CN116313089B - Method for predicting risk of atrial fibrillation after stroke, computer equipment and medium - Google Patents
Method for predicting risk of atrial fibrillation after stroke, computer equipment and medium Download PDFInfo
- Publication number
- CN116313089B CN116313089B CN202310263744.1A CN202310263744A CN116313089B CN 116313089 B CN116313089 B CN 116313089B CN 202310263744 A CN202310263744 A CN 202310263744A CN 116313089 B CN116313089 B CN 116313089B
- Authority
- CN
- China
- Prior art keywords
- image
- stroke
- atrial fibrillation
- post
- risk
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 206010003658 Atrial Fibrillation Diseases 0.000 title claims abstract description 107
- 238000000034 method Methods 0.000 title claims abstract description 38
- 239000013598 vector Substances 0.000 claims description 50
- 238000000605 extraction Methods 0.000 claims description 42
- 210000004556 brain Anatomy 0.000 claims description 28
- JXSJBGJIGXNWCI-UHFFFAOYSA-N diethyl 2-[(dimethoxyphosphorothioyl)thio]succinate Chemical compound CCOC(=O)CC(SP(=S)(OC)OC)C(=O)OCC JXSJBGJIGXNWCI-UHFFFAOYSA-N 0.000 claims description 21
- 230000006870 function Effects 0.000 claims description 15
- 238000013058 risk prediction model Methods 0.000 claims description 13
- 238000011176 pooling Methods 0.000 claims description 12
- 230000004913 activation Effects 0.000 claims description 11
- 230000006835 compression Effects 0.000 claims description 10
- 238000007906 compression Methods 0.000 claims description 10
- 230000005284 excitation Effects 0.000 claims description 10
- 238000007781 pre-processing Methods 0.000 claims description 10
- 238000012549 training Methods 0.000 claims description 10
- 230000003902 lesion Effects 0.000 claims description 9
- 239000011159 matrix material Substances 0.000 claims description 9
- 230000004927 fusion Effects 0.000 claims description 8
- 238000004590 computer program Methods 0.000 claims description 7
- 238000013135 deep learning Methods 0.000 claims description 6
- 238000003709 image segmentation Methods 0.000 claims description 3
- 238000012545 processing Methods 0.000 claims description 3
- 230000011218 segmentation Effects 0.000 claims description 3
- 208000006011 Stroke Diseases 0.000 description 88
- 238000004891 communication Methods 0.000 description 8
- 206010008190 Cerebrovascular accident Diseases 0.000 description 7
- 208000032382 Ischaemic stroke Diseases 0.000 description 7
- 238000010586 diagram Methods 0.000 description 4
- 238000003384 imaging method Methods 0.000 description 4
- 238000001514 detection method Methods 0.000 description 3
- 238000012544 monitoring process Methods 0.000 description 3
- 238000005481 NMR spectroscopy Methods 0.000 description 2
- 208000032109 Transient ischaemic attack Diseases 0.000 description 2
- 238000009825 accumulation Methods 0.000 description 2
- 230000001154 acute effect Effects 0.000 description 2
- 238000011156 evaluation Methods 0.000 description 2
- 238000004393 prognosis Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 201000010875 transient cerebral ischemia Diseases 0.000 description 2
- LFQSCWFLJHTTHZ-UHFFFAOYSA-N Ethanol Chemical compound CCO LFQSCWFLJHTTHZ-UHFFFAOYSA-N 0.000 description 1
- 206010019280 Heart failures Diseases 0.000 description 1
- 206010020772 Hypertension Diseases 0.000 description 1
- 102000004895 Lipoproteins Human genes 0.000 description 1
- 108090001030 Lipoproteins Proteins 0.000 description 1
- 208000005764 Peripheral Arterial Disease Diseases 0.000 description 1
- 208000030831 Peripheral arterial occlusive disease Diseases 0.000 description 1
- 230000003143 atherosclerotic effect Effects 0.000 description 1
- 239000008280 blood Substances 0.000 description 1
- 210000004369 blood Anatomy 0.000 description 1
- 210000003710 cerebral cortex Anatomy 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 206010012601 diabetes mellitus Diseases 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000013467 fragmentation Methods 0.000 description 1
- 238000006062 fragmentation reaction Methods 0.000 description 1
- 208000019622 heart disease Diseases 0.000 description 1
- 230000000302 ischemic effect Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 208000030159 metabolic disease Diseases 0.000 description 1
- 208000010125 myocardial infarction Diseases 0.000 description 1
- 230000003449 preventive effect Effects 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 230000033764 rhythmic process Effects 0.000 description 1
- 230000009863 secondary prevention Effects 0.000 description 1
- 230000000391 smoking effect Effects 0.000 description 1
- BGRJTUBHPOOWDU-UHFFFAOYSA-N sulpiride Chemical compound CCN1CCCC1CNC(=O)C1=CC(S(N)(=O)=O)=CC=C1OC BGRJTUBHPOOWDU-UHFFFAOYSA-N 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 238000010200 validation analysis Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
- G16H50/30—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for calculating health indices; for individual health risk assessment
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0012—Biomedical image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/30—Determination of transform parameters for the alignment of images, i.e. image registration
- G06T7/33—Determination of transform parameters for the alignment of images, i.e. image registration using feature-based methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20112—Image segmentation details
- G06T2207/20132—Image cropping
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Medical Informatics (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Data Mining & Analysis (AREA)
- Public Health (AREA)
- General Health & Medical Sciences (AREA)
- Radiology & Medical Imaging (AREA)
- Artificial Intelligence (AREA)
- Quality & Reliability (AREA)
- Primary Health Care (AREA)
- Epidemiology (AREA)
- Pathology (AREA)
- Databases & Information Systems (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Evolutionary Computation (AREA)
- Biomedical Technology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Measuring And Recording Apparatus For Diagnosis (AREA)
Abstract
The invention discloses a post-stroke atrial fibrillation risk prediction method, computer equipment and a medium, and relates to the technical field of post-stroke atrial fibrillation risk prediction.
