second total score; and the first and second total scores are compared, if the first total score is greater than or equal to the second total score, the first total score is taken as the corresponding current total score N c , and the corresponding first affected side type is set to left side, if the first total score is less than the second total score, the second total score is taken as the corresponding current total score N c , and the corresponding first affected side type is set to right side;

[0083] If the first evaluation side is left side, the sum of the values of the first left brain partition value set is calculated and the calculation result is taken as the corresponding current total score N c , and the corresponding first affected side type is set to left side;

[0084] If the first evaluation side is right side, the sum of the values of the first right brain partition value set is calculated and the calculation result is taken as the corresponding current total score N c , and the corresponding first affected side type is set to right side;

[0085] If the first evaluation side is bilateral, the sum of the values of the first left brain partition value set is calculated to obtain a corresponding third total score, and the sum of the values of the first right brain partition value set is calculated to obtain a corresponding fourth total score; and the third and fourth total scores are compared, if the third total score is greater than or equal to the fourth total score, the third total score is taken as the corresponding current total score N c , if the third total score is less than the fourth total score, the second total score is taken as the corresponding current total score N c ; and the corresponding first affected side type is set to bilateral;

[0086] and based on a preset ASPECTS score total score N max and the current total score N c corresponding to the first total score, the first total score = N max -N c ; the ASPECTS score total score N max is 10 by default; the current total score N c and the first total score are both integers with a value between 0 and 10;

[0087] and the first summary report corresponding to the first total score and the first affected side type.

[0088] The second aspect of the embodiment of the application provides a device for implementing the processing method for predicting an ASPECTS score based on a CT image according to the first aspect, and the device comprises a model construction and training module, a data receiving module, an image segmentation and classification module, a side comprehensive evaluation module, a subregion image noise reduction and disturbance removal module, a subregion scoring module, and an overall summary module.

[0089] The model construction and training module is configured to construct an image semantic segmentation model for segmenting a cerebral fissure region of a brain CT image, denoted as a corresponding first segmentation model; construct an image semantic segmentation model for segmenting an old infarction region of a brain artery of a brain CT image, denoted as a corresponding second segmentation model; construct an image semantic segmentation model for segmenting a high-density arterial sign region of a brain artery of a brain CT image, denoted as a corresponding third segmentation model; construct an image semantic segmentation model for segmenting an ASPECTS subregion of a left and right brain and a brain ventricle region of a brain CT image sequence, denoted as a corresponding fourth segmentation model; construct a first classification model for classifying and identifying brain side information of brain tissue cytotoxic edema on a brain CT image; construct a first scoring model for scoring an ASPECTS subregion left and right brain CT image pair; and based on a preset plain CT image library, collectively train the first, second, third, and fourth segmentation models, the first classification model, and the first scoring model.

[0090] The data receiving module is configured to, after the collective model training is completed, receive an axial brain plain CT image located at a brain basal ganglia nuclear layer and an axial brain plain CT image located above the brain basal ganglia nuclear layer, denoted as a corresponding first CT image and a second CT image; and form a first CT image sequence corresponding to the first CT image and the second CT image.

[0091] The image segmentation and classification module is configured to perform image segmentation on the first CT image sequence by the first, second, third and fourth segmentation models to obtain a corresponding first semantic graph group, a second semantic graph group, a third semantic graph group and a fourth semantic graph group, and to perform classification and identification on the first CT image sequence by the first classification model to obtain a corresponding first prediction side group; the first prediction side group includes a first image prediction side and a second image prediction side; the first and second image prediction sides both include empty, left side, right side and bilateral sides;

[0092] The side comprehensive evaluation module is configured to perform comprehensive evaluation of side information based on the third semantic graph group, the fourth semantic graph group, the first prediction side group and the first CT image sequence to obtain a corresponding first evaluation side; the first evaluation side includes empty, left side, right side and bilateral sides;

[0093] The partition image noise reduction and decontamination module is configured to perform left and right brain ASPECTS partition image extraction processing on the first CT image sequence to obtain a corresponding first partition image set; and perform filtering on each partition image in the first partition image set based on a median filtering method; and after the filtering is completed, perform sulcal region and cerebral arterial blood vessel old infarction region elimination processing on each partition image in the first partition image set based on the first and second semantic graph groups to obtain a corresponding second partition image set; and perform ASPECTS partition left and right brain CT image pair extraction and sorting based on the second partition image set to obtain a corresponding first partition image pair sequence;

[0094] The partition scoring module is configured to perform ASPECTS scoring on the first partition image pair sequence by the first scoring model to obtain a corresponding first partition score pair sequence;

[0095] The overall summary module is configured to perform overall score summary and affected side summary based on the first evaluation side and the first partition score pair sequence to obtain a corresponding first summary report; the first summary report includes a first total score and a first affected side type; the first total score is an integer with a value between 0 and 10; and the first affected side type includes left side, right side and bilateral sides.

[0096] The third aspect of the embodiment of the application provides an electronic device, including a memory, a processor and a transceiver;

[0097] The processor is configured to be coupled with the memory, read and execute instructions in the memory to realize the method steps of the first aspect;

[0098] The transceiver is coupled with the processor, and the transceiver is controlled by the processor to perform message transmission and reception.

[0099] The fourth aspect of the embodiment of the present application provides a computer readable storage medium, which stores computer instructions, and when the computer instructions are executed by a computer, the computer executes the instructions of the method in the first aspect.

[0100] The embodiment of the present application provides a processing method and device for predicting ASPECTS score based on CT images, electronic equipment and computer readable storage medium. From the above content, it can be known that the embodiment of the present application constructs four image segmentation models (a first segmentation model for sulci region segmentation, a second segmentation model for cerebral arterial old infarction region segmentation, a third segmentation model for cerebral vascular high density arterial sign region segmentation, and a fourth segmentation model for left and right brain ASPECTS partition and ventricle region segmentation), a first classification model for classifying brain tissue cytotoxic edema brain side information, and a first scoring model for scoring ten groups of left and right brain partition CT images. After training of all models is completed, an axial brain plain CT image located at a basal ganglia nuclear group level of the brain and an axial brain plain CT image located above the basal ganglia nuclear group level of the brain form a CT image sequence. The first / second / third / fourth segmentation model respectively performs image segmentation on the CT image sequence to obtain corresponding first / second / third / fourth semantic graph groups, and the first classification model classifies and identifies the CT image sequence to obtain a corresponding prediction side group. Based on the third / fourth semantic graph group, the prediction side group and the CT image sequence, side information comprehensive evaluation is performed to obtain a corresponding evaluation side. Left and right brain ASPECTS partition images of the CT image sequence are extracted to obtain a corresponding partition image set. Each partition image in the partition image set is filtered based on a median filtering method, and after the filtering is completed, the first / second semantic graph group is used to eliminate interference regions (sulci region and cerebral arterial old infarction region) of each partition image. The first scoring model scores the partition image sequence to obtain a corresponding partition score pair sequence. Finally, based on the evaluation side and the partition score pair sequence, overall score summary and affected side summary are performed to obtain a corresponding summary report. As can be seen, the fourth segmentation model is used to improve the boundary accuracy of the partition image, the median filtering is used to reduce the noise of the partition image, the first / second segmentation model is used to eliminate two types of interference factor images (sulci region image and cerebral arterial old infarction region image) in the partition image, the third segmentation model+the first classification model and a side information comprehensive evaluation mechanism are used to improve the positioning accuracy of the affected side, and the batch processing mechanism of the first scoring model is used to improve the partition score efficiency. Through the embodiment of the present application, the overall prediction efficiency is further improved, the boundary accuracy of the partition image is improved, the noise and interference of the partition image are reduced, the positioning accuracy of the affected side is improved, and the prediction accuracy is improved. BRIEF DESCRIPTION OF DRAWINGS

[0101] Fig. 1 is a schematic diagram of a processing method for predicting ASPECTS score based on CT image according to an embodiment of the present application;

[0102] Fig. 2 is a module structure diagram of a first classification model according to an embodiment of the present application;

[0103] Fig. 3 is a module structure diagram of a first scoring model according to an embodiment of the present application;

[0104] Fig. 4 is a module structure diagram of a processing device for predicting ASPECTS score based on CT image according to an embodiment of the present application;

[0105] Fig. 5 is a structural schematic diagram of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0106] In order to make the objects, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work fall within the scope of protection of the present application.

[0107] An embodiment of the present application provides a processing method for predicting ASPECTS score based on CT image. As shown in Fig. 1, the method mainly includes the following steps:

[0108] Step 1, constructing an image semantic segmentation model for segmenting sulcal region of brain CT image, denoted as a corresponding first segmentation model; and constructing an image semantic segmentation model for segmenting brain arterial old infarction region of brain CT image, denoted as a corresponding second segmentation model; and constructing an image semantic segmentation model for segmenting brain vascular high-density arterial sign region of brain CT image, denoted as a corresponding third segmentation model; and constructing an image semantic segmentation model for segmenting left and right brain ASPECTS partition and ventricular region of brain CT image sequence, denoted as a corresponding fourth segmentation model; and constructing a first classification model for classifying and identifying brain side information of brain tissue cytotoxic edema on brain CT image; and constructing a first scoring model for ASPECTS scoring of ten groups of ASPECTS partition left and right brain CT image pairs; and collectively training the first, second, third and fourth segmentation models, the first classification model and the first scoring model based on a preset plain CT image library;

[0109] Specifically comprising: step 11, constructing an image semantic segmentation model for sulci region segmentation of brain CT images, denoted as a corresponding first segmentation model;

[0110] Here, the first segmentation model of the embodiment of the application is implemented based on a U-Net image segmentation network; the first segmentation model is used for sulci region segmentation processing of a current CT image input by the model to obtain a corresponding first semantic graph; wherein the current CT image here is an axial position brain plain CT image located at a basal ganglia nuclear team level or an axial position brain plain CT image located above the basal ganglia nuclear team level; the first semantic graph comprises a plurality of first pixel points; each first pixel point corresponds to a first semantic type; the first semantic type comprises sulci region semantics and background semantics;

[0111] Step 12, and constructing an image semantic segmentation model for old infarction region segmentation of brain arterial vessels of brain CT images, denoted as a corresponding second segmentation model;

[0112] Here, the second segmentation model of the embodiment of the application is implemented based on a U-Net image segmentation network; the second segmentation model is used for old infarction region segmentation processing of a current CT image input by the model to obtain a corresponding second semantic graph; wherein the current CT image here is an axial position brain plain CT image located at a basal ganglia nuclear team level or an axial position brain plain CT image located above the basal ganglia nuclear team level; the second semantic graph comprises a plurality of second pixel points; each second pixel point corresponds to a second semantic type; the second semantic type comprises old infarction region semantics and background semantics;

[0113] Step 13, and constructing an image semantic segmentation model for brain vascular high-density arterial sign region segmentation of brain CT images, denoted as a corresponding third segmentation model;

[0114] Here, it should be noted that the brain vascular high-density arterial sign mentioned here and the brain tissue cytotoxic edema mentioned below are each a typical sign of ischemic stroke;

[0115] The third segmentation model of the embodiment of the application is implemented based on a U-Net image segmentation network; the third segmentation model is used for brain vascular high-density arterial sign region segmentation processing of a current CT image input by the model to obtain a corresponding third semantic graph; wherein the current CT image here is an axial position brain plain CT image located at a basal ganglia nuclear team level or an axial position brain plain CT image located above the basal ganglia nuclear team level; the third semantic graph comprises a plurality of third pixel points; each third pixel point corresponds to a third semantic type; the third semantic type comprises high-density arterial sign region semantics and background semantics;

[0116] Step 14, and construct an image semantic segmentation model for left and right brain ASPECTS partition and ventricle region segmentation of the brain CT image sequence, denoted as a corresponding fourth segmentation model;

[0117] Here, it should be noted that the ASPECTS partition of the embodiment of the application is composed of ten types of brain tissue partitions, which are C partition, L partition, IC partition, I partition, M1 partition, M2 partition, M3 partition, M4 partition, M5 partition and M6 partition; the C partition is the caudate nucleus region of the brain; the L partition is the lenticular nucleus region of the brain; the IC partition is the internal capsule region of the brain; the I partition is the insular cortex region of the brain; the M1 partition is the anterior cortical region of the middle cerebral artery; the M2 partition is the lateral cortical region of the middle cerebral artery outside the insula; the M3 partition is the posterior cortical region of the middle cerebral artery; the M4 partition is the middle cerebral artery cortical region above the M1 partition; the M5 partition is the middle cerebral artery cortical region above the M2 partition; the M6 partition is the middle cerebral artery cortical region above the M3 partition; each type of brain tissue partition of the ASPECTS partition of the embodiment of the application is symmetrical between the left and right brain;

[0118] The fourth segmentation model of the embodiment of the application is implemented based on two U-Net image segmentation networks, denoted as a corresponding first U-Net network and a second U-Net network; the fourth segmentation model is used to perform left and right brain ASPECTS partition and ventricle region segmentation processing on the current CT image sequence input to the model to obtain a corresponding fourth semantic image sequence, specifically: the fourth segmentation model is used to take the first CT image in the current CT image sequence input to the model as a corresponding first network input image, and take the second CT image in the current CT image sequence as a corresponding second network input image; and the first U-Net network is used to perform fourteen types of left and right brain ASPECTS partition and ventricle region segmentation processing on the first network input image to obtain a corresponding fourth one semantic image; and the second U-Net network is used to perform six types of left and right brain ASPECTS partition and ventricle region segmentation processing on the second network input image to obtain a corresponding fourth two semantic image; and the obtained fourth one and fourth two semantic images are sequentially sorted to form a corresponding fourth semantic image sequence output;

[0119] Among them, the current CT image sequence here is composed of two CT images in order, wherein the first CT image is an axial brain plain CT image located at the basal ganglia nuclear layer, and the second CT image is an axial brain plain CT image located above the basal ganglia nuclear layer;

[0120] The fourth one semantic graph includes a plurality of fourth one pixel points; each fourth one pixel point corresponds to a fourth one semantic type; the fourth one semantic type includes fourteen kinds of left and right brain ASPECTS partition semantics, ventricle region semantics and background semantics; the fourteen kinds of left and right brain ASPECTS partition semantics include left / right brain C partition semantics, left / right brain L partition semantics, left / right brain IC partition semantics, left / right brain I partition semantics, left / right brain M1 partition semantics, left / right brain M2 partition semantics, left / right brain M3 partition semantics;

[0121] The fourth two semantic graph includes a plurality of fourth two pixel points; each fourth two pixel point corresponds to a fourth two semantic type; the fourth two semantic type includes six kinds of left and right brain ASPECTS partition semantics, ventricle region semantics and background semantics; the six kinds of left and right brain ASPECTS partition semantics include left / right brain M4 region semantics, left / right brain M5 region semantics, left / right brain M6 region semantics;

[0122] Step 15, and a first classification model for classifying and identifying the brain side information of the brain tissue cytotoxic edema on the CT image of the brain is constructed;

[0123] Here, as shown in FIG. 2, which is a module structure diagram of the first classification model provided by the first embodiment of the present application, the first classification model of the present application includes a first feature extraction module and a first classification module; the first feature extraction module is realized based on the ResNet network; the first classification module is realized based on the MLP network; the first classification model is used for classifying and identifying the brain side information of the brain tissue cytotoxic edema on the current CT image input by the model to obtain the corresponding predicted side, specifically: the first classification model is used for inputting the current CT image input by the model into the first feature extraction module to obtain the corresponding feature tensor; and inputting the feature tensor into the first classification module to obtain the corresponding prediction type; wherein the prediction type includes empty, left side, right side and bilateral;