Description
Technical Field
The invention relates to the technical field of prediction of risk of post-stroke atrial fibrillation, in particular to a method, computer equipment and medium for predicting risk of post-stroke atrial fibrillation.
Background
Atrial fibrillation-related ischemic strokes and transient ischemic attacks are very high in proportion, and atrial fibrillation users who combine ischemic strokes can be classified into past atrial fibrillation (atrial fibrillation found before ischemic strokes) and atrial fibrillation found after ischemic strokes. Compared with sinus rhythm users, ischemic stroke post-atrial fibrillation users have higher risk of recurrence of stroke and death, so that early recognition of the post-atrial fibrillation is particularly important for optimizing a secondary prevention strategy of the users and improving prognosis of the users. Which ischemic stroke users are the high risk group of post-stroke atrial fibrillation, how to quickly and efficiently identify this group in clinical work remains the difficulty of current research. Although various predictive scores and factors are found at present, and whether the risk of atrial fibrillation after the stroke exists is judged according to the predictive scores and factors, the adopted imaging information is mostly based on manual interpretation, so that the subjectivity is strong, and the problem of offset misjudgment of an interpreter exists.
Disclosure of Invention
The invention aims to provide a method, computer equipment and medium for predicting risk of post-stroke atrial fibrillation, which can improve accuracy of predicting risk of post-stroke atrial fibrillation.
In order to achieve the above object, the present invention provides the following solutions:
a method of predicting risk of atrial fibrillation after a stroke, the method comprising:
acquiring risk prediction information of a user; the risk prediction information comprises multi-modal images and baseline clinical data information; the multi-mode image comprises a T1 image, a T2 image, a FLAIR image and a DWI image; the baseline clinical data information includes age information, gender information, past history information, and NIHSS scoring information;
preprocessing the risk prediction information to obtain preprocessed risk prediction information; the preprocessed risk prediction information comprises preprocessed multi-mode images, preprocessed baseline clinical data information, focus masks and focus volume characteristics;
inputting the preprocessed risk prediction information into a trained post-stroke atrial fibrillation risk prediction model, and predicting the probability of occurrence of post-stroke atrial fibrillation of the user to obtain a post-stroke atrial fibrillation risk prediction probability;
and comparing the predicted probability of the risk of the post-stroke atrial fibrillation with a preset probability value to obtain the risk level of the post-stroke atrial fibrillation of the user.
Optionally, the preprocessing the risk prediction information specifically includes:
carrying out structural image registration on the multi-modal image and the registration template image to obtain a registered multi-modal image and a registration matrix;
performing brain extraction on the registered multimodal images to obtain brain extraction results; clipping the registered multi-modal image according to the brain extraction result to obtain a clipped multi-modal image; the cut multi-mode image is the preprocessed multi-mode image and comprises a cut T1 image, a cut T2 image, a cut FLAIR image and a cut DWI image;
dividing the stroke focus of the cut FLAIR image to obtain a focus mask of the FLAIR image; dividing the stroke focus of the cut DWI image to obtain a DWI image focus mask; the DWI image focus mask and the FLAIR image focus mask form the focus mask;
determining focal volume characteristics according to the DWI image focal mask;
and carrying out coding pretreatment on the baseline clinical data information to obtain pretreated baseline clinical data information.
Optionally, the dividing specifically includes:
the segmentation is performed by using a manual outlining method or a deep learning image segmentation model.
Optionally, the determining the focal volume feature according to the DWI image focal mask specifically includes:
calculating a brain region map of the user using the registration matrix and the registration template image;
and obtaining the number of focus voxels of the stroke according to the brain region map and the DWI image focus mask, wherein the number of focus voxels of the stroke is the focus volume characteristic.
Optionally, the encoding preprocessing of the baseline clinical data information specifically includes:
and performing Z-score standardization on the age information, performing One-hot encoding on the gender information, performing One-hot encoding on the past history information, and performing Z-score standardization on the NIHSS scoring information.
Optionally, the trained risk prediction model for post-stroke atrial fibrillation comprises five sub-models; the five sub-models are all models which are obtained by taking sample risk prediction information of a sample user as input and taking whether the sample user has suffered a stroke or not as label training; the pre-processed risk prediction information is input into a trained post-stroke atrial fibrillation risk prediction model, the probability of occurrence of post-stroke atrial fibrillation of the user is predicted, and the post-stroke atrial fibrillation risk prediction probability is obtained, and specifically comprises the following steps:
the preprocessed risk prediction information is respectively input into five sub-models to obtain five risk prediction sub-probabilities of the post-stroke atrial fibrillation;
and calculating the average value of the five risk predictors of the post-stroke atrial fibrillation to obtain the risk predictors of the post-stroke atrial fibrillation.
Optionally, the sub-model comprises an image feature extraction module, a compression excitation module, a data feature extraction module, a splicing module and a prediction module;
the image feature extraction module is used for extracting features of the preprocessed multi-mode image and the focus mask to obtain an image feature vector;
the compression excitation module is used for carrying out global pooling on the image feature vectors, processing the image feature vectors subjected to global pooling by utilizing an activation function to obtain weight vectors, and multiplying the weight vectors and the image feature vectors in channel dimensions to obtain channel multiplication feature vectors;
the data feature extraction module is used for carrying out feature extraction on the preprocessed baseline clinical data information and the focus volume feature to obtain a data feature vector;
the splicing module is used for splicing the channel product feature vector and the data feature vector to obtain fusion features;
and the prediction module is used for obtaining the risk predictor probability of the post-stroke atrial fibrillation according to the fusion characteristics.
Optionally, the image feature extraction module includes a plurality of residual units; each residual error unit comprises a plurality of convolution layers which are connected in sequence;
the compression excitation module comprises a global pooling layer, a ReLU layer and a Sigmoid layer which are sequentially connected;
the data characteristic extraction module comprises five data characteristic extraction sub-modules; each data characteristic extraction submodule comprises a first full-connection layer, a ReLu layer and a second full-connection layer which are sequentially connected;
the prediction module comprises a third full-connection layer, a first ReLu layer, a fourth full-connection layer, a second ReLu layer and a fifth full-connection layer which are sequentially connected.
The invention also provides a computer device comprising a memory and a processor, the memory storing a computer program which, when executed by the processor, causes the processor to perform the steps of the method for predicting risk of atrial fibrillation after stroke.
The invention also provides a computer readable storage medium storing a computer program adapted to be loaded by a processor and to perform the method of post-stroke atrial fibrillation risk prediction method described above.
According to the specific embodiment provided by the invention, the invention discloses the following technical effects: the invention provides a risk prediction method, computer equipment and medium for post-stroke atrial fibrillation, which comprises the steps of firstly obtaining risk prediction information of a user, preprocessing the risk prediction information, inputting the preprocessed risk prediction information into a trained post-stroke atrial fibrillation risk prediction model, predicting the probability of occurrence of post-stroke atrial fibrillation of the user to obtain the risk prediction probability of post-stroke atrial fibrillation, and comparing the risk prediction probability of post-stroke atrial fibrillation with a preset probability value to obtain the risk level of post-stroke atrial fibrillation of the user.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions of the prior art, the drawings that are needed in the embodiments will be briefly described below, it being obvious that the drawings in the following description are only some embodiments of the present invention, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
Fig. 1 is a flowchart of a method for predicting risk of atrial fibrillation after stroke according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of a sub-model structure according to an embodiment of the present invention;
fig. 3 is a schematic diagram of a pro-actionhead network structure according to an embodiment of the present invention;
fig. 4 is a schematic structural diagram of a computer device according to an embodiment of the present invention.
Symbol description:
1000-a computer device; 1001-a processor; 1002-a communication bus; 1003-user interface; 1004-a network interface; 1005-memory.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Most current assessment models for risk of post-stroke atrial fibrillation events rely on a large amount of multidimensional data, such as blood markers, echocardiogram, electrocardiographic monitoring data and the like of users. However, the time consumption of the multi-dimensional information collection leads to predictable time delay, the data dimension collected among different hospitals is different, and different users can cooperate with the collected data content to have the same difference, so that the external verification and popularization of the model are difficult. In addition, the imaging information adopted in the current research is mostly based on manual interpretation, so that bias and heterogeneity among readers exist, and the whole imaging information cannot be included.
Based on the defects in the prior art, the invention provides a prediction method, computer equipment and medium for risk of post-stroke atrial fibrillation, which are used for predicting the risk probability of post-stroke atrial fibrillation of a user through a machine learning method, have strong objectivity, avoid the influence of artificial subjective factors and further improve the accuracy of risk prediction of post-stroke atrial fibrillation. The invention adopts the currently accepted noninvasive craniocerebral MRI detection, the applied sequence is a detection sequence T1/T2/DWI/FLAIR commonly developed in clinical routine, no additional evaluation is needed, the input baseline clinical data is gender, age, past history information and NIHSS score of the user, the information can be acquired at the first time of user admission and is not dependent on other additional detection, the invention utilizes the homogeneity data (information for risk prediction) which can be acquired early in the real world to carry out risk prediction, reduces the dependence and hysteresis of multidimensional data, carries out risk probability prediction of post-stroke atrial fibrillation of the user in early stage of morbidity, screens out effective secondary preventive intervention for atrial fibrillation users, and has obvious advantages and clinical application value for reducing functional disability of the user and improving prognosis.
In order that the above-recited objects, features and advantages of the present invention will become more readily apparent, a more particular description of the invention will be rendered by reference to the appended drawings and appended detailed description.
As shown in fig. 1, the present invention provides a method for predicting risk of atrial fibrillation after a stroke, the method comprising:
s1: acquiring risk prediction information of a user; the risk prediction information comprises multi-modal images and baseline clinical data information; the multi-mode image comprises a T1 image, a T2 image, a FLAIR image and a DWI image; the baseline clinical data information includes age information, gender information, past history information, and NIHSS scoring information.
S2: preprocessing the risk prediction information to obtain preprocessed risk prediction information; the preprocessed risk prediction information comprises preprocessed multi-modal images, preprocessed baseline clinical data information, focus masks and focus volume features.
S3: and inputting the preprocessed risk prediction information into a trained post-stroke atrial fibrillation risk prediction model, and predicting the probability of occurrence of post-stroke atrial fibrillation of the user to obtain post-stroke atrial fibrillation risk prediction probability.
S4: and comparing the predicted probability of the risk of the post-stroke atrial fibrillation with a preset probability value to obtain the risk level of the post-stroke atrial fibrillation of the user.
In this embodiment, the preprocessing the risk prediction information in step S2 specifically includes:
and carrying out structural image registration on the multi-modal image and the registration template image to obtain a registered multi-modal image and a registration matrix.
Performing brain extraction on the registered multimodal images to obtain brain extraction results; clipping the registered multi-modal image according to the brain extraction result to obtain a clipped multi-modal image; the cut multi-mode image is the preprocessed multi-mode image, and comprises a cut T1 image, a cut T2 image, a cut FLAIR image and a cut DWI image.
Dividing the stroke focus of the cut FLAIR image to obtain a focus mask of the FLAIR image; dividing the stroke focus of the cut DWI image to obtain a DWI image focus mask; the DWI image lesion mask and the FLAIR image lesion mask constitute the lesion mask.
And determining focal volume characteristics according to the DWI image focal mask.
And carrying out coding pretreatment on the baseline clinical data information to obtain pretreated baseline clinical data information.
The determining the focus volume feature according to the DWI image focus mask specifically includes:
and calculating a brain region map of the user by using the registration matrix and the registration template image.
And obtaining the number of focus voxels of the stroke according to the brain region map and the DWI image focus mask, wherein the number of focus voxels of the stroke is the focus volume characteristic.
The preprocessing of the baseline clinical data information comprises the following steps:
and performing Z-score standardization on the age information, performing One-hot encoding on the gender information, performing One-hot encoding on the past history information, and performing Z-score standardization on the NIHSS scoring information.