[0124] Step 16, and a first scoring model for scoring ten groups of ASPECTS partition left and right brain CT image pairs is constructed;

[0125] Here, as shown in Figure 3, the first scoring model of the embodiment of the present application is composed of ten parallel partition scoring models and a partition score merging module; each partition scoring model is composed of a left-right brain image sorting module, a left brain feature extraction module, a right brain feature extraction module, a left-right brain feature difference module, a left brain feature fusion module, a right brain feature fusion module, a left brain score prediction module, a right brain score prediction module and a left-right brain score merging module; the left and right brain feature extraction modules are realized based on a fixed omics feature extraction program or a dynamically set omics feature extraction tool interface, such as a PyRadiomics software interface; the left and right brain score prediction modules are realized based on a binary classification prediction model for binary 0 / 1 classification prediction according to the input feature vector / tensor, and the left and right brain score prediction modules can also be realized based on a regression calculation model for binary 0 / 1 prediction data regression calculation according to the input feature vector / tensor;

[0126] The first scoring model of the embodiment of the present application is used to perform ASPECTS scoring on the input partition image pair sequence of the model to obtain a corresponding partition score pair sequence, specifically: the first scoring model is used to input each partition image pair in the input partition image pair sequence of the model into the corresponding partition scoring model, and perform left-right brain ASPECTS scoring on the corresponding partition image pair by each partition scoring model to obtain a corresponding partition score pair and send it to the partition score merging module; and the partition score merging module sequentially sorts all obtained partition score pairs to form a corresponding partition score pair sequence;

[0127] Wherein, the partition image pair sequence of the embodiment of the present application is sequentially sorted by ten partition image pairs; each partition image pair corresponds to a type of partition in the ASPECTS partition; the partition image pair corresponds to the partition scoring model in the first scoring model one by one; each partition image is composed of a corresponding left brain partition image and a right brain partition image;

[0128] The partition score pair sequence of the embodiment of the present application is sequentially sorted by ten partition score pairs; the partition score pair corresponds to the partition image pair one by one; each partition score pair is composed of a corresponding left brain partition score and a right brain partition score; the left and right brain partition scores are 0 or 1;

[0129] Each partition scoring model performs left-right brain ASPECTS scoring on the corresponding partition image pair to obtain a corresponding partition score pair and sends it to the partition score merging module, specifically including:

[0130] Step A1, each sub-region scoring model inputs the corresponding sub-region image pair into the left and right brain image sorting module, and the left and right brain image sorting module extracts the corresponding left brain sub-region image and right brain sub-region image from the sub-region image pair and sends them to the corresponding left brain feature extraction module and right brain feature extraction module;

[0131] Step A2, and the left brain feature extraction module extracts the imageomics features of the left brain sub-region image to obtain the corresponding left brain imageomics features, which are sent to the left and right brain feature difference module and the left brain feature fusion module;

[0132] Step A3, and the right brain feature extraction module extracts the imageomics features of the right brain sub-region image to obtain the corresponding right brain imageomics features, which are sent to the left and right brain feature difference module and the right brain feature fusion module;

[0133] Step A4, and the left and right brain feature difference module subtracts the difference features of the left brain imageomics features from the right brain imageomics features, and the difference features of the right brain imageomics features from the left brain imageomics features, and records the left-right brain imageomics difference features and the right-left brain imageomics difference features as the corresponding left-right brain imageomics difference features and the right-left brain imageomics difference features, and sends the left-right brain imageomics difference features to the corresponding left brain feature fusion module, and sends the right-left brain imageomics difference features to the right brain feature fusion module;

[0134] Step A5, and the left brain feature fusion module performs feature splicing processing on the left brain imageomics features and the left-right brain imageomics difference features to obtain the corresponding left brain splicing features, which are sent to the left brain score prediction module;

[0135] Step A6, and the right brain feature fusion module performs feature splicing processing on the right brain imageomics features and the right-left brain imageomics difference features to obtain the corresponding right brain splicing features, which are sent to the right brain score prediction module;

[0136] Step A7, and the left brain score prediction module performs binary ASPECTS score prediction processing according to the left brain splicing features to obtain the corresponding left brain sub-region score, which is sent to the left and right brain score merging module;

[0137] Step A8, and the right brain score prediction module performs binary ASPECTS score prediction processing according to the right brain splicing features to obtain the corresponding right brain sub-region score, which is sent to the left and right brain score merging module;

[0138] Step A9, and the left and right brain score merging module groups the left brain sub-region score and the right brain sub-region score into the corresponding sub-region score pair, which is sent to the sub-region score merging module;

[0139] Step 17, and based on the pre-set plain CT image library, the first, second, third and fourth segmentation models, the first classification model and the first scoring model are collectively trained;

[0140] Wherein, the plain CT image library of the embodiment of the present application comprises a plurality of first plain CT image groups; each first plain CT image group corresponds to two plain CT acquisition images of an image acquisition object; the first plain CT image group comprises a first acquisition object type, a first CT acquisition image and a second CT acquisition image; the first acquisition object type is used to mark the characteristic type of the corresponding image acquisition object, at least including a healthy type without brain disease, an old infarction history type with brain arterial vessel old infarction history but without ischemic stroke signs, an A type ischemic stroke type without brain arterial vessel old infarction history but with ischemic stroke signs, and a B type ischemic stroke type with brain arterial vessel old infarction history and ischemic stroke signs; here, the ischemic stroke signs include one or all of brain vessel high-density arterial sign and brain tissue cytotoxic edema; the first CT acquisition image is an axial brain plain CT image of the corresponding image acquisition object located at the level of basal ganglia nuclear mass; the second CT acquisition image is an axial brain plain CT image of the corresponding image acquisition object located above the level of basal ganglia nuclear mass; the acquisition mode of the first and second CT acquisition images is: once axial brain CT scanning of the image acquisition object to obtain a corresponding current plain CT image sequence, and extracting a CT image located at the level of basal ganglia nuclear mass from the current plain CT image sequence as the corresponding first CT acquisition image, and extracting a CT image located above the level of basal ganglia nuclear mass from the current plain CT image sequence as the corresponding second CT acquisition image;

[0141] The current step 17 specifically comprises:

[0142] Specifically comprises:

[0143] Step 171, taking each first CT acquisition image or second CT acquisition image in the plain CT image library as a corresponding first training image; and marking the sulcus region and the background region on each first training image by artificial marking or other tool marking mode, and creating a corresponding first label semantic map based on the marking result to obtain the corresponding first label semantic map; and each first training image and the corresponding first label semantic map form a corresponding first data record; and all the obtained first data records form a corresponding first data set;

[0144] Here, the first data set obtained by the embodiment of the present application comprises a plurality of first data records; the first data record comprises a first training image and a first label semantic map; the first label semantic map comprises a plurality of first label image pixels; each first label image pixel corresponds to a first semantic type;

[0145] Step 172, each first CT acquisition image or second CT acquisition image in the plain scan CT image library is taken as a corresponding second training image; and the brain arterial vessel old infarction area and the background area are marked on each second training image by manual marking or other tool marking manner, and a corresponding second label semantic map is created based on the marking result; each second training image and the corresponding second label semantic map form a corresponding second data record; all the second data records form a corresponding second data set;

[0146] Here, the second data set obtained by the embodiment of the application includes a plurality of second data records; the second data record includes a second training image and a second label semantic map; the second label semantic map includes a plurality of second label image pixels; each second label image pixel corresponds to a second semantic type;

[0147] Step 173, each first CT acquisition image or second CT acquisition image in the plain scan CT image library is taken as a corresponding third training image; and the brain vessel high density arterial sign area and the background area are marked on each third training image by manual marking or other tool marking manner, and a corresponding third label semantic map is created based on the marking result; each third training image and the corresponding third label semantic map form a corresponding third data record; all the third data records form a corresponding third data set;

[0148] Here, the third data set obtained by the embodiment of the application includes a plurality of third data records; the third data record includes a third training image and a third label semantic map; the third label semantic map includes a plurality of third label image pixels; each third label image pixel corresponds to a third semantic type;

[0149] Step 174, and the first CT acquisition image and the second CT acquisition image of each first plain CT image group in the plain CT image library are taken as a corresponding fourth first training image and a corresponding fourth second training image, and each first plain CT image group is sequentially sorted according to the corresponding fourth first training image and the corresponding fourth second training image to form a corresponding fourth training image sequence; and the corresponding fourth first training image is marked with the corresponding sixteen semantic regions of the corresponding fourth first semantic type by manual marking or other tool marking methods, and a corresponding fourth first label semantic map is created based on the marking result; and the corresponding fourth second training image is marked with the corresponding eight semantic regions of the corresponding fourth second semantic type by manual marking or other tool marking methods, and a corresponding fourth second label semantic map is created based on the marking result; and the corresponding fourth first label semantic map and the corresponding fourth second label semantic map of each fourth training image sequence are sequentially sorted to form a corresponding fourth label semantic map sequence; and each fourth training image sequence and the corresponding fourth label semantic map sequence form a corresponding fourth data record; and all the obtained fourth data records form a corresponding fourth data set;

[0150] Here, the fourth data set obtained by the embodiment of the application includes a plurality of fourth data records; the fourth data record includes a fourth training image sequence and a fourth label semantic map sequence; the fourth training image sequence is sequentially sorted by a fourth first training image and a fourth second training image; the fourth label semantic map sequence is sequentially sorted by a fourth first label semantic map and a fourth second label semantic map; the fourth first label semantic map includes a plurality of fourth first label image pixels; each fourth first label image pixel corresponds to a fourth first semantic type; the fourth second label semantic map includes a plurality of fourth second label image pixels; each fourth second label image pixel corresponds to a fourth second semantic type;

[0151] Step 175, and each first CT acquisition image or second CT acquisition image in the plain scan CT image library is set as a corresponding fifth training image; whether brain tissue cytotoxic edema appears on each fifth training image is identified through artificial identification or other tool identification; if brain tissue cytotoxic edema does not appear on the current fifth training image, a corresponding first label prediction side is set as empty; if brain tissue cytotoxic edema appears on the current fifth training image, left and right brain regions where brain tissue cytotoxic edema appears are further identified through artificial identification or other tool identification; if brain tissue cytotoxic edema only appears in the left brain region, a corresponding first label prediction side is set as the left side; if brain tissue cytotoxic edema only appears in the right brain region, a corresponding first label prediction side is set as the right side; if brain tissue cytotoxic edema appears in both left and right brain regions, a corresponding first label prediction side is set as the bilateral side; each fifth training image and a corresponding first label prediction side constitute a corresponding fifth data record; and all obtained fifth data records constitute a corresponding fifth data set;

[0152] Here, the fifth data set obtained by the embodiment of the application includes a plurality of fifth data records; the fifth data record includes a fifth training image and a first label prediction side; the first label prediction side includes empty, left side, right side and bilateral side;

[0153] Step 176, each first plain scan CT image group in the plain scan CT image library is taken as a corresponding current CT image group; and the corresponding left and right brain C area image pairs, left and right brain L area image pairs, left and right brain IC area image pairs, left and right brain I area image pairs, left and right brain M1 area image pairs, left and right brain M2 area image pairs, left and right brain M3 area image pairs, left and right brain M4 area image pairs, left and right brain M5 area image pairs and left and right brain M6 area image pairs are extracted from the first and second CT collection images of the current CT image group by means of manual image extraction or image extraction of other tools, and each image pair is taken as a corresponding first training subarea image pair, and the ten first training subarea image pairs are sequentially sorted to obtain a corresponding first training subarea image pair sequence; and the left and right brain subarea score of each first training subarea image pair is set as 0 / 1 by means of manual score setting or score setting of other tools, and is recorded as a corresponding first label left brain subarea score and a first label right brain subarea score; and a corresponding first label subarea score pair is composed of the first label left brain subarea score and the first label right brain subarea score of each first training subarea image pair; and all the first label subarea score pairs are sequentially sorted to obtain a corresponding first label subarea score pair sequence; and a corresponding sixth data record is composed of the first training subarea image pair sequence and the first label subarea score pair sequence; and a corresponding sixth data set is composed of all the sixth data records obtained;

[0154] Here, the sixth data set obtained by the embodiment of the application includes a plurality of sixth data records; the sixth data record includes a first training subarea image pair sequence and a first label subarea score pair sequence;

[0155] Step 177, after obtaining the first, second, third, fourth, fifth and sixth data sets, the first model training is performed on the first segmentation model based on the first data set; the second model training is performed on the second segmentation model based on the second data set; the third model training is performed on the third segmentation model based on the third data set; the fourth model training is performed on the fourth segmentation model based on the fourth data set; the fifth model training is performed on the first classification model based on the fifth data set; and the sixth model training is performed on the first scoring model based on the sixth data set;

[0156] Specifically, step 1771, the first model training is performed on the first segmentation model based on the first data set;

[0157] Specifically,

[0158] Step 1771-1, the first training data set and the first evaluation data set are obtained by randomly dividing the first data set into training / evaluation subsets based on a preset first training evaluation ratio;

[0159] The first training evaluation ratio is a preset ratio; the first training data set and the first evaluation data set are composed of a plurality of first data records; the ratio of the total number of records of the first training data set to the total number of records of the first evaluation data set meets the first training evaluation ratio;

[0160] Step 1771-2, the first data record of the first training data set is extracted as the corresponding current training record;

[0161] Step 1771-3, the first training image of the current training record is input into the first segmentation model for sulcus region segmentation processing to obtain the corresponding first predicted semantic map;

[0162] Step 1771-4, the first predicted semantic map and the first label semantic map of the current training record are input into the preset first model loss function to obtain the corresponding first loss value;

[0163] The loss function type of the first model loss function at least includes L1 loss function, L2 loss function and binary cross-entropy loss function;

[0164] Step 1771-5, whether the first loss value meets the preset first loss value range is identified; if the first loss value meets the first loss value range, whether the current training record is the last first data record of the first training data set is identified, if yes, it is turned to step 1771-6, if not, the next first data record of the first training data set is extracted as a new current training record and returned to step 1771-3; if the first loss value does not meet the first loss value range, a round of parameter optimization is performed on the first segmentation model in the direction of making the first model loss function reach the minimum value based on the preset first model parameter optimizer, and step 1771-3 is returned when the round of parameter optimization is completed.