Specifically, since brain morphologies of different user individuals are different, in order to compare the differences in brain structures and functions of different user individuals, it is necessary to register the multimodal images of different user individuals onto one standard brain template (registered template image). Therefore, by using registration software ANTs, each MRI multi-modal data (multi-modal image) of the user and the registration template image are subjected to structural image registration, so as to obtain each MRI modal image (multi-modal image after registration) consistent with the registration template image and a registration matrix, wherein the registration matrix can deform the original image (multi-modal image) to enable the properties of the brain of the individual and the template image to be consistent as much as possible, and the properties comprise the size, the spatial resolution and the like of the image.
And performing brain extraction on the registered multi-modal images, and obtaining a non-zero region of the multi-modal images according to a brain extraction result. Clipping the registered multi-mode image based on a non-zero region, and storing clipped multi-mode image data (a clipped T1 image, a clipped T2 image, a clipped FLAIR image and a clipped DWI image) as npz files, wherein the related attributes are stored as pkl files so as to accelerate the loading speed of later data. And then dividing the stroke focus in the cut DWI image and the cut FLAIR image by using a manual outlining method or a deep learning image division model to obtain a DWI image focus mask and a FLAIR image focus mask. The deep learning image segmentation model in this embodiment may select models such as 3DUNet and nnUNet.
The present example also uses the accumulation of stroke lesions in different cortical areas as risk prediction information. Specifically, there is a standard brain region template (registration template image), and the registration matrix is multiplied by the template to obtain a brain region mask (brain region map) under the brain space of the user individual. Based on the brain region mask and DWI image focus mask, the number of focus voxels under each brain region can be calculated, which can be approximated as the volume of the focus, thereby obtaining a focus volume feature with a size of 1×16. Wherein the selected brain region map comprises island leaf regions.
Baseline clinical data included age information, gender information, past history information (hypertension, diabetes, lipoprotein metabolic disorders, current smoking, current alcohol consumption, coronary atherosclerotic heart disease, heart failure, myocardial infarction, peripheral arterial disease, stroke, transient ischemic attacks) and NIHSS score information for the user. Before being input into a trained prediction model of risk of post-stroke atrial fibrillation, the baseline clinical data needs to be preprocessed to obtain preprocessed baseline clinical data. Specifically, the age information of the user is subjected to Z-score standardization to obtain age characteristics, wherein the size is 1*1; performing One-hot encoding on the gender information of the user to obtain gender characteristics, wherein the size is 1*2; performing One-hot coding on the past history information of the user to obtain past history characteristics, wherein the size of the past history characteristics is 1 x 11; the NIHSS score was Z-score normalized to yield a NIHSS score feature, size 1*1. Wherein the size of the dimension is the size of each feature vector and represents the length of the feature vector. For example, after the sex information is encoded by One-hot, [0,1] represents the sex of the user as female, and [1,0] represents the sex of the user as male, the length of the obtained sex characteristic is 2. The length of the past history feature is 11.
The preprocessed risk prediction information can be obtained through the preprocessing, and comprises preprocessed multi-mode images, preprocessed baseline clinical data information, focus masks and focus volume characteristics. And the preprocessed risk prediction information is used as the input of a trained risk prediction model for the post-stroke atrial fibrillation, so that the risk prediction probability of the post-stroke atrial fibrillation can be predicted.
In this embodiment, the trained risk prediction model for post-stroke atrial fibrillation includes five sub-models; the five sub-models are all models which are obtained by taking sample risk prediction information of a sample user as input and taking whether the sample user has a post-stroke atrial fibrillation as a label training. Inputting the preprocessed risk prediction information into a trained post-stroke atrial fibrillation risk prediction model, predicting the probability of occurrence of post-stroke atrial fibrillation of the user, and obtaining post-stroke atrial fibrillation risk prediction probability, wherein the method specifically comprises the following steps of:
and respectively inputting the preprocessed risk prediction information into five sub-models to obtain five risk prediction sub-probabilities of the post-stroke atrial fibrillation.
And calculating the average value of the five risk predictors of the post-stroke atrial fibrillation to obtain the risk predictors of the post-stroke atrial fibrillation.
The structure of the sub-model is shown in fig. 2, and the sub-model comprises an image feature extraction module, a compression excitation module, a data feature extraction module, a splicing module and a prediction module.
The image feature extraction module is used for extracting features of the preprocessed multi-mode images and the focus mask to obtain image feature vectors.
The compression excitation module is used for carrying out global pooling on the image feature vectors, processing the image feature vectors subjected to global pooling by utilizing an activation function to obtain weight vectors, and multiplying the weight vectors and the image feature vectors in channel dimensions to obtain channel multiplication feature vectors.
The data feature extraction module is used for carrying out feature extraction on the preprocessed baseline clinical data information and the focus volume feature to obtain a data feature vector.
And the splicing module is used for splicing the channel product feature vector and the data feature vector to obtain fusion features.
And the prediction module is used for obtaining the risk predictor probability of the post-stroke atrial fibrillation according to the fusion characteristics.
The image feature extraction module comprises a plurality of residual error units; each residual unit comprises a plurality of convolution layers which are connected in sequence.
The compression excitation module comprises a global pooling layer, a ReLU layer and a Sigmoid layer which are sequentially connected.
The data characteristic extraction module comprises five data characteristic extraction sub-modules; each data characteristic extraction submodule comprises a first full-connection layer, a ReLu layer and a second full-connection layer which are sequentially connected.
The prediction module comprises a third full-connection layer, a first ReLu layer, a fourth full-connection layer, a second ReLu layer and a fifth full-connection layer which are sequentially connected.
Before the preprocessed risk prediction information is input into a trained risk prediction model of the post-stroke atrial fibrillation, training of the sub-model is needed, and the training process is as follows:
the data set is divided into a training set and a test set according to a ratio of 7:3. The training set is divided into a number of negative data sets (the negative data sets refer to the information for predicting the risk of the preprocessed sample of the patient without the occurrence of the post-stroke atrial fibrillation, corresponding to the label 0) in equal proportion according to the number of positive samples (the positive samples are screened according to CRF, and are known as the information for predicting the risk of the preprocessed sample of the post-stroke atrial fibrillation patient, corresponding to the label 1) to obtain a plurality of groups of sub-data sets A, B, C, D, E. Each sub-dataset was partitioned into a sub-training set and a sub-validation set according to a 9:1 ratio. The five sub-models model_A, model_B, model_C, model_D, model_E are trained separately using a sub-training set, and the training process uses a cross entropy loss function and an Adadelta optimizer to update the model parameters. The label of the sub model is a binary value, 0 represents that no post-stroke atrial fibrillation occurs, and 1 represents that post-stroke atrial fibrillation occurs. It is obtained through a Case Report Form (CRF) corresponding to the data set, for example, if the sample user has atrial fibrillation, the case report form will show that the corresponding data is then marked as 1.
For each sub-model: the structure is as shown in fig. 2, and the input of the submodel is composed of two branches, namely an image branch (corresponding to an image feature extraction module) and a data branch (corresponding to a data feature extraction module). The image feature extraction module is composed of a residual network 3 dresent 34, the input of which is a preprocessed multi-mode image and focus mask, the size of which is 6×32×256×256 (6 is the number of channels of input data, 6 channels are respectively a preprocessed T1 image, a preprocessed T2 image, a preprocessed FLAIR image, a preprocessed DWI image, a FLAIR image focus mask and a DWI image focus mask, and 32×256×256 are respectively the depth, the length and the width of the image), and the image feature vector with the length of 128 is output through global average pooling (GlobalAverage Pooling). And the compression excitation module (SEBlock) performs global pooling (GlobalPooling) on the input image feature vector to obtain a feature vector of 1 x C, wherein C is the channel number, then obtains a weight vector through a ReLU function and Sigmoid, and multiplies the weight vector by the input image feature vector in the channel dimension to be used as the output (channel multiplication feature vector) of SEBlock.
Data branches include age characteristic branches, gender characteristic branches, past history characteristic branches, NIHSS scoring characteristic branches, and lesion volume characteristic branches. The input of the age characteristic branch is an age characteristic with the size of 1*1, and an embedding vector with the size of 1 x 32 is obtained through a ProjectionHead1 network. The input of the gender feature branch is gender feature with the size of 1*2, and the embedded vector with the size of 1 x 32 is obtained through the ProjectionHead2 network. The input of the past history feature branch is the past history feature with the size of 1 x 11, and the embedded vector with the size of 1 x 32 is obtained through the ProjectionHead3 network. The input of the NIHSS scoring feature branch is a NIHSS scoring feature of 1*1, and the NIHSS scoring feature branch passes through a ProjectionHead4 network to obtain a feature vector of 1×32. The input of the lesion volume feature branch is a lesion volume feature with the size of 1×116, and a feature vector with the size of 1×32 is obtained through a ProjectionHead5 network. The vectors output by the ProjectionHead networks form data feature vectors.
Splicing the channel product feature vector and the data feature vector through concat to obtain a fusion feature with the size of 1 x 288, inputting the fusion feature into a ProjectionHead6 network (prediction module) to obtain an embedded vector with the size of 1 x 128, and inputting the embedded vector into a third full-connection layer of the ProjectionHead6 network to obtain a stroke posterior atrial fibrillation risk predictor probability value with the size of 1*1. And carrying out integrated prediction by utilizing the five sub-models, namely, respectively and independently predicting the five sub-models to obtain five predicted sub-probability values of the risk of the atrial fibrillation after the apoplexy, wherein the size of the predicted sub-probability values is 1*1, calculating the average value of the five probability values to obtain predicted probability of the risk of the atrial fibrillation after the apoplexy, and comparing the predicted probability of the risk of the atrial fibrillation after the apoplexy with the predicted probability value to obtain the risk level of the atrial fibrillation after the apoplexy. In this embodiment, the predicted probability value is selected to be 0.5, if the predicted probability of the risk of the atrial fibrillation after the stroke is greater than 0.5, the risk of the atrial fibrillation after the stroke is high, otherwise, the risk of the atrial fibrillation after the stroke is low.
As shown in fig. 3, the above-described structure of the ProjectionHead1-ProjectionHead5 network includes two full connection layers (FullConnection, FC) and a ReLU activation function (ReLU layer), and the first full connection layer, the ReLU layer, and the second full connection layer in each ProjectionHead network are sequentially connected. Specifically, the input of the first full connection layer of the ProjectionHead1 network is set to 1 node, the output is set to 32 nodes, then the input of the second full connection layer is set to 32 nodes through the ReLu activation function (ReLu layer), and the output is set to 32 nodes. The input of the first full connection layer of the ProjectionHead2 network is set to 2 nodes, the output is set to 32 nodes, then the input of the second full connection layer is set to 32 nodes through the ReLu activation function, and the output is set to 32 nodes. The input of the first full connection layer of the ProjectionHead3 network is set to 11 nodes, the output is set to 32 nodes, then the input of the second full connection layer is set to 32 nodes through the ReLu activation function, and the output is set to 32 nodes. The input of the first full connection layer of the ProjectionHead4 network is set to 1 node, the output is set to 32 nodes, then the input of the second full connection layer is set to 32 nodes through the ReLu activation function, and the output is set to 32 nodes. The input of the first full connection layer of the ProjectionHead5 network is set to 16 nodes, the output is set to 32 nodes, then the input of the second full connection layer is set to 32 nodes through the ReLu activation function, and the output is set to 32 nodes. The input of the first full connection layer (third full connection layer) of the ProjectionHead6 network is set to 288 nodes, the output is set to 128 nodes, the input of the second full connection layer (fourth full connection layer) is set to 128 nodes after being processed by the ReLu activation function (first ReLu layer), the output is set to 64 nodes, then the input of the third full connection layer (fifth full connection layer) is set to 64 nodes after being processed again by the ReLu activation function (second ReLu layer), and the output is set to 1 node.