[0165] The first loss value range is a preset loss value range; the first model parameter optimizer at least includes SGD optimizer, ADAM optimizer;

[0166] Step 1771-6, a round of traversal is performed on all first data records of the first evaluation data set; during the traversal, a currently-traversed first data record is taken as a corresponding current evaluation record; a first training image of the current evaluation record is input into the first segmentation model to perform sulcus region segmentation processing to obtain a corresponding second predicted semantic map; and a pixel-level semantic two-classification precision and a pixel-level semantic two-classification recall of the second predicted semantic map are calculated based on a conventional two-classification evaluation method based on precision, recall, and F-score, taking the first label semantic map of the current evaluation record as a true value reference to obtain a corresponding first precision and a first recall, and based on the first precision and the first recall, a corresponding F-score is calculated to obtain a corresponding first F-score; and when the round of traversal ends, a mean value of all obtained first precisions is calculated to obtain a corresponding first average precision, a mean value of all obtained first recalls is calculated to obtain a corresponding first average recall, and a mean value of all obtained first F-scores is calculated to obtain a corresponding first average F-score;

[0167] Step 1771-7, the first average precision, the first average recall, and the first average F-score are identified based on a preset first precision range, a first recall range, and a first F-score range; if the first average precision does not satisfy the first precision range or the first average recall does not satisfy the first recall range or the first average F-score does not satisfy the first F-score range, step 1771-1 is returned to continue training; if the first average precision satisfies the first precision range, the first average recall satisfies the first recall range, and the first average F-score satisfies the first F-score range, the training is stopped and it is confirmed that the first model training is completed;

[0168] Here, the first precision range, the first recall range, and the first F-score range are a set of pre-set precision, recall, and F-score ranges;

[0169] Step 1772, and based on the second data set, a second model training is performed on the second segmentation model;

[0170] Specifically includes:

[0171] Step 1772-1, based on a preset second training evaluation ratio, a random training / evaluation subset division is performed on the second data set to obtain a corresponding second training data set and a second evaluation data set;

[0172] Wherein, the second training evaluation ratio is a pre-set ratio; the second training data set and the second evaluation data set are both composed of a plurality of second data records; the ratio of the total number of records of the second training data set to the total number of records of the second evaluation data set satisfies the second training evaluation ratio;

[0173] Step 1772-2, the first second data record of the second training data set is extracted as a corresponding current training record;

[0174] Step 1772-3, the second training image of the current training record is input into the second segmentation model for brain arterial vessel old infarction area segmentation processing to obtain a corresponding third prediction semantic map;

[0175] Step 1772-4, the third prediction semantic map and the second label semantic map of the current training record are brought into a preset second model loss function for calculation to obtain a corresponding second loss value;

[0176] Wherein, the loss function type of the second model loss function at least includes L1 loss function, L2 loss function and binary classification cross-entropy loss function;

[0177] Step 1772-5, whether the second loss value meets a preset second loss value range is identified; if the second loss value meets the second loss value range, whether the current training record is the last second data record of the second training data set is identified, if yes, it is turned to step 1772-6, if not, the next second data record of the second training data set is extracted as a new current training record and returns to step 1772-3; if the second loss value does not meet the second loss value range, a second model parameter optimizer is used to optimize the second segmentation model in the direction of making the second model loss function reach the minimum value, and returns to step 1772-3 when the parameter optimization is finished;

[0178] Wherein, the second loss value range is a pre-set loss value range; the second model parameter optimizer at least includes SGD optimizer, ADAM optimizer;

[0179] Step 1772-6, a round of traversal is performed on all second data records of the second evaluation data set; and during the traversal, the second data record currently traversed is taken as a corresponding current evaluation record; and the second training image of the current evaluation record is input into the second segmentation model to perform brain arterial vessel old infarction area segmentation processing to obtain a corresponding fourth predicted semantic map; and according to a conventional binary classification evaluation method based on precision, recall and F-score, the binary classification precision and the binary classification recall of the pixel-level semantics of the fourth predicted semantic map are calculated by taking the second label semantic map of the current evaluation record as the true value reference to obtain a corresponding second precision and a second recall, and the F-score corresponding to the second precision and the second recall is calculated to obtain a corresponding second F-score; and when the current round of traversal ends, the mean value of all the second precisions obtained is calculated to obtain a corresponding second average precision, the mean value of all the second recalls obtained is calculated to obtain a corresponding second average recall, and the mean value of all the second F-scores obtained is calculated to obtain a corresponding second average F-score;

[0180] Step 1772-7, the second average precision, the second average recall and the second average F-score are identified based on a preset second precision range, a second recall range and a second F-score range; if the second average precision does not satisfy the second precision range or the second average recall does not satisfy the second recall range or the second average F-score does not satisfy the second F-score range, return to step 1772-1 to continue training; if the second average precision satisfies the second precision range, the second average recall satisfies the second recall range and the second average F-score satisfies the second F-score range, stop training and confirm that the second model training is completed;

[0181] Here, the second precision range, the second recall range and the second F-score range are a set of pre-set precision, recall and F-score ranges;

[0182] Step 1773, and based on the third data set, the third segmentation model is subjected to third model training;

[0183] Specifically includes:

[0184] Step 1773-1, the third data set is subjected to random training / evaluation subset division based on a preset third training evaluation ratio to obtain a corresponding third training data set and a third evaluation data set;

[0185] Wherein, the third training evaluation ratio is a pre-set ratio; the third training data set and the third evaluation data set are both composed of a plurality of third data records; the ratio of the total number of records of the third training data set to the total number of records of the third evaluation data set satisfies the third training evaluation ratio;

[0186] Step 1773-2, the first third data record of the third training data set is extracted as a corresponding current training record;

[0187] Step 1773-3, the third training image of the current training record is input into the third segmentation model for brain blood vessel high-density arterial sign region segmentation processing to obtain a corresponding fifth predicted semantic map;

[0188] Step 1773-4, the fifth predicted semantic map and the third label semantic map of the current training record are brought into a preset third model loss function for calculation to obtain a corresponding third loss value;

[0189] Wherein, the loss function type of the third model loss function at least includes L1 loss function, L2 loss function and binary classification cross-entropy loss function;

[0190] Step 1773-5, whether the third loss value meets a preset third loss value range is identified; if the third loss value meets the third loss value range, whether the current training record is the last third data record of the third training data set is identified, if yes, it is turned to step 1773-6, if not, the next third data record of the third training data set is extracted as a new current training record and returns to step 1773-3; if the third loss value does not meet the third loss value range, a third model parameter optimizer is used to perform a round of parameter optimization on the third segmentation model in the direction of making the third model loss function reach the minimum value, and returns to step 1773-3 when the round of parameter optimization is completed;

[0191] Wherein, the third loss value range is a pre-set loss value range; the third model parameter optimizer at least includes SGD optimizer, ADAM optimizer;

[0192] Step 1773-6, a round of traversal is performed on all third data records of the third evaluation data set; and during the traversal, the third data record currently being traversed is taken as a corresponding current evaluation record; the third training image of the current evaluation record is input into the third segmentation model to perform brain blood vessel high-density arterial sign region segmentation processing to obtain a corresponding sixth predicted semantic map; and according to a conventional binary classification evaluation method based on precision, recall and F-score, the binary classification precision and the binary classification recall of the pixel-level semantics of the sixth predicted semantic map are calculated by taking the third label semantic map of the current evaluation record as the true value reference to obtain a corresponding third precision and a third recall, and the F-score corresponding to the third precision and the third recall is calculated to obtain a corresponding third F-score; and when the current round of traversal ends, the mean value of all the third precisions obtained is calculated to obtain a corresponding third average precision, the mean value of all the third recalls obtained is calculated to obtain a corresponding third average recall, and the mean value of all the third F-scores obtained is calculated to obtain a corresponding third average F-score;

[0193] Step 1773-7, the third average precision, the third average recall and the third average F-score are identified based on a preset third precision range, a third recall range and a third F-score range; if the third average precision does not satisfy the third precision range or the third average recall does not satisfy the third recall range or the third average F-score does not satisfy the third F-score range, return to step 1773-1 to continue training; if the third average precision satisfies the third precision range, the third average recall satisfies the third recall range and the third average F-score satisfies the third F-score range, stop training and confirm that the third model training is completed;

[0194] Here, the third precision range, the third recall range and the third F-score range are a set of pre-set precision, recall and F-score ranges;

[0195] Step 1774, and based on the fourth data set, the fourth segmentation model is subjected to fourth model training;

[0196] Specifically includes:

[0197] Step 1774-1, the fourth data set is subjected to random training / evaluation subset division based on a preset fourth training evaluation ratio to obtain a corresponding fourth training data set and a fourth evaluation data set;

[0198] Wherein, the fourth training evaluation ratio is a pre-set ratio; the fourth training data set and the fourth evaluation data set are both composed of multiple fourth data records; the ratio of the total number of records of the fourth training data set to the total number of records of the fourth evaluation data set satisfies the fourth training evaluation ratio;

[0199] Step 1774-2, the first third data record of the fourth training data set is extracted as a corresponding current training record;

[0200] Step 1774-3, the fourth training image sequence of the current training record is input into the fourth segmentation model for left-right brain ASPECTS partitioning and ventricle region segmentation processing to obtain a corresponding seventh predicted semantic graph sequence; the seventh predicted semantic graph sequence is sequentially sorted by a seventh first predicted semantic graph and a seventh second predicted semantic graph;

[0201] Step 1774-4, the seventh predicted semantic graph sequence and the fourth label semantic graph sequence of the current training record are input into a preset fourth model loss function for calculation to obtain a corresponding fourth loss value;

[0202] wherein,

[0203] w1 and w2 are two preset weighting coefficients, is a preset fourth first model loss function, is a preset fourth second model loss function, Y1 is the seventh first predicted semantic graph of the seventh predicted semantic graph sequence, and Y2 is the seventh second predicted semantic graph of the seventh predicted semantic graph sequence, is a fourth first label semantic graph of the fourth label semantic graph sequence of the current training record, is a fourth second label semantic graph of the fourth label semantic graph sequence of the current training record; the loss function types of the fourth first and second model loss functions are the same, and the loss function types of the fourth first and second model loss functions at least include an L1 loss function, an L2 loss function, and a binary cross-entropy loss function;

[0204] Step 1774-5, whether the fourth loss value meets a preset fourth loss value range is identified; if the fourth loss value meets the fourth loss value range, whether the current training record is the last fourth data record of the fourth training data set is identified; if yes, step 1774-6 is entered; if no, the next fourth data record of the fourth training data set is extracted as a new current training record and step 1774-3 is returned; if the fourth loss value does not meet the fourth loss value range, a fourth segmentation model is subjected to a round of parameter optimization in a direction towards making the fourth first and second model loss functions reach minimum values based on a preset fourth model parameter optimizer, and step 1774-3 is returned at the end of the round of parameter optimization;

[0205] wherein, the fourth loss value range is a preset loss value range; the fourth model parameter optimizer at least includes an SGD optimizer and an ADAM optimizer;

[0206] Step 1774-6, a round of traversal is performed on all fourth data records of the fourth evaluation data set; and in the traversal, a currently-traversed fourth data record is taken as a corresponding current evaluation record; and a fourth training image sequence of the current evaluation record is input into the fourth segmentation model to perform left-right brain ASPECTS partitioning and ventricle region segmentation processing to obtain a corresponding eighth predicted semantic graph sequence; and according to a conventional multi-classification evaluation method based on precision, recall and F-score, a fourth one label semantic graph of the fourth label semantic graph sequence of the current evaluation record is taken as a true value reference to calculate multi-classification precision and multi-classification recall of pixel-level semantics of a fourth one predicted semantic graph of the eighth predicted semantic graph sequence to obtain corresponding fourth one precision and fourth one recall, and based on the fourth one precision and the fourth one recall, an F-score is calculated to obtain corresponding fourth one F-score; and according to the conventional multi-classification evaluation method based on precision, recall and F-score, a fourth two label semantic graph of the fourth label semantic graph sequence of the current evaluation record is taken as a true value reference to calculate multi-classification precision and multi-classification recall of pixel-level semantics of a fourth two predicted semantic graph of the eighth predicted semantic graph sequence to obtain corresponding fourth two precision and fourth two recall, and based on the fourth two precision and the fourth two recall, an F-score is calculated to obtain corresponding fourth two F-score; and based on the fourth one and fourth two precisions, corresponding fourth precision = a1*fourth one precision + a2*fourth two precision is calculated, a1 and a2 being two preset weighting coefficients; and based on the fourth one and fourth two recalls, corresponding fourth recall = a1*fourth one recall + a2*fourth two recall is calculated; and based on the fourth one and fourth two F-scores, corresponding fourth F-score = a1*fourth one F-score + a2*fourth two F-score is calculated; and when the round of traversal ends, all the fourth precisions obtained are averaged to obtain corresponding fourth average precision, all the fourth recalls obtained are averaged to obtain corresponding fourth average recall, and all the fourth F-scores obtained are averaged to obtain corresponding fourth average F-score;

[0207] The eighth predicted semantic graph sequence is sequentially sorted by the fourth one predicted semantic graph and the fourth two predicted semantic graph.

[0208] Step 1774-7, the fourth average precision, the fourth average recall and the fourth average F-score are identified based on a preset fourth precision range, a fourth recall range and a fourth F-score range; if the fourth average precision does not satisfy the fourth precision range or the fourth average recall does not satisfy the fourth recall range or the fourth average F-score does not satisfy the fourth F-score range, step 1774-1 is returned to continue training; if the fourth average precision satisfies the fourth precision range, the fourth average recall satisfies the fourth recall range and the fourth average F-score satisfies the fourth F-score range, training is stopped and it is confirmed that the fourth model training is ended.

[0209] Here, the fourth precision rate range, the fourth recall rate range, and the fourth F-score range are a set of pre-set precision rates, recall rates, and F-scores;

[0210] Step 1775, and performing fifth model training on the first classification model based on the fifth data set;

[0211] Specifically includes:

[0212] Step 1775-1, performing random training / evaluation subset division on the fifth data set based on a pre-set fifth training evaluation ratio to obtain a corresponding fifth training data set and a fifth evaluation data set;

[0213] The fifth training evaluation ratio is a pre-set ratio; the fifth training data set and the fifth evaluation data set each consist of a plurality of fifth data records; and the ratio of the total number of records in the fifth training data set to the total number of records in the fifth evaluation data set satisfies the fifth training evaluation ratio.

[0214] Step 1775-2, extracting the first fifth data record of the fifth training data set as a corresponding current training record;

[0215] Step 1775-3, inputting the fifth training image of the current training record into the first classification model to classify and identify the brain tissue cell cytotoxic edema side information on the fifth training image of the current training record by the first classification model to obtain a corresponding first training prediction side;

[0216] Step 1775-4, inputting the first training prediction side and the first label prediction side of the current training record into a pre-set fifth model loss function to calculate a corresponding fifth loss value;

[0217] The loss function type of the fifth model loss function at least includes an L1 loss function, an L2 loss function, and a multi-class cross-entropy loss function.

[0218] Step 1775-5, identifying whether the fifth loss value satisfies a pre-set fifth loss value range; if the fifth loss value satisfies the fifth loss value range, identifying whether the current training record is the last fifth data record of the fifth training data set; if yes, proceeding to step 1775-6; if no, extracting the next fifth data record of the fifth training data set as a new current training record and returning to step 1775-3; if the fifth loss value does not satisfy the fifth loss value range, performing a round of parameter optimization on the first classification model in the direction of minimizing the fifth model loss function based on a pre-set fifth model parameter optimizer, and returning to step 1775-3 when the round of parameter optimization ends;

[0219] The fifth loss value range is a pre-set loss value range; and the fifth model parameter optimizer at least includes an SGD optimizer and an ADAM optimizer.