The invention utilizes a plurality of clinically collected sequence image data and baseline clinical information to carry out early evaluation on the risk of post-stroke atrial fibrillation of an acute ischemic stroke patient; in the real world, patients in the acute phase are difficult to effectively evaluate the risk of atrial fibrillation after stroke depending on a small amount of fragmentation information before multi-dimensional clinical information is acquired, so that suitable people for long-range electrocardiographic monitoring cannot be screened out in an early and efficient mode. Therefore, in the acute phase of the apoplexy, the invention applies the homogeneous imaging data and demographic characteristics which are easy to acquire in early stage, based on the baseline clinical information and nuclear magnetic resonance examination information which can be acquired in initial diagnosis by a deep learning technical method, fully utilizes a residual network (an image characteristic extraction module) to extract the high-dimensional characteristic of nuclear magnetic resonance and utilizes a ProjectionHead network to extract the baseline clinical information characteristic and focus volume characteristic, comprehensively predicts the onset risk of atrial fibrillation after the apoplexy, and has important significance for identifying the high risk group of atrial fibrillation after the ischemic apoplexy as soon as possible, searching the target group suitable for long-range electrocardiographic monitoring and further providing a more accurate and personalized treatment scheme for the patient; the invention also adds the accumulation condition of the stroke focus in different cortex areas to the deep learning network as a characteristic (namely, adopts the focus volume characteristic as the input of a stroke posterior atrial fibrillation risk prediction model).
The invention also provides a computer device comprising a memory and a processor, the memory storing a computer program which, when executed by the processor, causes the processor to perform the steps of the method for predicting risk of atrial fibrillation after stroke.
Referring to fig. 4, fig. 4 is a schematic structural diagram of a computer device provided in the present application. As shown in fig. 4, the computer device 1000 may include: processor 1001, network interface 1004, and memory 1005, in addition, computer device 1000 may further comprise: a user interface 1003, and at least one communication bus 1002. Wherein the communication bus 1002 is used to enable connected communication between these components. The user interface 1003 may include a Display (Display), a Keyboard (Keyboard), and the optional user interface 1003 may further include a standard wired interface, a wireless interface, among others. The network interface 1004 may optionally include a standard wired interface, a wireless interface (e.g., WI-FI interface). The memory 1005 may be a high-speed RAM memory or a non-volatile memory (non-volatile memory), such as at least one disk memory. The memory 1005 may also optionally be at least one storage device located remotely from the processor 1001. As shown in fig. 4, an operating system, a network communication module, a user interface module, and a device control application program may be included in the memory 1005, which is one type of computer storage medium.
In the computer device 1000 shown in FIG. 4, the network interface 1004 may provide network communication functions; while user interface 1003 is primarily used as an interface for providing input to a user; the processor 1001 may be configured to invoke the device control application stored in the memory 1005 to implement the method for predicting risk of atrial fibrillation according to the above embodiment, which will not be described herein.
The present invention also provides a computer readable storage medium storing a computer program adapted to be loaded by a processor and to execute the method for predicting risk of atrial fibrillation after stroke according to the above embodiment, which will not be described in detail herein.
The above-described program may be deployed to be executed on one computer device or on multiple computer devices that are deployed at one site or on multiple computer devices that are distributed across multiple sites and interconnected by a communication network, and the multiple computer devices that are distributed across multiple sites and interconnected by a communication network may constitute a blockchain network.
The computer readable storage medium may be an internal storage unit of the computer device, such as a hard disk or a memory of the computer device. The computer readable storage medium may also be an external storage device of the computer device, such as a plug-in hard disk, a smart card (SMC), a Secure Digital (SD) card, a flash memory card (flashcard), etc. which are provided on the computer device. Further, the computer-readable storage medium may also include both internal storage units and external storage devices of the computer device. The computer-readable storage medium is used to store the computer program and other programs and data required by the computer device. The computer-readable storage medium may also be used to temporarily store data that has been output or is to be output.
In the present specification, each embodiment is described in a progressive manner, and each embodiment is mainly described in a different point from other embodiments, and identical and similar parts between the embodiments are all enough to refer to each other. For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant points refer to the description of the method section.
The principles and embodiments of the present invention have been described herein with reference to specific examples, the description of which is intended only to assist in understanding the methods of the present invention and the core ideas thereof; also, it is within the scope of the present invention to be modified by those of ordinary skill in the art in light of the present teachings. In view of the foregoing, this description should not be construed as limiting the invention.
Claims (9)
1. A method of predicting risk of atrial fibrillation after a stroke, the method comprising:
acquiring risk prediction information of a user; the risk prediction information comprises multi-modal images and baseline clinical data information; the multi-mode image comprises a T1 image, a T2 image, a FLAIR image and a DWI image; the baseline clinical data information includes age information, gender information, past history information, and NIHSS scoring information;
preprocessing the risk prediction information to obtain preprocessed risk prediction information; the preprocessed risk prediction information comprises preprocessed multi-mode images, preprocessed baseline clinical data information, focus masks and focus volume characteristics;
inputting the preprocessed risk prediction information into a trained post-stroke atrial fibrillation risk prediction model, and predicting the probability of occurrence of post-stroke atrial fibrillation of the user to obtain a post-stroke atrial fibrillation risk prediction probability;
the trained risk prediction model for the post-stroke atrial fibrillation comprises five sub-models; each sub-model comprises an image feature extraction module, a compression excitation module, a data feature extraction module, a splicing module and a prediction module;
the image feature extraction module is used for extracting features of the preprocessed multi-mode image and the focus mask to obtain an image feature vector;
the compression excitation module is used for carrying out global pooling on the image feature vectors, processing the image feature vectors subjected to global pooling by utilizing an activation function to obtain weight vectors, and multiplying the weight vectors and the image feature vectors in channel dimensions to obtain channel multiplication feature vectors;
the data feature extraction module is used for carrying out feature extraction on the preprocessed baseline clinical data information and the focus volume feature to obtain a data feature vector;
the splicing module is used for splicing the channel product feature vector and the data feature vector to obtain fusion features;
the prediction module is used for obtaining a risk predictor probability of the post-stroke atrial fibrillation according to the fusion characteristics;
and comparing the predicted probability of the risk of the post-stroke atrial fibrillation with a preset probability value to obtain the risk level of the post-stroke atrial fibrillation of the user.