[0220] Step 1775-6, a round of traversal is performed on all fifth data records of the fifth evaluation data set; and during the traversal, a currently-traversed fifth data record is taken as a corresponding current evaluation record; the fifth training image of the current evaluation record is input into the first classification model; the brain tissue cell cytotoxic edema brain side information on the fifth training image of the current evaluation record is classified and recognized by the first classification model to obtain a corresponding second training prediction side; and a corresponding prediction-label data pair is formed by the second training prediction side and the first label prediction side of the current evaluation record; and at the end of the round of traversal, all prediction-label data pairs obtained are taken as a corresponding prediction-label data pair set;

[0221] Step 1775-7, according to a conventional multi-classification evaluation method based on precision, recall and F-score, the prediction-label data pair set is used to calculate the precision and the recall to obtain a corresponding fifth precision and a corresponding fifth recall; and the F-score is calculated based on the fifth precision and the fifth recall to obtain a corresponding fifth F-score;

[0222] Step 1775-8, the fifth precision, the fifth recall and the fifth F-score are identified based on a pre-set fifth precision range, a fifth recall range and a fifth F-score range; if the fifth precision does not satisfy the fifth precision range, or the fifth recall does not satisfy the fifth recall range, or the fifth F-score does not satisfy the fifth F-score range, the step 1775-1 is returned to continue the training; if the fifth precision satisfies the fifth precision range, the fifth recall satisfies the fifth recall range, and the fifth F-score satisfies the fifth F-score range, the training is stopped and the fifth model training is confirmed to be completed;

[0223] Here, the fifth precision range, the fifth recall range and the fifth F-score range are a group of pre-set precision, recall and F-score ranges;

[0224] Step 1776, the first scoring model is trained based on the sixth data set;

[0225] Specifically, it includes:

[0226] Step 1776-1, the sixth data set is randomly divided into a sixth training data set and a sixth evaluation data set based on a pre-set sixth training evaluation ratio;

[0227] The sixth training evaluation ratio is a preset ratio; the sixth training data set and the sixth evaluation data set each comprise a plurality of sixth data records; and a ratio of a total number of records in the sixth training data set to a total number of records in the sixth evaluation data set satisfies the sixth training evaluation ratio.

[0228] Step 1776-2, the first sixth data record of the sixth training data set is extracted as a corresponding current training record;

[0229] Step 1776-3, the first training partition image pair sequence of the current training record is input into the first scoring model to obtain a corresponding first training partition score pair sequence;

[0230] Step 1776-4, the first training partition score pair sequence and the first label partition score pair sequence of the current training record are input into a preset sixth model loss function to obtain a corresponding sixth loss value;

[0231] The loss function type of the sixth model loss function at least includes an L1 loss function, an L2 loss function, and a multi-class cross-entropy loss function.

[0232] Step 1776-5, whether the sixth loss value satisfies a preset sixth loss value range is identified; if the sixth loss value satisfies the sixth loss value range, whether the current training record is the last sixth data record of the sixth training data set is identified; if yes, step 1776-6 is performed; if no, the next sixth data record of the sixth training data set is extracted as a new current training record, and step 1776-3 is returned; if the sixth loss value does not satisfy the sixth loss value range, a round of parameter optimization of the first scoring model is performed in a direction of minimizing the sixth model loss function based on a preset sixth model parameter optimizer, and step 1776-3 is returned at the end of the round of parameter optimization.

[0233] The sixth loss value range is a preset loss value range; and the sixth model parameter optimizer at least includes an SGD optimizer and an ADAM optimizer.

[0234] Step 1776-6, all sixth data records of the sixth evaluation data set are iterated for a round; during the iteration, a currently iterated sixth data record is taken as a corresponding current evaluation record; the first training partition image pair sequence of the current evaluation record is input into the first scoring model to obtain a corresponding second training partition score pair sequence; a corresponding prediction-label sequence pair is formed by the second training partition score pair sequence and the first label partition score pair sequence of the current evaluation record; and at the end of the round of iteration, a corresponding prediction-label sequence pair set is formed by all obtained prediction-label sequence pairs.

[0235] Step 1776-7, according to the prediction-label sequence pair set, the sixth precision and the sixth recall are calculated by the conventional multi-classification evaluation method based on the precision, the recall and the F-score; and the corresponding sixth F-score is calculated based on the sixth precision and the sixth recall.

[0236] Step 1776-8, the sixth precision, the sixth recall and the sixth F-score are identified based on the preset sixth precision range, the sixth recall range and the sixth F-score range; if the sixth precision does not satisfy the sixth precision range or the sixth recall does not satisfy the sixth recall range or the sixth F-score does not satisfy the sixth F-score range, return to step 1776-1 to continue training; if the sixth precision satisfies the sixth precision range, the sixth recall satisfies the sixth recall range and the sixth F-score satisfies the sixth F-score range, stop training and confirm that the sixth model training is completed;

[0237] Here, the sixth precision range, the sixth recall range and the sixth F-score range are a set of pre-set precision, recall and F-score ranges;

[0238] Step 178, after the first, second, third, fourth, fifth and sixth model training are all completed, confirm that the collective model training is completed.

[0239] Step 2, after the collective model training is completed, receive an axial brain CT image located at the basal ganglia nuclear layer of the brain and an axial brain CT image located above the basal ganglia nuclear layer of the brain, denoted as a corresponding first CT image and a second CT image; and the first CT image sequence is composed of the first and second CT images.

[0240] Step 3, the first CT image sequence is respectively subjected to image segmentation processing by the first, second, third and fourth segmentation models to obtain the corresponding first semantic graph group, the second semantic graph group, the third semantic graph group and the fourth semantic graph group, and the first classification model is used to classify and identify the first CT image sequence to obtain the corresponding first prediction side group.

[0241] Specifically, it includes: step 31, the first and second CT images of the first CT image sequence are respectively input into the first segmentation model for sulcus region segmentation processing to obtain the corresponding first image semantic graph A 11 and the second image semantic graph A 12 to form the corresponding first semantic graph group;

[0242] Among them, the first semantic graph group includes the first image semantic graph A 11 and the second image semantic graph A 12 ; the first image semantic graph A 11 includes a plurality of first semantic image pixels a11 each first semantic image pixel point a 11 corresponds to a first semantic type; the second image semantic graph A 12 includes a plurality of second semantic image pixel points a 12 each second semantic image pixel point a 12 corresponds to a first semantic type;

[0243] Step 32, input the first and second CT images of the first CT image sequence into the second segmentation model respectively to perform brain arterial vessel old infarction area segmentation processing to obtain corresponding first and second image semantic graphs A 21 and A 22 to form a corresponding second semantic graph group;

[0244] The second semantic graph group includes the first and second image semantic graphs A 21 and A 22 ; the first image semantic graph A 21 includes a plurality of first semantic image pixel points a 21 each first semantic image pixel point a 21 corresponds to a second semantic type; the second image semantic graph A 22 includes a plurality of second semantic image pixel points a 22 each second semantic image pixel point a 22 corresponds to a second semantic type;

[0245] Step 33, input the first and second CT images of the first CT image sequence into the third segmentation model respectively to perform brain vessel high density arterial sign area segmentation processing to obtain corresponding first and second image semantic graphs A 31 and A 32 to form a corresponding third semantic graph group;

[0246] The third semantic graph group includes the first and second image semantic graphs A 31 and A 32 ; the first image semantic graph A 31 includes a plurality of first semantic image pixel points a 31 each first semantic image pixel point a 31 corresponds to a third semantic type; the second image semantic graph A 32 includes a plurality of second semantic image pixel points a 32 each second semantic image pixel point a 32 corresponds to a third semantic type;

[0247] Step 34, input the first CT image sequence into the fourth segmentation model to perform left and right brain ASPECTS partitioning and brain ventricle region segmentation processing to obtain a corresponding semantic graph sequence; and sequentially take the first and second semantic graphs in the obtained semantic graph sequence as corresponding first image semantic graph A 41 and second image semantic graph A 42 to form a corresponding fourth semantic graph group;

[0248] The fourth semantic graph group includes the first image semantic graph A 41 and the second image semantic graph A 42 ; the first image semantic graph A 41 includes a plurality of first semantic image pixels a 41 , and each first semantic image pixel a 41 corresponds to a fourth first semantic type; the second image semantic graph A 42 includes a plurality of second semantic image pixels a 42 , and each second semantic image pixel a 42 corresponds to a fourth second semantic type.

[0249] Step 35, input the first and second CT images of the first CT image sequence into the first classification model, and perform classification recognition on the brain tissue cytotoxic edema brain side information generated on the first or second CT image by the first classification model to obtain a corresponding first image prediction side and a second image prediction side to form a corresponding first prediction type group;

[0250] The first prediction side group includes the first image prediction side and the second image prediction side; the first and second image prediction sides both include empty, left side, right side, and bilateral.

[0251] Step 4, based on the third semantic graph group, the fourth semantic graph group, the first prediction side group, and the first CT image sequence, perform side information comprehensive evaluation to obtain a corresponding first evaluation side;

[0252] The first evaluation side includes empty, left side, right side, and bilateral.

[0253] Specifically, step 4-1, in the first image semantic graph A 31 of the third semantic graph group, each first semantic image pixel a 31 corresponding to the high-density arterial sign region semantics is recorded as a corresponding first type of point; and based on a preset point clustering algorithm, all first type of points are clustered to obtain a plurality of first type of point sets; and based on each first type of point set, a corresponding high-density arterial sign region is marked on the first CT image of the first CT image sequence to obtain a corresponding first marked region.

[0254] The point clustering algorithm at least includes a K-means clustering algorithm, a DBSCAN clustering algorithm, and an OPTICS clustering algorithm.

[0255] Step 4-2, and the second image semantic graph A of the third semantic graph group is obtained by performing semantic segmentation on the second CT image in the first CT image sequence. 32 The second semantic image pixel points a corresponding to the semantic of the high-density arterial sign region in the second CT image are grouped into a corresponding second type of point set; and the second type of point set is labeled on the second CT image in the first CT image sequence to obtain a corresponding second labeled region. 32 The second semantic image pixel points a corresponding to the semantic of the high-density arterial sign region in the second CT image are grouped into a corresponding second type of point set; and the second type of point set is labeled on the second CT image in the first CT image sequence to obtain a corresponding second labeled region.

[0256] Step 4-3, and the first image semantic graph A of the fourth semantic graph group is obtained by performing semantic segmentation on the first CT image in the first CT image sequence. 41 The first semantic image pixel points a corresponding to the semantic of the fourteen types of left and right brain ASPECTS partition regions in the first CT image are grouped into a corresponding third type of point set; and the fourteen third type of point sets are labeled on the first CT image to obtain fourteen first labeled partitions; and seven left brain partitions form a corresponding first left brain partition set, and seven right brain partitions form a corresponding first right brain partition set. 41 The first semantic image pixel points a corresponding to the semantic of the fourteen types of left and right brain ASPECTS partition regions in the first CT image are grouped into a corresponding third type of point set; and the fourteen third type of point sets are labeled on the first CT image to obtain fourteen first labeled partitions; and seven left brain partitions form a corresponding first left brain partition set, and seven right brain partitions form a corresponding first right brain partition set.

[0257] Step 4-4, and the second image semantic graph A of the fourth semantic graph group is obtained by performing semantic segmentation on the second CT image in the first CT image sequence. 42 The second semantic image pixel points a corresponding to the semantic of the six types of left and right brain ASPECTS partition regions in the second CT image are grouped into a corresponding fourth type of point set; and the six fourth type of point sets are labeled on the second CT image to obtain six second labeled partitions; and three left brain partitions form a corresponding second left brain partition set, and three right brain partitions form a corresponding second right brain partition set; and the first and second left brain partition sets are merged to obtain a corresponding third left brain partition set, and the first and second right brain partition sets are merged to obtain a corresponding third right brain partition set. 42 The second semantic image pixel points a corresponding to the semantic of the six types of left and right brain ASPECTS partition regions in the second CT image are grouped into a corresponding fourth type of point set; and the six fourth type of point sets are labeled on the second CT image to obtain six second labeled partitions; and three left brain partitions form a corresponding second left brain partition set, and three right brain partitions form a corresponding second right brain partition set; and the first and second left brain partition sets are merged to obtain a corresponding third left brain partition set, and the first and second right brain partition sets are merged to obtain a corresponding third right brain partition set.

[0258] Step 4-5, a first side evaluation result is set based on the region intersection relationship between the third left brain partition set and the third right brain partition set and all first and second labeled regions; and the third average precision obtained by the third segmentation model in the last model training is taken as a first confidence of the first side evaluation result.

[0259] Specifically comprising: step 4-5-1, setting the corresponding first side evaluation result based on the region intersection relationship of the third left brain partition set and the third right brain partition set with all the first and second marker regions;

[0260] The first side evaluation result includes null, left side, right side and bilateral;

[0261] Specifically comprising: step 4-5-1-1, if all the left brain partitions in the third left brain partition set have no intersection with all the first and second marker regions, and all the right brain partitions in the third right brain partition set have no intersection with all the first and second marker regions, then setting the corresponding first side evaluation result as null;

[0262] Step 4-5-1-2, if at least one left brain partition in the third left brain partition set has intersection with at least one first or second marker region, and at least one right brain partition in the third right brain partition set has intersection with at least one first or second marker region, then setting the corresponding first side evaluation result as bilateral;

[0263] Step 4-5-1-3, if at least one left brain partition in the third left brain partition set has intersection with at least one first or second marker region, and all the right brain partitions in the third right brain partition set have no intersection with all the first and second marker regions, then setting the corresponding first side evaluation result as left side;

[0264] Step 4-5-1-4, if at least one right brain partition in the third right brain partition set has intersection with at least one first or second marker region, and all the left brain partitions in the third left brain partition set have no intersection with all the first and second marker regions, then setting the corresponding first side evaluation result as right side;

[0265] Step 4-5-2, and taking the third average precision obtained by the third segmentation model in the last model training as the first confidence corresponding to the first side evaluation result;

[0266] Step 4-6, and identifying the first and second image prediction sides of the first predicted side group; if both the first and second image prediction sides are null, then setting the corresponding second side evaluation result as null; if one of the first and second image prediction sides is left side, and the other is left side or null, then setting the corresponding second side evaluation result as left side; if one of the first and second image prediction sides is right side, and the other is right side or null, then setting the corresponding second side evaluation result as right side; if one of the first and second image prediction sides is left side, and the other is right side, then setting the corresponding second side evaluation result as bilateral; and taking the fifth precision obtained by the first classification model in the last model training as the second confidence corresponding to the second side evaluation result;

[0267] wherein the second side evaluation result comprises empty, left side, right side and bilateral;

[0268] Step 4-7, and identify the switch state of the preset manual side evaluation switch; if the switch state of the manual side evaluation switch is the open state, set the corresponding third side evaluation result as empty, and set the third confidence degree corresponding to the third side evaluation result as 0; if the switch state of the manual side evaluation switch is the closed state, send the first CT image sequence to the preset first manual side evaluation interface, and receive the third side evaluation result and the corresponding third confidence degree returned by the first manual side evaluation interface;

[0269] wherein the switch state of the manual side evaluation switch comprises the open state and the closed state;

[0270] Step 4-8, identify whether the first, second and third side evaluation results are all the same; if all the same, take any one of the first, second and third side evaluation results as the corresponding current evaluation side and go to step 4-16; if not all the same, go to step 4-9;

[0271] Step 4-9, identify whether the third confidence degree is 0; if the third confidence degree is 0, go to step 4-10; if the third confidence degree is not 0, go to step 4-11;

[0272] Step 4-10, identify whether one of the first and second side evaluation results is empty; if yes, take the other side evaluation result which is not empty as the corresponding current evaluation side and go to step 4-16; if no, take the side evaluation result with the maximum confidence degree in the first and second side evaluation results as the corresponding current evaluation side and go to step 4-16;

[0273] Step 4-11, identify the preset first evaluation mode; if the first evaluation mode is the first mode, go to step 4-12; if the first evaluation mode is the second mode, go to step 4-13; if the first evaluation mode is the third mode, go to step 4-14; if the first evaluation mode is the fourth mode, go to step 4-15;

[0274] wherein the first evaluation mode comprises the first mode, the second mode, the third mode and the fourth mode;

[0275] Step 4-12, take the side evaluation result with the maximum confidence degree in the first, second and third side evaluation results as the corresponding current evaluation side and go to step 4-16;

[0276] Step 4-13, set four initialized empty ballot box sets for the four side type categories of empty, left side, right side and bilateral, and record them as the first, second, third and fourth ballot box sets respectively; and when the first, second or third side evaluation result matches one of the four side type categories, add the current side evaluation result to the first, second, third or fourth ballot box set corresponding to the current side type category; and add the total number of side evaluation results in the finally obtained first, second, third and fourth ballot box sets to the first, second, third and fourth ballot box votes respectively; and take the maximum value of the first, second, third and fourth ballot box votes as the corresponding maximum vote; and identify whether the number of the ballot box set corresponding to the maximum vote is unique; if the number of the ballot box set corresponding to the maximum vote is unique, take the side type category corresponding to the maximum vote among the four side type categories of empty, left side, right side and bilateral as the corresponding current evaluation side and go to step 4-16; if the number of the ballot box set corresponding to the maximum vote is not unique, take the side evaluation result with the maximum confidence among all side evaluation results in all ballot box sets corresponding to the maximum vote as the corresponding current evaluation side and go to step 4-16;

[0277] Step 4-14, take the difference between the first, second or third confidence and a preset baseline confidence threshold as the corresponding first, second or third differential confidence; and set four initialized differential confidence sets for the four side type categories of empty, left side, right side and bilateral, and record them as the first, second, third and fourth differential confidence sets respectively; and when the first, second or third side evaluation result matches one of the four side type categories, add the first, second or third differential confidence corresponding to the current side evaluation result to the first, second, third or fourth differential confidence set corresponding to the current side type category; and calculate the sum of all differential confidences in the finally obtained first, second, third or fourth differential confidence set to obtain the corresponding first, second, third or fourth set sum; and take the maximum value of the first, second, third and fourth set sums as the corresponding maximum sum; and identify whether the number of the differential confidence set corresponding to the maximum sum is unique; if the number of the differential confidence set corresponding to the maximum sum is unique, take the side type category corresponding to the maximum sum among the four side type categories of empty, left side, right side and bilateral as the corresponding current evaluation side and go to step 4-16; if the number of the differential confidence set corresponding to the maximum sum is not unique, take the side evaluation result with the maximum confidence among all side evaluation results in all differential confidence sets corresponding to the maximum sum as the corresponding current evaluation side and go to step 4-16;

[0278] Step 4-15, subtract a baseline confidence threshold from the first, second, or third confidence to obtain a corresponding fourth, fifth, or sixth differential confidence; set four differential confidence sets initialized as empty for the four side type categories of null, left, right, and bilateral, denoted as the fifth, sixth, seventh, and eighth differential confidence sets; when the first, second, or third side evaluation result matches one of the four side type categories, add the fourth, fifth, or sixth differential confidence corresponding to the current side evaluation result to the fifth, sixth, seventh, or eighth differential confidence set corresponding to the current side type category; calculate the mean of all differential confidences in the fifth, sixth, seventh, or eighth differential confidence set to obtain a corresponding first, second, third, or fourth set mean; take the maximum value among the first, second, third, and fourth set means as a corresponding maximum mean; identify whether the number of differential confidence sets corresponding to the maximum mean is unique; if the number of differential confidence sets corresponding to the maximum mean is unique, take the side type category corresponding to the maximum mean among the four side type categories of null, left, right, and bilateral as the corresponding current evaluation side and proceed to step 4-16; if the number of differential confidence sets corresponding to the maximum mean is not unique, take the side evaluation result with the maximum confidence among all side evaluation results in all differential confidence sets corresponding to the maximum mean as the corresponding current evaluation side and proceed to step 4-16;

[0279] Step 4-16, output the final current evaluation side as the corresponding first evaluation side.

[0280] Step 5, perform left and right brain ASPECTS partition image extraction processing on the first CT image sequence to obtain a corresponding first partition image set; filter each partition image in the first partition image set based on a median filtering method; after filtering, perform sulcal region and cerebral arterial blood vessel old infarction region elimination processing on each partition image in the first partition image set based on the first and second semantic graph groups to obtain a corresponding second partition image set; perform ASPECTS partition left and right brain CT image pair extraction and sorting based on the second partition image set to obtain a corresponding first partition image pair sequence.

[0281] Specifically includes: step 51, performing left and right brain ASPECTS partition image extraction processing on the first CT image sequence to obtain a corresponding first partition image set;

[0282] The first partition image set includes twenty first partition images, and the first to twentieth first partition images are respectively corresponding left / right brain C area images, left / right brain L area images, left / right brain IC area images, left / right brain I area images, left / right brain M1 area images, left / right brain M2 area images, left / right brain M3 area images, left / right brain M4 area images, left / right brain M5 area images and left / right brain M6 area images.

[0283] Specifically, the step 511 comprises: extracting the left brain C area image, the left brain L area image, the left brain IC area image, the left brain I area image, the left brain M1 area image, the left brain M2 area image, the left brain M3 area image, the left brain M4 area image, the left brain M5 area image and the left brain M6 area image corresponding to the ten left brain partitions of the third left brain partition set on the first and second CT images of the first CT image sequence as the corresponding ten first partition images.

[0284] The step 512 comprises: extracting the right brain C area image, the right brain L area image, the right brain IC area image, the right brain I area image, the right brain M1 area image, the right brain M2 area image, the right brain M3 area image, the right brain M4 area image, the right brain M5 area image and the right brain M6 area image corresponding to the ten right brain partitions of the third right brain partition set on the first and second CT images of the first CT image sequence as the corresponding ten first partition images.

[0285] The step 513 comprises: forming the corresponding first partition image set by the obtained twenty first partition images.

[0286] The step 52 comprises: filtering each partition image in the first partition image set based on a median filtering mode.

[0287] The step 53 comprises: performing sulcus region and cerebral arterial blood vessel obsolete infarction region elimination processing on each partition image in the first partition image set based on the first and second semantic graph groups after the filtering to obtain the corresponding second partition image set.

[0288] The second partition image set includes twenty second partition images, and the first to twentieth second partition images are respectively corresponding left / right brain C area images, left / right brain L area images, left / right brain IC area images, left / right brain I area images, left / right brain M1 area images, left / right brain M2 area images, left / right brain M3 area images, left / right brain M4 area images, left / right brain M5 area images and left / right brain M6 area images.

[0289] Specifically, the step 531 comprises: extracting the first image semantic graph A 11 Each first semantic image pixel point a 11the corresponding fifth type of points; and clustering all the fifth type of points based on a point clustering algorithm to obtain a plurality of fifth type of point sets; and marking the corresponding sulcus region on the first CT image based on each fifth type of point set to obtain a corresponding third marked region;

[0290] Step 532, and the second image semantic graph A of the first semantic graph group is obtained; 12 each second semantic graph pixel point a corresponding to the sulcus region semantic 12 the corresponding sixth type of points; and clustering all the sixth type of points based on a point clustering algorithm to obtain a plurality of sixth type of point sets; and marking the corresponding sulcus region on the second CT image based on each sixth type of point set to obtain a corresponding fourth marked region;

[0291] Step 533, and the first image semantic graph A of the second semantic graph group is obtained; 21 each first semantic graph pixel point a corresponding to the old infarction region semantic 21 the corresponding seventh type of points; and clustering all the seventh type of points based on a point clustering algorithm to obtain a plurality of seventh type of point sets; and marking the corresponding old infarction region on the first CT image based on each seventh type of point set to obtain a corresponding fifth marked region;

[0292] Step 534, and the second image semantic graph A of the second semantic graph group is obtained; 22 each second semantic graph pixel point a corresponding to the old infarction region semantic 22 the corresponding eighth type of points; and clustering all the eighth type of points based on a point clustering algorithm to obtain a plurality of eighth type of point sets; and marking the corresponding old infarction region on the second CT image based on each eighth type of point set to obtain a corresponding sixth marked region;

[0293] Step 535, and each first partition image in the first partition image set is sequentially taken as a corresponding current partition image; and the intersection region of any third, fourth, fifth or sixth marked region in the current partition image is taken as a corresponding first intersection region; the total number of the obtained first intersection regions is counted to obtain a corresponding first intersection total number; and the first intersection total number is identified; if the first intersection total number is 0, the current partition image is taken as a corresponding second partition image; if the first intersection total number is greater than 0, the local region image of the current partition image covered by each first intersection region is deleted, and the current partition image after the deletion is completed is taken as a corresponding second partition image;

[0294] Step 536, and the second partition image set is composed of all the obtained second partition images;

[0295] Step 54, and based on the second partition image set, ASPECTS partition left and right brain CT image pair extraction and sorting to obtain the corresponding first partition image pair sequence;

[0296] Wherein, the first partition image pair sequence is sequentially sorted by ten first partition image pairs; the first to tenth first partition image pairs are respectively corresponding left and right brain C region image pairs, left and right brain L region image pairs, left and right brain IC region image pairs, left and right brain I region image pairs, left and right brain M1 region image pairs, left and right brain M2 region image pairs, left and right brain M3 region image pairs, left and right brain M4 region image pairs, left and right brain M5 region image pairs and left and right brain M6 region image pairs;

[0297] Specifically includes: step 541, two second partition images corresponding to left / right brain C region image in the second partition image set are combined into a left and right brain C region image pair, which is recorded as a corresponding first partition image pair;

[0298] Step 542, and two second partition images corresponding to left / right brain L region image in the second partition image set are combined into a left and right brain L region image pair, which is recorded as a corresponding first partition image pair;

[0299] Step 543, and two second partition images corresponding to left / right brain IC region image in the second partition image set are combined into a left and right brain IC region image pair, which is recorded as a corresponding first partition image pair;

[0300] Step 544, and two second partition images corresponding to left / right brain I region image in the second partition image set are combined into a left and right brain I region image pair, which is recorded as a corresponding first partition image pair;

[0301] Step 545, and two second partition images corresponding to left / right brain M1 region image in the second partition image set are combined into a left and right brain M1 region image pair, which is recorded as a corresponding first partition image pair;

[0302] Step 546, and two second partition images corresponding to left / right brain M2 region image in the second partition image set are combined into a left and right brain M2 region image pair, which is recorded as a corresponding first partition image pair;

[0303] Step 547, and two second partition images corresponding to left / right brain M3 region image in the second partition image set are combined into a left and right brain M3 region image pair, which is recorded as a corresponding first partition image pair;

[0304] Step 548, and two second partition images corresponding to left / right brain M4 region image in the second partition image set are combined into a left and right brain M4 region image pair, which is recorded as a corresponding first partition image pair;

[0305] Step 549, and two second partition image sets corresponding to the left / right brain M5 region image are combined into a left / right brain M5 region image pair recorded as a corresponding first partition image pair;

[0306] Step 550, and two second partition image sets corresponding to the left / right brain M6 region image are combined into a left / right brain M6 region image pair recorded as a corresponding first partition image pair; and ten first partition image pairs obtained are sequentially sorted to form a corresponding first partition image pair sequence.

[0307] Step 6, the first partition image pair sequence is subjected to ASPECTS scoring by the first scoring model to obtain a corresponding first partition score pair sequence.

[0308] Here, the first partition score pair sequence output by the first scoring model of the embodiment of the application is sequentially sorted by ten first partition score pairs; the first partition score pair in the output sequence, i.e., the first partition score pair sequence, corresponds to the first partition image pair in the input sequence, i.e., the first partition image pair sequence; the first first partition score pair is composed of corresponding left and right brain C region values; the second first partition score pair is composed of corresponding left and right brain L region values; the third first partition score pair is composed of corresponding left and right brain IC region values; the fourth first partition score pair is composed of corresponding left and right brain I region values; the fifth first partition score pair is composed of corresponding left and right brain M1 region values; the sixth first partition score pair is composed of corresponding left and right brain M2 region values; the seventh first partition score pair is composed of corresponding left and right brain M3 region values; the eighth first partition score pair is composed of corresponding left and right brain M4 region values; the ninth first partition score pair is composed of corresponding left and right brain M5 region values; and the tenth first partition score pair is composed of corresponding left and right brain M6 region values; all left / right brain partition values in the first partition score pair sequence are 0 or 1; any left / right brain partition value being 1 indicates that the possibility of ischemic stroke in the corresponding left / right brain partition is relatively high, and any left / right brain partition value being 0 indicates that the possibility of ischemic stroke in the corresponding left / right brain partition is relatively low.

[0309] Step 7, overall score summary and affected side summary are performed based on the first evaluation side and the first partition score pair sequence to obtain a corresponding first summary report;

[0310] The first summary report includes a first total score and a first affected side type; the first total score is an integer between 0 and 10; and the first affected side type includes left side, right side and bilateral side.

[0311] Specifically comprising: step 71, grouping all the left brain partition values with a value of 1 in the first partition score pair sequence into a corresponding first left brain partition value set; and grouping all the right brain partition values with a value of 1 in the first partition score pair sequence into a corresponding first right brain partition value set;

[0312] Step 72, identifying the first evaluation side;

[0313] Step 73, if the first evaluation side is empty, calculating the sum of the values in the first left brain partition value set to obtain a corresponding first total score, and calculating the sum of the values in the first right brain partition value set to obtain a corresponding second total score; and comparing the first total score with the second total score, if the first total score is greater than or equal to the second total score, taking the first total score as a corresponding current total score N c , and setting a corresponding first affected side type as left side, if the first total score is less than the second total score, taking the second total score as a corresponding current total score N c , and setting a corresponding first affected side type as right side;

[0314] Step 74, if the first evaluation side is left side, calculating the sum of the values in the first left brain partition value set and taking the calculation result as a corresponding current total score N c , and setting a corresponding first affected side type as left side;

[0315] Step 75, if the first evaluation side is right side, calculating the sum of the values in the first right brain partition value set and taking the calculation result as a corresponding current total score N c , and setting a corresponding first affected side type as right side;

[0316] Step 76, if the first evaluation side is bilateral, calculating the sum of the values in the first left brain partition value set to obtain a corresponding third total score, and calculating the sum of the values in the first right brain partition value set to obtain a corresponding fourth total score; and comparing the third total score with the fourth total score, if the third total score is greater than or equal to the fourth total score, taking the third total score as a corresponding current total score N c , if the third total score is less than the fourth total score, taking the second total score as a corresponding current total score N c ; and setting a corresponding first affected side type as bilateral;

[0317] Step 77, calculating a corresponding first total score based on a preset ASPECTS score total score N max and a current total score N c ;

[0318] Wherein, the first total score = N max -N c ;

[0319] Here, the ASPECTS score total score N max is 10 by default; the current total score N c and the first total score are each an integer between 0 and 10;

[0320] Step 78, and the corresponding first summary report is composed of the obtained first total score and the first affected side type.

[0321] Fig. 4 is a module structure diagram of a processing device for predicting an ASPECTS score based on CT images according to an embodiment of the present application. The device is a terminal device or a server for implementing the method embodiments, or a device capable of enabling the terminal device or the server to implement the method embodiments, such as a device or a chip system of the terminal device or the server. As shown in Fig. 4, the device includes a model construction and training module 201, a data receiving module 202, an image segmentation and classification module 203, a side comprehensive evaluation module 204, a partition image denoising and disturbance removal module 205, a partition score module 206, and an overall summary module 207.