2. The method for predicting risk of post-stroke atrial fibrillation according to claim 1, wherein the preprocessing the risk prediction information specifically comprises:
carrying out structural image registration on the multi-modal image and the registration template image to obtain a registered multi-modal image and a registration matrix;
performing brain extraction on the registered multimodal images to obtain brain extraction results; clipping the registered multi-modal image according to the brain extraction result to obtain a clipped multi-modal image; the cut multi-mode image is the preprocessed multi-mode image and comprises a cut T1 image, a cut T2 image, a cut FLAIR image and a cut DWI image;
dividing the stroke focus of the cut FLAIR image to obtain a focus mask of the FLAIR image; dividing the stroke focus of the cut DWI image to obtain a DWI image focus mask; the DWI image focus mask and the FLAIR image focus mask form the focus mask;
determining focal volume characteristics according to the DWI image focal mask;
and carrying out coding pretreatment on the baseline clinical data information to obtain pretreated baseline clinical data information.
3. The method of predicting risk of post-stroke atrial fibrillation according to claim 2, characterized in that said segmentation comprises in particular:
the segmentation is performed by using a manual outlining method or a deep learning image segmentation model.
4. The method for predicting risk of post-stroke atrial fibrillation according to claim 2, wherein said determining lesion volume features from said DWI image lesion mask specifically comprises:
calculating a brain region map of the user using the registration matrix and the registration template image;
and obtaining the number of focus voxels of the stroke according to the brain region map and the DWI image focus mask, wherein the number of focus voxels of the stroke is the focus volume characteristic.
5. The method of predicting risk of post-stroke atrial fibrillation according to claim 2, wherein said pre-coding the baseline clinical data information comprises:
and performing Z-score standardization on the age information, performing One-hot encoding on the gender information, performing One-hot encoding on the past history information, and performing Z-score standardization on the NIHSS scoring information.
6. The method for predicting risk of post-stroke atrial fibrillation according to claim 1, wherein the five sub-models are models obtained by taking information for predicting sample risk of a sample user as input and taking whether the sample user has suffered post-stroke atrial fibrillation as label training; the pre-processed risk prediction information is input into a trained post-stroke atrial fibrillation risk prediction model, the probability of occurrence of post-stroke atrial fibrillation of the user is predicted, and the post-stroke atrial fibrillation risk prediction probability is obtained, and specifically comprises the following steps:
the preprocessed risk prediction information is respectively input into five sub-models to obtain five risk prediction sub-probabilities of the post-stroke atrial fibrillation;
and calculating the average value of the five post-stroke atrial fibrillation risk predictors to obtain the post-stroke atrial fibrillation risk predictors.
7. The method of claim 1, wherein the risk prediction of post-stroke atrial fibrillation,
the image feature extraction module comprises a plurality of residual error units; each residual error unit comprises a plurality of convolution layers which are connected in sequence;
the compression excitation module comprises a global pooling layer, a ReLU layer and a Sigmoid layer which are sequentially connected;
the data characteristic extraction module comprises five data characteristic extraction sub-modules; each data characteristic extraction submodule comprises a first full-connection layer, a ReLu layer and a second full-connection layer which are sequentially connected;
the prediction module comprises a third full-connection layer, a first ReLu layer, a fourth full-connection layer, a second ReLu layer and a fifth full-connection layer which are sequentially connected.
8. A computer device comprising a memory and a processor, the memory storing a computer program that, when executed by the processor, causes the processor to perform the steps of the method of any of claims 1-7.
9. A computer readable storage medium, characterized in that the computer readable storage medium stores a computer program adapted to be loaded by a processor and to perform the method of any of claims 1-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310263744.1A CN116313089B (en) | 2023-03-13 | 2023-03-13 | Method for predicting risk of atrial fibrillation after stroke, computer equipment and medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310263744.1A CN116313089B (en) | 2023-03-13 | 2023-03-13 | Method for predicting risk of atrial fibrillation after stroke, computer equipment and medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116313089A CN116313089A (en) | 2023-06-23 |
| CN116313089B true CN116313089B (en) | 2024-01-16 |
Family
ID=86779349
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310263744.1A Active CN116313089B (en) | 2023-03-13 | 2023-03-13 | Method for predicting risk of atrial fibrillation after stroke, computer equipment and medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116313089B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117953027B (en) * | 2024-03-22 | 2024-06-25 | 首都医科大学附属北京天坛医院 | DWI-FLAIR mismatch evaluation method, device, medium and product |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107145756A (en) * | 2017-05-17 | 2017-09-08 | 上海辉明软件有限公司 | A kind of stroke types Forecasting Methodology and device |
| CA3160827A1 (en) * | 2018-09-24 | 2019-07-04 | Amarin Pharmaceuticals Ireland Limited | Methods of reducing the risk of cardiovascular events in a subject |