[0322] The model construction and training module 201 is configured to construct an image semantic segmentation model for segmenting a cerebral sulcus region of a brain CT image, denoted as a corresponding first segmentation model; construct an image semantic segmentation model for segmenting a cerebral arterial old infarction region of a brain CT image, denoted as a corresponding second segmentation model; construct an image semantic segmentation model for segmenting a cerebral vascular high-density arterial sign region of a brain CT image, denoted as a corresponding third segmentation model; construct an image semantic segmentation model for segmenting an ASPECTS partition and a cerebral ventricle region of a brain CT image sequence, denoted as a corresponding fourth segmentation model; construct a first classification model for classifying and identifying brain side information of brain tissue cytotoxic edema on a brain CT image; construct a first scoring model for scoring ASPECTS of ten sets of ASPECTS partition left and right brain CT image pairs; and perform collective model training on the first, second, third, and fourth segmentation models, the first classification model, and the first scoring model based on a pre-set plain CT image library.

[0323] The data receiving module 202 is configured to, after the collective model training is completed, receive an axial brain plain CT image located at a brain basal ganglia nuclear layer and an axial brain plain CT image located above the brain basal ganglia nuclear layer, denoted as a corresponding first CT image and a second CT image; and compose a corresponding first CT image sequence from the first and second CT images.

[0324] The image segmentation and classification module 203 is configured to perform image segmentation on the first CT image sequence by the first, second, third and fourth segmentation models to obtain a corresponding first semantic graph set, a second semantic graph set, a third semantic graph set and a fourth semantic graph set, and perform classification and identification on the first CT image sequence by the first classification model to obtain a corresponding first predicted side type set; the first predicted side type set includes a first image predicted side type and a second image predicted side type; the first and second image predicted side types both include empty, left side, right side and bilateral.

[0325] The side type comprehensive evaluation module 204 is configured to perform comprehensive evaluation on side type information based on the third semantic graph set, the fourth semantic graph set, the first predicted side type set and the first CT image sequence to obtain a corresponding first evaluation side type; the first evaluation side type includes empty, left side, right side and bilateral.

[0326] The partition image denoising and decontamination module 205 is configured to perform left and right brain ASPECTS partition image extraction on the first CT image sequence to obtain a corresponding first partition image set; perform filtering on each partition image in the first partition image set based on a median filtering method; after the filtering is completed, perform sulcal region and cerebral arterial blood vessel old infarction region elimination processing on each partition image in the first partition image set based on the first and second semantic graph sets to obtain a corresponding second partition image set; and perform ASPECTS partition left and right brain CT image pair extraction and sorting based on the second partition image set to obtain a corresponding first partition image pair sequence.

[0327] The partition score module 206 is configured to perform ASPECTS scoring on the first partition image pair sequence by the first scoring model to obtain a corresponding first partition score pair sequence.

[0328] The overall summary module 207 is configured to perform overall score summarization and affected side summarization based on the first evaluation side type and the first partition score pair sequence to obtain a corresponding first summary report; the first summary report includes a first total score and a first affected side type; the first total score is an integer with a value between 0 and 10; and the first affected side type includes left side, right side and bilateral.

[0329] The processing device for predicting ASPECTS score based on CT image provided by the embodiment of the present application can execute the method steps in the method embodiment, and the implementation principle and technical effects are similar, which will not be repeated here.

[0330] It should be noted that the division of the various modules of the above apparatus is only a logical functional division, and all or part of them can be integrated into one physical entity or physically separated when actually implemented. These modules can all be implemented in the form of software invoked by a processing element; all in the form of hardware; or some modules are implemented in the form of software invoked by a processing element, and some modules are implemented in the form of hardware. For example, the model construction and training module can be a separately established processing element, or can be integrated into a chip of the above apparatus, in addition, it can also be stored in the form of program code in the memory of the above apparatus, and the function of the above determination module is invoked and executed by a processing element of the above apparatus. The implementation of other modules is similar. In addition, all or part of these modules can be integrated together or independently implemented. The processing element described herein can be an integrated circuit with signal processing capability. In the implementation process, each step of the above method or each module can be completed by the integrated logic circuit of hardware or the instruction of software in the processing element.

[0331] For example, the above modules can be one or more integrated circuits configured to implement the above method, such as one or more application specific integrated circuits (ASICs), or one or more digital signal processors (DSPs), or one or more field programmable gate arrays (FPGAs), etc. For another example, when a certain module above is implemented in the form of program code invoked by a processing element, the processing element can be a general purpose processor, such as a central processing unit (CPU) or other processor that can invoke program code. For another example, these modules can be integrated together to implement in the form of system on a chip (SOC).

[0332] In the above embodiments, all or part of the embodiments can be implemented by software, hardware, firmware or any combination thereof. When implemented by software, all or part of the embodiments can be implemented in the form of a computer program product. The computer program product includes one or more computer instructions. When loaded and executed by a computer, all or part of the computer program instructions generate the processes or functions described in the foregoing method embodiments. The computer can be a general purpose computer, a special purpose computer, a computer network or other programmable apparatus. The computer instructions can be stored in a computer readable storage medium or transferred from one computer readable storage medium to another computer readable storage medium, for example, the computer instructions can be transferred from one website, computer, server or data center to another website, computer, server or data center through wired (such as coaxial cable, optical fiber, digital subscriber line (Digital Subscriber Line, DSL)) or wireless (such as infrared, wireless, Bluetooth, microwave, etc.) mode. The computer readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server, data center, etc. that includes one or more available media sets. The available medium can be a magnetic medium (for example, floppy disk, hard disk, magnetic tape), an optical medium (for example, DVD), or a semiconductor medium (for example, solid state disk (solid state disk, SSD)) and the like.

[0333] Fig. 5 is a structural schematic diagram of an electronic device according to an embodiment of the present application. The electronic device can be a terminal device or a server for implementing the method of the foregoing embodiments, or a terminal device or a server connected to the terminal device or the server for implementing the method of the foregoing embodiments. As shown in Fig. 5, the electronic device can include a processor 301 (such as a CPU), a memory 302, and a transceiver 303. The transceiver 303 is coupled to the processor 301, and the processor 301 controls the transceiving action of the transceiver 303. The memory 302 can store various instructions for completing various processing functions and implementing the processing steps described in the foregoing embodiments. Preferably, the electronic device related to the embodiments of the present application further includes a power supply 304, a system bus 305 and a communication port 306. The system bus 305 is used to realize the communication connection between elements. The communication port 306 is used for the connection and communication between the electronic device and other external devices.

[0334] The system bus 305 mentioned in FIG. 5 can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. The system bus can be divided into an address bus, a data bus, a control bus, etc. For the convenience of representation, only one thick line is used in the figure, but it does not mean that there is only one bus or only one type of bus. The communication interface is used to realize the communication between the database access device and other devices (such as the client, the read-write library and the read-only library). The memory can contain a Random Access Memory (RAM), and can also include a Non-Volatile Memory, such as at least one disk memory.

[0335] The processor described above can be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), a Graphics Processing Unit (GPU), etc.; can also be a Digital Signal Processor (DSP), an Application-Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA) or other programmable logic device, a discrete gate or transistor logic device, a discrete hardware component.

[0336] It should be noted that the embodiments of the present application also provide a computer readable storage medium, which stores instructions, and when the instructions run on a computer, the computer executes the method and processing procedure provided in the above embodiments.

[0337] The embodiments of the present application also provide a chip for running instructions, which is used to execute the processing steps described in the foregoing method embodiments.

[0338] The embodiment of the present application provides a processing method and device for predicting ASPECTS score based on CT image, electronic equipment and computer readable storage medium. According to the above content, the embodiment of the present application constructs four image segmentation models (a first segmentation model for sulci region segmentation, a second segmentation model for cerebral arterial old infarction region segmentation, a third segmentation model for cerebral vascular high density arterial sign region segmentation, and a fourth segmentation model for left and right brain ASPECTS partition and ventricle region segmentation), a first classification model for classifying brain tissue cytotoxic edema brain side information, and a first scoring model for ASPECTS scoring of ten sets of left and right brain partition CT images. After training of all models is completed, an axial brain CT image located at the basal ganglia nuclear group level of the brain and an axial brain CT image located above the basal ganglia nuclear group level of the brain form a CT image sequence. The first / second / third / fourth segmentation model respectively performs image segmentation on the CT image sequence to obtain corresponding first / second / third / fourth semantic graph sets, and the first classification model classifies and identifies the CT image sequence to obtain a corresponding prediction side group. Based on the third / fourth semantic graph set, the prediction side group and the CT image sequence, a comprehensive evaluation of the side information is performed to obtain a corresponding evaluation side. The CT image sequence is subjected to left and right brain ASPECTS partition image extraction to obtain a corresponding partition image set. The median filtering method is used to filter each partition image in the partition image set, and after filtering is completed, the first / second semantic graph set is used to eliminate the interference region (sulci region and cerebral arterial old infarction region) of each partition image. The first scoring model performs ASPECTS scoring on the partition image sequence to obtain a corresponding partition score pair sequence. Finally, based on the evaluation side and the partition score pair sequence, overall score summary and affected side summary are performed to obtain a corresponding summary report. As can be seen, the fourth segmentation model is used to improve the boundary accuracy of the partition image, the median filtering is used to reduce the noise of the partition image, the first / second segmentation model is used to eliminate two types of interference factor images (sulci region image and cerebral arterial old infarction region image) in the partition image, the third segmentation model+first classification model and a side information comprehensive evaluation mechanism are used to improve the positioning accuracy of the affected side, and the batch processing mechanism of the first scoring model is used to improve the partition scoring efficiency. Through the embodiment of the present application, the overall prediction efficiency is further improved, the partition image boundary accuracy is improved, the noise and interference of the partition image are reduced, the positioning accuracy of the affected side is improved, and the prediction accuracy is improved.

[0339] Those skilled in the art should further appreciate that the elements and algorithms described in connection with the examples disclosed herein can be embodied in electronic hardware, computer software, or in combinations of both. To clearly illustrate this interchangeability of hardware and software, various examples have been described herein in terms of their general functionality. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the overall system. Skilled artisans can implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present application.

[0340] The steps of a method or algorithm described in connection with the examples disclosed herein can be embodied directly in hardware, in a software module executed by a processor, or in a combination of the two. A software module can reside in random access memory (RAM), flash memory, read-only memory (ROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), registers, hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

[0341] The specific implementation described above is further to the purposes, technical solutions, and beneficial effects of the present application. It should be understood that the above description is merely a specific implementation of the present application and is not intended to limit the protection scope of the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A processing method for predicting an ASPECTS score based on CT images, characterized by, The method comprises: constructing an image semantic segmentation model for sulci region segmentation of brain CT images, denoted as a corresponding first segmentation model; constructing an image semantic segmentation model for brain arterial vascular old infarction region segmentation of brain CT images, denoted as a corresponding second segmentation model; constructing an image semantic segmentation model for brain vascular high-density arterial sign region segmentation of brain CT images, denoted as a corresponding third segmentation model; constructing an image semantic segmentation model for left and right brain ASPECTS partition and brain ventricle region segmentation of brain CT image sequences, denoted as a corresponding fourth segmentation model; constructing a first classification model for classifying and identifying brain side information of brain tissue cytotoxic edema on brain CT images; constructing a first scoring model for ASPECTS scoring of ten sets of ASPECTS partition left and right brain CT image pairs; and collectively training the first, second, third and fourth segmentation models, the first classification model and the first scoring model based on a preset plain CT image library; after the collective model training is completed, receiving an axial brain plain CT image located at a brain basal ganglia nuclear layer and an axial brain plain CT image located above the brain basal ganglia nuclear layer, denoted as a corresponding first CT image and a second CT image; and forming a corresponding first CT image sequence from the first and second CT images; performing image segmentation processing on the first CT image sequence by the first, second, third and fourth segmentation models to obtain a corresponding first semantic graph set, a second semantic graph set, a third semantic graph set and a fourth semantic graph set, and performing classification and identification on the first CT image sequence by the first classification model to obtain a corresponding first predicted side set; the first predicted side set comprises a first image predicted side and a second image predicted side; the first and second image predicted sides both comprise empty, left side, right side and bilateral; performing side information comprehensive evaluation based on the third semantic graph set, the fourth semantic graph set, the first predicted side set and the first CT image sequence to obtain a corresponding first evaluation side; the first evaluation side comprises empty, left side, right side and bilateral; performing left and right brain ASPECTS partition image extraction processing on the first CT image sequence to obtain a corresponding first partition image set; performing filtering on each partition image in the first partition image set based on a median filtering method; after the filtering is completed, performing sulci region and brain arterial vascular old infarction region elimination processing on each partition image in the first partition image set based on the first and second semantic graph sets to obtain a corresponding second partition image set; and performing ASPECTS partition left and right brain CT image pair extraction and sorting based on the second partition image set to obtain a corresponding first partition image pair sequence; performing ASPECTS scoring on the first partition image pair sequence by the first scoring model to obtain a corresponding first partition scoring pair sequence; The first evaluation side and the first partition score are used to obtain a first overall report, which includes a first total score and a first affected side type; the first total score is an integer between 0 and 10; and the first affected side type includes left side, right side and bilateral.

2. The processing method for predicting an ASPECTS score based on a CT image according to claim 1, wherein The cerebral vessel hyperdense artery sign and the cerebral tissue cytotoxic edema are typical signs of a type of ischemic stroke; The ASPECTS partition is composed of ten types of brain tissue partitions, namely, a C partition, an L partition, an IC partition, an I partition, an M1 partition, an M2 partition, an M3 partition, an M4 partition, an M5 partition and an M6 partition; the C partition is a caudate nucleus region of the brain; the L partition is a lenticular nucleus region of the brain; the IC partition is an internal capsule region of the brain; the I partition is an insular cortex region of the brain; the M1 partition is a pre-cortical region of the middle cerebral artery; the M2 partition is a lateral cortical region of the middle cerebral artery outside the insula; the M3 partition is a post-cortical region of the middle cerebral artery; the M4 partition is a cortical region of the middle cerebral artery above the M1 partition; the M5 partition is a cortical region of the middle cerebral artery above the M2 partition; the M6 partition is a cortical region of the middle cerebral artery above the M3 partition; and each type of brain tissue partition of the ASPECTS partition is symmetrical between the left and right brain.

3. The processing method for predicting an ASPECTS score based on a CT image according to claim 2, wherein The first segmentation model is implemented based on a U-Net image segmentation network; the first segmentation model is used to perform sulcus region segmentation processing on a current CT image input into the model to obtain a corresponding first semantic map; the current CT image is an axial position brain plain CT image located at a basal ganglia nuclear layer or an axial position brain plain CT image located above the basal ganglia nuclear layer; the first semantic map includes a plurality of first pixel points; each first pixel point corresponds to a first semantic type; and the first semantic type includes a sulcus region semantic and a background semantic; The second segmentation model is implemented based on a U-Net image segmentation network; the second segmentation model is used to perform brain arterial vessel old infarction region segmentation processing on a current CT image input into the model to obtain a corresponding second semantic map; the current CT image is an axial position brain plain CT image located at a basal ganglia nuclear layer or an axial position brain plain CT image located above the basal ganglia nuclear layer; the second semantic map includes a plurality of second pixel points; each second pixel point corresponds to a second semantic type; and the second semantic type includes an old infarction region semantic and a background semantic. The third segmentation model is based on a U-Net image segmentation network; the third segmentation model is used for brain blood vessel high-density arterial sign region segmentation processing of a current CT image input by the model to obtain a corresponding third semantic map; the current CT image is an axial position brain plain CT image located at a basal ganglia nuclear team level or an axial position brain plain CT image located above the basal ganglia nuclear team level; the third semantic map comprises a plurality of third pixel points; each third pixel point corresponds to a third semantic type; the third semantic type comprises a high-density arterial sign region semantic and a background semantic; The fourth segmentation model is based on two U-Net image segmentation networks, denoted as a corresponding first U-Net network and a second U-Net network; the fourth segmentation model is used for left and right brain ASPECTS partition and ventricle region segmentation processing of a current CT image sequence input by the model to obtain a corresponding fourth semantic map sequence, specifically: the fourth segmentation model is used for taking a first CT image in the current CT image sequence input by the model as a corresponding first network input image, and taking a second CT image in the current CT image sequence as a corresponding second network input image; and performing fourteen-class left and right brain ASPECTS partition and ventricle region segmentation processing on the first network input image by the first U-Net network to obtain a corresponding fourth one semantic map; and performing six-class left and right brain ASPECTS partition and ventricle region segmentation processing on the second network input image by the second U-Net network to obtain a corresponding fourth two semantic map; and sequentially sorting the obtained fourth one and fourth two semantic maps to form a corresponding fourth semantic map sequence output; The current CT image sequence is sorted by two CT images, wherein the first CT image is an axial position brain plain CT image located at a basal ganglia nuclear team level, and the second CT image is an axial position brain plain CT image located above the basal ganglia nuclear team level; The fourth one semantic map comprises a plurality of fourth one pixel points; each fourth one pixel point corresponds to a fourth one semantic type; the fourth one semantic type comprises fourteen-class left and right brain ASPECTS partition semantics, ventricle region semantics and background semantics; the fourteen-class left and right brain ASPECTS partition semantics comprise left / right brain C partition semantics, left / right brain L partition semantics, left / right brain IC partition semantics, left / right brain I partition semantics, left / right brain M1 partition semantics, left / right brain M2 partition semantics, and left / right brain M3 partition semantics; The fourth two semantic map comprises a plurality of fourth two pixel points; each fourth two pixel point corresponds to a fourth two semantic type; the fourth two semantic type comprises six-class left and right brain ASPECTS partition semantics, ventricle region semantics and background semantics; the six-class left and right brain ASPECTS partition semantics comprise left / right brain M4 region semantics, left / right brain M5 region semantics, and left / right brain M6 region semantics; The first classification model comprises a first feature extraction module and a first classification module; the first feature extraction module is realized based on a ResNet network; the first classification module is realized based on an MLP network; the first classification model is used for classifying and identifying brain side information of brain tissue cytotoxic edema on a current CT image input by the model to obtain a corresponding predicted side, specifically: the first classification model is used for inputting the current CT image input by the model into the first feature extraction module for feature extraction processing to obtain a corresponding feature tensor; and inputting the feature tensor into the first classification module for classification prediction processing to obtain a corresponding predicted type; the predicted type comprises empty, left side, right side and bilateral; The first scoring model is composed of ten parallel partition scoring models and a partition score merging module; each partition scoring model is composed of a left and right brain image sorting module, a left brain feature extraction module, a right brain feature extraction module, a left and right brain feature difference module, a left brain feature fusion module, a right brain feature fusion module, a left brain score prediction module, a right brain score prediction module and a left and right brain score merging module; the left and right brain score prediction modules are realized based on a binary classification prediction model; The first scoring model is used for ASPECTS scoring of a partition image pair sequence input by the model to obtain a corresponding partition score pair sequence, specifically: the first scoring model is used for inputting each partition image pair in the partition image pair sequence input by the model into the corresponding partition scoring model, sending corresponding partition score pairs obtained by each partition scoring model according to the corresponding partition image pair to the partition score merging module by left and right brain ASPECTS scoring; and sequentially sorting all the partition score pairs obtained by the partition score merging module to form the corresponding partition score pair sequence; The corresponding partition score pair obtained by each partition score model according to the corresponding partition image pair is sent to the partition score merging module, specifically: each partition score model inputs the corresponding partition image pair into the left and right brain image sorting module, extracts the corresponding left brain partition image and right brain partition image from the partition image pair by the left and right brain image sorting module, and sends the corresponding left brain partition image and right brain partition image to the left brain feature extraction module and the right brain feature extraction module; and the left brain feature extraction module extracts radiomics features from the left brain partition image to obtain corresponding left brain radiomics features, which are sent to the left and right brain feature difference module and the left brain feature fusion module; and the right brain feature extraction module extracts radiomics features from the right brain partition image to obtain corresponding right brain radiomics features, which are sent to the left and right brain feature difference module and the right brain feature fusion module; and the left and right brain feature difference module subtracts the difference features of the left brain radiomics features from the right brain radiomics features, and the difference features of the right brain radiomics features from the left brain radiomics features, and records the difference features as corresponding left-right brain radiomics difference features and right-left brain radiomics difference features, and sends the left-right brain radiomics difference features to the corresponding left brain feature fusion module and the right-left brain radiomics difference features to the right brain feature fusion module; and the left brain feature fusion module performs feature splicing processing on the left brain radiomics features and the left-right brain radiomics difference features to obtain corresponding left brain splicing features, which are sent to the left brain score prediction module; and the right brain feature fusion module performs feature splicing processing on the right brain radiomics features and the right-left brain radiomics difference features to obtain corresponding right brain splicing features, which are sent to the right brain score prediction module; and the left brain score prediction module performs binary ASPECTS score prediction processing on the left brain splicing features to obtain corresponding left brain partition score values, which are sent to the left and right brain score merging module; and the right brain score prediction module performs binary ASPECTS score prediction processing on the right brain splicing features to obtain corresponding right brain partition score values, which are sent to the left and right brain score merging module; and the left and right brain score merging module combines the left brain partition score values and the right brain partition score values to obtain the corresponding partition score pair, which is sent to the partition score merging module; The partition image pair sequence is sequentially sorted by ten partition image pairs; each partition image pair corresponds to a type of partition in the ASPECTS partition; the partition image pair corresponds to the partition score model in the first score model; each partition image is composed of the corresponding left brain partition image and right brain partition image; The partition score pair is sequentially ordered by ten partition score pairs; the partition score pairs correspond to the partition image pairs one by one; each partition score pair is composed of a corresponding left brain partition score and a right brain partition score; the left and right brain partition scores are valued as 0 or 1.

4. The processing method for predicting an ASPECTS score based on CT images according to claim 3, characterized in that, The plain CT image library comprises a plurality of first plain CT image groups; each first plain CT image group corresponds to two plain CT acquisition images of an image acquisition object; the first plain CT image group comprises a first acquisition object type, a first CT acquisition image and a second CT acquisition image; the first acquisition object type is used to mark the characteristic type of the corresponding image acquisition object, and at least includes a healthy type without brain diseases, an old infarction history type with old infarction history of brain arterial vessels but without ischemic stroke symptoms, an A type ischemic stroke type without old infarction history of brain arterial vessels but with the ischemic stroke symptoms, and a B type ischemic stroke type with old infarction history of brain arterial vessels and with the ischemic stroke symptoms; the ischemic stroke symptoms include one or all of the brain vessel high-density arterial sign and the brain tissue cytotoxic edema; the first CT acquisition image is an axial brain plain CT image of the corresponding image acquisition object located at the level of basal ganglia nuclear mass; the second CT acquisition image is an axial brain plain CT image of the corresponding image acquisition object located above the level of basal ganglia nuclear mass; the first and second CT acquisition images are acquired in the following manner: once axial brain CT scanning is performed on the image acquisition object to obtain a corresponding current plain CT image sequence, and a CT image located at the level of basal ganglia nuclear mass is extracted from the current plain CT image sequence as the corresponding first CT acquisition image, and a CT image located above the level of basal ganglia nuclear mass is extracted from the current plain CT image sequence as the corresponding second CT acquisition image; The first semantic graph set comprises a first image semantic graph A 11 and a second image semantic graph A 12 ; the first image semantic graph A 11 comprises a plurality of first semantic image pixels a 11 , each of the first semantic image pixels a 11 corresponds to one of the first semantic types; the second image semantic graph A 12 comprises a plurality of second semantic image pixels a 12 , each of the second semantic image pixels a 12 corresponds to one of the first semantic types; The second semantic graph set comprises a first image semantic graph A 21 and a second image semantic graph A 22 ; the first image semantic graph A 21 comprises a plurality of first semantic image pixels a 21 , each of the first semantic image pixels a 21 corresponds to one of the second semantic types; the second image semantic graph A 22 comprises a plurality of second semantic image pixels a 22 , each of the second semantic image pixels a 22 corresponds to one of the second semantic types; The third semantic graph set comprises a first image semantic graph A 31 and a second image semantic graph A 32 ; the first image semantic graph A 31 comprises a plurality of first semantic image pixels a 31 , each of the first semantic image pixels a 31 corresponds to one of the third semantic types; the second image semantic graph A 32 comprises a plurality of second semantic image pixels a 32 , each of the second semantic image pixels a 32 corresponds to one of the third semantic types; The fourth semantic graph set comprises a first image semantic graph A 41 and a second image semantic graph A 42 ; the first image semantic graph A 41 comprises a plurality of first semantic image pixels a 41 , each of the first semantic image pixels a 41 corresponds to a fourth first semantic type; the second image semantic graph A 42 comprises a plurality of second semantic image pixels a 42 , each of the second semantic image pixels a 42 corresponds to a fourth second semantic type; The first partition image set comprises twenty first partition images, and the first to twentieth first partition images are corresponding left / right brain C region images, left / right brain L region images, left / right brain IC region images, left / right brain I region images, left / right brain M1 region images, left / right brain M2 region images, left / right brain M3 region images, left / right brain M4 region images, left / right brain M5 region images and left / right brain M6 region images, respectively; The second partition image set comprises twenty second partition images, and the first to twentieth second partition images are corresponding left / right brain C region images, left / right brain L region images, left / right brain IC region images, left / right brain I region images, left / right brain M1 region images, left / right brain M2 region images, left / right brain M3 region images, left / right brain M4 region images, left / right brain M5 region images and left / right brain M6 region images, respectively; The second partition image set comprises twenty second partition images, and the first to twentieth second partition images are corresponding left / right brain C region images, left / right brain L region images, left / right brain IC region images, left / right brain I region images, left / right brain M1 region images, left / right brain M2 region images, left / right brain M3 region images, left / right brain M4 region images, left / right brain M5 region images and left / right brain M6 region images, respectively; The first partition image pair sequence is sequentially arranged by ten first partition image pairs; the first to tenth first partition image pairs are corresponding left and right brain C region image pairs, left and right brain L region image pairs, left and right brain IC region image pairs, left and right brain I region image pairs, left and right brain M1 region image pairs, left and right brain M2 region image pairs, left and right brain M3 region image pairs, left and right brain M4 region image pairs, left and right brain M5 region image pairs and left and right brain M6 region image pairs respectively; The first partition score pair sequence is sequentially arranged by ten first partition score pairs; the first partition score pairs correspond to the first partition image pairs one by one; the first first partition score pair is composed of corresponding left and right brain C region scores; the second first partition score pair is composed of corresponding left and right brain L region scores; the third first partition score pair is composed of corresponding left and right brain IC region scores; the fourth first partition score pair is composed of corresponding left and right brain I region scores; the fifth first partition score pair is composed of corresponding left and right brain M1 region scores; the sixth first partition score pair is composed of corresponding left and right brain M2 region scores; the seventh first partition score pair is composed of corresponding left and right brain M3 region scores; the eighth first partition score pair is composed of corresponding left and right brain M4 region scores; the ninth first partition score pair is composed of corresponding left and right brain M5 region scores; and the tenth first partition score pair is composed of corresponding left and right brain M6 region scores; all left / right brain partition scores in the first partition score pair sequence are 0 or 1; any left / right brain partition score being 1 indicates that the possibility of ischemic stroke in the corresponding left / right brain partition is relatively high, and any left / right brain partition score being 0 indicates that the possibility of ischemic stroke in the corresponding left / right brain partition is relatively low.

5. The processing method for predicting ASPECTS score based on CT images according to claim 4, characterized in that, The side information comprehensive evaluation based on the third semantic group, the fourth semantic group, the first prediction side group and the first CT image sequence obtains a corresponding first evaluation side, specifically including: Step 501, obtaining a first image semantic graph A of the third semantic graph set 31 Each first semantic image pixel point a corresponding to a high-density arterial sign region semantic is recorded as a first type point; and all the first type points are clustered based on a preset point clustering algorithm to obtain a plurality of first type point sets; and a corresponding first marking region is marked on the first CT image of the first CT image sequence based on each first type point set. 31 Each first semantic image pixel point a corresponding to a high-density arterial sign region semantic is recorded as a first type point; and all the first type points are clustered based on a preset point clustering algorithm to obtain a plurality of first type point sets; and a corresponding first marking region is marked on the first CT image of the first CT image sequence based on each first type point set. The point clustering algorithm at least includes a K-means clustering algorithm, a DBSCAN clustering algorithm and an OPTICS clustering algorithm; Step 502, and the second image semantic graph A of the third semantic graph group 32 Each of the second semantic image pixels a corresponding to the high-density arterial sign region semantics 32 Corresponding to the second type of point; and based on the point clustering algorithm, all the second type of points are clustered to obtain a plurality of second type of point sets; and based on each of the second type of point sets, a corresponding high-density arterial sign region label is performed on the second CT image of the first CT image sequence to obtain a corresponding second label region; Step 503, and the first image semantic graph A of the fourth semantic graph group is obtained 41 Corresponding to each type of semantic partition of the fourteen types of left and right brain ASPECTS partition semantics 41 are grouped into a corresponding third type of point set; and based on the obtained fourteen third type of point sets, corresponding left and right brain ASPECTS partition marking is performed on the first CT image to obtain corresponding fourteen first marked partitions; and seven left brain partitions in the fourteen first marked partitions form a corresponding first left brain partition set, and seven right brain partitions form a corresponding first right brain partition set; Step 504, and the second image semantic graph A of the fourth semantic graph group 42 Corresponding to each type of semantic of the six types of left and right brain ASPECTS partition semantics 42 Grouped into a corresponding fourth type of point set; and based on the obtained six fourth type point sets, corresponding left and right brain ASPECTS partition labels are marked on the second CT image to obtain corresponding six second labeled partitions; and three left brain partitions in the six second labeled partitions form a corresponding second left brain partition set, and three right brain partitions form a corresponding second right brain partition set; and the first and second left brain partition sets are merged to obtain a corresponding third left brain partition set, and the first and second right brain partition sets are merged to obtain a corresponding third right brain partition set; In step 505, a first side evaluation result is set based on the region intersection relationship between the third left brain partition set and the third right brain partition set and all the first and second marking regions; and the third average precision obtained by the third segmentation model in the last model training is taken as a first confidence degree corresponding to the first side evaluation result; The first side evaluation result includes empty, left side, right side and bilateral sides. Step 506, identifying the first image prediction side and the second image prediction side of the first prediction side group; if both the first and second image prediction sides are empty, setting the corresponding second side evaluation result as empty; if one of the first and second image prediction sides is left side and the other is left side or empty, setting the corresponding second side evaluation result as left side; if one of the first and second image prediction sides is right side and the other is right side or empty, setting the corresponding second side evaluation result as right side; if one of the first and second image prediction sides is left side and the other is right side, setting the corresponding second side evaluation result as bilateral; and taking the fifth precision rate obtained by the first classification model in the last model training as the second confidence degree corresponding to the second side evaluation result; The second side evaluation result includes empty, left side, right side and bilateral. Step 507, identifying the switch state of the preset manual side evaluation switch; if the switch state of the manual side evaluation switch is the open state, setting the corresponding third side evaluation result as empty and setting the third confidence degree corresponding to the third side evaluation result as 0; if the switch state of the manual side evaluation switch is the closed state, sending the first CT image sequence to the preset first manual side evaluation interface and receiving the third side evaluation result and the corresponding third confidence degree returned by the first manual side evaluation interface; The switch state of the manual side evaluation switch includes the open state and the closed state. Step 508, identifying whether the first, second and third side evaluation results are all the same; if they are all the same, taking any one of the first, second and third side evaluation results as the corresponding current evaluation side and turning to step 516; if they are not all the same, turning to step 509; Step 509, identifying whether the third confidence degree is 0; if the third confidence degree is 0, turning to step 510; if the third confidence degree is not 0, turning to step 511; Step 510, identifying whether one of the first and second side evaluation results is empty; if yes, taking the other side evaluation result which is not empty as the corresponding current evaluation side and turning to step 516; if no, taking the side evaluation result with the highest confidence degree as the corresponding current evaluation side and turning to step 516; Step 511, identifying the preset first evaluation mode; if the first evaluation mode is the first mode, turning to step 512; if the first evaluation mode is the second mode, turning to step 513; if the first evaluation mode is the third mode, turning to step 514; if the first evaluation mode is the fourth mode, turning to step 515; The first evaluation mode includes the first mode, the second mode, the third mode and the fourth mode. Step 512, taking the side evaluation result with the highest confidence among the first, second and third side evaluation results as the current evaluation side and going to step 516; Step 513, setting four initialized empty ballot box sets corresponding to the first, second, third and fourth ballot box sets for the four side type categories of empty, left side, right side and bilateral; adding the current side evaluation result to the first, second, third or fourth ballot box set corresponding to the current side type category when the first, second or third side evaluation result matches one of the four side type categories; taking the total number of side evaluation results added in the final first, second, third and fourth ballot box sets as the first, second, third and fourth ballot box votes; taking the maximum value among the first, second, third and fourth ballot box votes as the maximum vote; identifying whether the number of ballot box sets corresponding to the maximum vote is unique; taking the side type category corresponding to the maximum vote among the four side type categories of empty, left side, right side and bilateral as the current evaluation side when the number of ballot box sets corresponding to the maximum vote is unique and going to step 516; taking the side evaluation result with the highest confidence among all side evaluation results in all ballot box sets corresponding to the maximum vote as the current evaluation side when the number of ballot box sets corresponding to the maximum vote is not unique and going to step 516; Step 514, subtracting the difference between the first, second or third confidence and a preset baseline confidence threshold as the corresponding first, second or third differential confidence; and setting four differential confidence sets initialized as empty for the four side type categories of empty, left, right and bilateral, denoted as the first, second, third and fourth differential confidence sets respectively; and when the first, second or third side evaluation result matches one of the four side type categories, adding the first, second or third differential confidence corresponding to the current side evaluation result to the first, second, third or fourth differential confidence set corresponding to the current side type category; and summing all differential confidences in the final first, second, third or fourth differential confidence set to obtain the corresponding first, second, third or fourth set sum; and taking the maximum value of the first, second, third and fourth set sums as the corresponding maximum sum; and identifying whether the number of differential confidence sets corresponding to the maximum sum is unique; if the number of differential confidence sets corresponding to the maximum sum is unique, taking the side type category corresponding to the maximum sum among the four side type categories of empty, left, right and bilateral as the corresponding current evaluation side type and proceeding to step 516; if the number of differential confidence sets corresponding to the maximum sum is not unique, taking the side evaluation result with the maximum confidence among all side evaluation results in all differential confidence sets corresponding to the maximum sum as the corresponding current evaluation side type and proceeding to step 516; Step 515, subtracting the difference between the first, second or third confidence and a baseline confidence threshold from the difference as a corresponding fourth, fifth or sixth differential confidence; and setting four differential confidence sets initialized as empty for four side type categories of null, left, right and bilateral, denoted as a corresponding fifth, sixth, seventh and eighth differential confidence set; and adding the fourth, fifth or sixth differential confidence corresponding to the current side type evaluation result to the fifth, sixth, seventh or eighth differential confidence set corresponding to the current side type category when the first, second or third side type evaluation result matches one of the four side type categories; and performing mean value calculation on all differential confidences in the final fifth, sixth, seventh or eighth differential confidence set to obtain a corresponding first, second, third or fourth set mean; and taking the maximum value of the first, second, third and fourth set mean as a corresponding maximum mean; and identifying whether the number of differential confidence sets corresponding to the maximum mean is unique; if the number of differential confidence sets corresponding to the maximum mean is unique, taking the side type category corresponding to the maximum mean among the four side type categories of null, left, right and bilateral as a corresponding current evaluation side type and proceeding to step 516; if the number of differential confidence sets corresponding to the maximum mean is not unique, taking the side type evaluation result with the maximum confidence among all side type evaluation results in all differential confidence sets corresponding to the maximum mean as a corresponding current evaluation side type and proceeding to step 516; Step 516, outputting the final current evaluation side type as a corresponding first evaluation side type.

6. The processing method for predicting an ASPECTS score based on CT images according to claim 5, characterized in that, The first side type evaluation result is set based on the region intersection relationship between the third left brain partition set and the third right brain partition set and all the first and second marker regions, specifically including: If all left brain partitions in the third left brain partition set have no intersection with all the first and second marker regions, and all right brain partitions in the third right brain partition set have no intersection with all the first and second marker regions, the corresponding first side type evaluation result is set as null; If at least one left brain partition in the third left brain partition set has intersection with at least one first or second marker region, and at least one right brain partition in the third right brain partition set has intersection with at least one first or second marker region, the corresponding first side type evaluation result is set as bilateral; If at least one left brain partition in the third left brain partition set has intersection with at least one first or second marker region, and all right brain partitions in the third right brain partition set have no intersection with all the first and second marker regions, the corresponding first side type evaluation result is set as left; If at least one of the third right brain partition sets has an intersection with at least one of the first or second marked areas, and all of the third left brain partition sets have no intersection with all of the first, second marked areas, the corresponding first side evaluation result is set to right side. 7.The processing method for predicting an ASPECTS score based on CT images according to claim 5, characterized in that, The first CT image sequence is subjected to left and right brain ASPECTS partition image extraction processing to obtain a corresponding first partition image set, specifically including: The left brain C region image, left brain L region image, left brain IC region image, left brain I region image, left brain M1 region image, left brain M2 region image, left brain M3 region image, left brain M4 region image, left brain M5 region image and left brain M6 region image corresponding to the ten left brain partitions of the third left brain partition set on the first and second CT images of the first CT image sequence are extracted as the corresponding ten first partition images; And the right brain C region image, right brain L region image, right brain IC region image, right brain I region image, right brain M1 region image, right brain M2 region image, right brain M3 region image, right brain M4 region image, right brain M5 region image and right brain M6 region image corresponding to the ten right brain partitions of the third right brain partition set on the first and second CT images of the first CT image sequence are extracted as the corresponding ten first partition images; And the twenty first partition images obtained form the corresponding first partition image set. 8.The processing method for predicting an ASPECTS score based on CT images according to claim 4, characterized in that, After filtering, the first and second semantic graph groups are used to eliminate sulcal regions and old infarction regions of brain arteries in each partition image in the first partition image set to obtain a corresponding second partition image set, specifically including: the first image semantic graph A of the first semantic graph group 11 each of the first semantic image pixels a corresponding to a sulcus region semantic 11 is recorded as a corresponding fifth type of point; and all the fifth type of points are clustered based on a preset point clustering algorithm to obtain a plurality of fifth type of point sets; and a corresponding third marking region is marked based on each of the fifth type of point sets on the first CT image; the point clustering algorithm at least includes a K-means clustering algorithm, a DBSCAN clustering algorithm, and an OPTICS clustering algorithm. and the second image semantic graph A of the first semantic graph group 12 Each of the second semantic image pixel points a corresponding to a sulcus region semantic 12 is recorded as a corresponding sixth type of point; and all the sixth type of points are clustered based on the point clustering algorithm to obtain a plurality of sixth type of point sets; and a corresponding fourth marking region is marked on the second CT image based on each of the sixth type of point sets. and the first image semantic graph A of the second semantic graph group 21 Each of the first semantic image pixels a corresponding to the old infarction area semantics 21 is recorded as a corresponding seventh type of point; and all the seventh type of points are clustered based on the point clustering algorithm to obtain a plurality of seventh type of point sets; and a corresponding fifth marking area is marked on the first CT image based on each of the seventh type of point sets. and the second image semantic graph A of the second semantic graph group 22 Each of the second semantic image pixels a corresponding to the old infarction area semantics 22 is recorded as a corresponding eighth type of point; and all the eighth type of points are clustered based on the point clustering algorithm to obtain a plurality of eighth type of point sets; and a corresponding sixth marking area is marked on the second CT image based on each of the eighth type of point sets. Each first partition image in the first partition image set is sequentially taken as a corresponding current partition image; the intersection region of the current partition image with any third, fourth, fifth or sixth marked area is taken as a corresponding first intersection region; the total number of first intersection regions obtained is counted to obtain a corresponding first intersection total number; the first intersection total number is identified; if the first intersection total number is 0, the current partition image is taken as a corresponding second partition image; if the first intersection total number is greater than 0, the local region image of the current partition image covered by each first intersection region is deleted, and the current partition image after deletion is taken as a corresponding second partition image; And all the second partition images obtained form the corresponding second partition image set. 9.The processing method for predicting an ASPECTS score based on CT images according to claim 4, characterized in that, The first evaluation side and the first partition score are used to obtain a corresponding first summary report, specifically including: All left brain partition values with a value of 1 in the first partition score sequence are grouped into a corresponding first left brain partition value set; and all right brain partition values with a value of 1 in the first partition score sequence are grouped into a corresponding first right brain partition value set; The first evaluation side is identified; If the first evaluation side is empty, the sum of the scores of the first left-brain sub-zone score set is calculated to obtain a corresponding first total score, and the sum of the scores of the first right-brain sub-zone score set is calculated to obtain a corresponding second total score; and the first total score and the second total score are compared, if the first total score is greater than or equal to the second total score, the first total score is taken as a corresponding current total score N c , and a corresponding first affected side type is set as left side, if the first total score is less than the second total score, the second total score is taken as the corresponding current total score N c , and the corresponding first affected side type is set as right side; If the first evaluation side is the left side, the sum of scores of the first left brain partition score set is calculated and the calculation result is taken as the corresponding current total score N c , and the corresponding first affected side type is set as the left side. If the first evaluation side is the right side, the sum of scores of the first right brain partition score set is calculated and the calculation result is taken as the corresponding current total score N c , and the corresponding first affected side type is set as the right side; If the first evaluation side is bilateral, the sum of the scores of the first left brain sub-zone score set is calculated to obtain a corresponding third total score, and the sum of the scores of the first right brain sub-zone score set is calculated to obtain a corresponding fourth total score; and the third and fourth total scores are compared, if the third total score is greater than or equal to the fourth total score, the third total score is taken as the corresponding current total score N c , if the third total score is less than the fourth total score, the second total score is taken as the corresponding current total score N c ; and the corresponding first affected side type is set to bilateral; and based on a preset ASPECTS score total score N max and the current total score N c corresponding to the first total score, the first total score = N max -N c ; the ASPECTS score total score N max is 10 by default; the current total score N c and the first total score are both integers with values between 0 and 10. and the first total score and the first affected side type are combined to form a corresponding first summary report.

10. An apparatus for performing the processing method of predicting an ASPECTS score based on CT images according to any one of claims 1-9, characterized in that, The device comprises a model construction and training module, a data receiving module, an image segmentation and classification module, a side comprehensive evaluation module, a partition image noise reduction and disturbance removal module, a partition score module, and an overall summary module. The model construction and training module is configured to construct an image semantic segmentation model for segmenting sulcal regions of a brain CT image, denoted as a corresponding first segmentation model; construct an image semantic segmentation model for segmenting old arterial infarction regions of a brain CT image, denoted as a corresponding second segmentation model; construct an image semantic segmentation model for segmenting high-density arterial sign regions of a brain CT image, denoted as a corresponding third segmentation model; construct an image semantic segmentation model for segmenting ASPECTS partitions and ventricular regions of a brain CT image sequence, denoted as a corresponding fourth segmentation model; construct a first classification model for classifying and identifying brain side information of brain tissue cytotoxic edema on a brain CT image; and construct a first scoring model for scoring ASPECTS of ten sets of ASPECTS partitioned left and right brain CT image pairs; and collectively train the first, second, third, and fourth segmentation models, the first classification model, and the first scoring model based on a pre-set plain CT image library. The data receiving module is configured to receive an axial brain plain CT image located at a brain basal ganglia level and an axial brain plain CT image located above the brain basal ganglia level, denoted as a corresponding first CT image and a second CT image, after the collective model training is completed; and form a corresponding first CT image sequence from the first and second CT images. The image segmentation and classification module is configured to perform image segmentation processing on the first CT image sequence by the first, second, third, and fourth segmentation models to obtain corresponding first, second, third, and fourth semantic graph groups, and perform classification and identification on the first CT image sequence by the first classification model to obtain a corresponding first predicted side group; the first predicted side group comprises a first image predicted side and a second image predicted side; the first and second image predicted sides both comprise empty, left side, right side, and bilateral sides. The side comprehensive evaluation module is configured to perform side information comprehensive evaluation based on the third semantic graph group, the fourth semantic graph group, the first predicted side group, and the first CT image sequence to obtain a corresponding first evaluation side; the first evaluation side comprises empty, left side, right side, and bilateral sides. The partition image denoising and decontamination module is configured to perform left and right brain ASPECTS partition image extraction processing on the first CT image sequence to obtain a corresponding first partition image set; perform filtering on each partition image in the first partition image set based on a median filtering method; perform sulcal region and cerebral arterial blood vessel old infarction region elimination processing on each partition image in the first partition image set based on the first and second semantic graph groups after the filtering is completed to obtain a corresponding second partition image set; and perform ASPECTS partition left and right brain CT image pair extraction and sorting based on the second partition image set to obtain a corresponding first partition image pair sequence; The partition score module is configured to perform ASPECTS scoring on the first partition image pair sequence by the first scoring model to obtain a corresponding first partition score pair sequence. The overall summary module is configured to perform overall score summarization and affected side summarization based on the first evaluation side and the first partition score pair sequence to obtain a corresponding first summary report; the first summary report includes a first total score and a first affected side type; the first total score is an integer with a value between 0 and 10; and the first affected side type includes left side, right side and bilateral.

11. An electronic device, comprising: Comprise: a memory, a processor and a transceiver; the processor is configured to be coupled with the memory, read and execute instructions in the memory to implement the method of any one of claims 1-9; the transceiver is coupled with the processor, and the transceiver is controlled by the processor to perform message transmission and reception.

12. A computer-readable storage medium, characterized in that, The computer readable storage medium stores computer instructions, when the computer instructions are executed by a computer, the computer executes the method of any one of claims 1-9.

Images