| CN110840402A (en) * | 2019-11-19 | 2020-02-28 | 山东大学 | Atrial fibrillation signal identification method and system based on machine learning |
| CN111091911A (en) * | 2019-12-30 | 2020-05-01 | 重庆同仁至诚智慧医疗科技股份有限公司 | System and method for screening stroke risk |
| KR102189622B1 (en) * | 2020-10-06 | 2020-12-11 | 주식회사 휴런 | STROKE DIAGNOSIS APPARATUS BASED ON AI(Artificial Intelligence) SYSTEM THAT CATEGORIZES CEREBRAL HEMORRHAGE STATUS AND TYPE |
| KR102219202B1 (en) * | 2020-10-06 | 2021-02-23 | 주식회사 휴런 | How to Provide Artificial Intelligence-Based Stroke Diagnostic Information |
| KR20210065768A (en) * | 2019-11-27 | 2021-06-04 | 연세대학교 산학협력단 | System for Processing Medical Image and Clinical Factor for Individualized Diagnosis of Stroke |
| CN113380413A (en) * | 2021-06-24 | 2021-09-10 | 首都医科大学附属北京天坛医院 | Method and device for constructing invalid re-circulation (FR) prediction model |
| CN114121274A (en) * | 2021-11-01 | 2022-03-01 | 上海理工大学 | Atrial fibrillation-related stroke risk assessment method, device and storage medium |
| CN115148363A (en) * | 2022-07-28 | 2022-10-04 | 华中科技大学同济医学院附属同济医院 | Method, device, medium and electronic equipment for predicting depression risk of patient after stroke |
| CN115736939A (en) * | 2022-11-24 | 2023-03-07 | 北京哈特瑞姆医疗技术服务有限公司 | Atrial fibrillation disease probability generation method and device, electronic equipment and storage medium |
Family Cites Families (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9913587B2 (en) * | 2013-11-01 | 2018-03-13 | Cardiio, Inc. | Method and system for screening of atrial fibrillation |
| KR102001398B1 (en) * | 2018-01-25 | 2019-07-18 | 재단법인 아산사회복지재단 | Method and apparatus for predicting brain desease change though machine learning and program for the same |
| US11540796B2 (en) * | 2018-05-23 | 2023-01-03 | Case Western Reserve University | Prediction of risk of post-ablation atrial fibrillation based on radiographic features of pulmonary vein morphology from chest imaging |
| JP2022523835A (en) * | 2019-09-06 | 2022-04-26 | 中國醫藥大學附設醫院 | Atrial fibrillation prediction model and its prediction system |
| US20210076960A1 (en) * | 2019-09-18 | 2021-03-18 | Tempus Labs | Ecg based future atrial fibrillation predictor systems and methods |
| US20220370017A1 (en) * | 2021-05-14 | 2022-11-24 | University Of Cincinnati | Personalized prediction and identification of the incidence of atrial arrhythmias from other cardiac rhythms |
-
2023
- 2023-03-13 CN CN202310263744.1A patent/CN116313089B/en active Active
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107145756A (en) * | 2017-05-17 | 2017-09-08 | 上海辉明软件有限公司 | A kind of stroke types Forecasting Methodology and device |
| CA3160827A1 (en) * | 2018-09-24 | 2019-07-04 | Amarin Pharmaceuticals Ireland Limited | Methods of reducing the risk of cardiovascular events in a subject |
| CN110840402A (en) * | 2019-11-19 | 2020-02-28 | 山东大学 | Atrial fibrillation signal identification method and system based on machine learning |
| KR20210065768A (en) * | 2019-11-27 | 2021-06-04 | 연세대학교 산학협력단 | System for Processing Medical Image and Clinical Factor for Individualized Diagnosis of Stroke |
| CN111091911A (en) * | 2019-12-30 | 2020-05-01 | 重庆同仁至诚智慧医疗科技股份有限公司 | System and method for screening stroke risk |
| KR102189622B1 (en) * | 2020-10-06 | 2020-12-11 | 주식회사 휴런 | STROKE DIAGNOSIS APPARATUS BASED ON AI(Artificial Intelligence) SYSTEM THAT CATEGORIZES CEREBRAL HEMORRHAGE STATUS AND TYPE |
| KR102219202B1 (en) * | 2020-10-06 | 2021-02-23 | 주식회사 휴런 | How to Provide Artificial Intelligence-Based Stroke Diagnostic Information |
| CN113380413A (en) * | 2021-06-24 | 2021-09-10 | 首都医科大学附属北京天坛医院 | Method and device for constructing invalid re-circulation (FR) prediction model |
| CN114121274A (en) * | 2021-11-01 | 2022-03-01 | 上海理工大学 | Atrial fibrillation-related stroke risk assessment method, device and storage medium |
| CN115148363A (en) * | 2022-07-28 | 2022-10-04 | 华中科技大学同济医学院附属同济医院 | Method, device, medium and electronic equipment for predicting depression risk of patient after stroke |
| CN115736939A (en) * | 2022-11-24 | 2023-03-07 | 北京哈特瑞姆医疗技术服务有限公司 | Atrial fibrillation disease probability generation method and device, electronic equipment and storage medium |
Non-Patent Citations (1)
| Title |
|---|
| 缺血性脑卒中患者心房纤颤诊断的研究进展;丁也;郭阳;;中国全科医学(17);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116313089A (en) | 2023-06-23 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113314205B (en) | Efficient medical image labeling and learning system | |
| Xu et al. | Texture-specific bag of visual words model and spatial cone matching-based method for the retrieval of focal liver lesions using multiphase contrast-enhanced CT images | |
| CN109544518B (en) | Method and system applied to bone maturity assessment | |
| CN114581662B (en) | Brain tumor image segmentation method, system, device and storage medium | |
| US20220254022A1 (en) | Method and system for automatic multiple lesion annotation of medical images | |
| CN111951281B (en) | Image segmentation method, device, equipment and storage medium | |
| CN111951280A (en) | Image segmentation method, device, equipment and storage medium | |
| CN116313089B (en) | Method for predicting risk of atrial fibrillation after stroke, computer equipment and medium | |
| CN111967495A (en) | Classification recognition model construction method | |
| CN113127661B (en) | Multi-supervision medical image retrieval method and system based on cyclic query expansion | |
| CN111863247B (en) | Brain age cascade refining prediction method and system based on structural magnetic resonance image | |
| CN113706442A (en) | Medical image processing method and device based on artificial intelligence and electronic equipment | |
| CN112597884B (en) | Training method of gesture recognition model, gesture recognition method and system | |
| CN113705595A (en) | Method, device and storage medium for predicting degree of abnormal cell metastasis | |
| CN111612739A (en) | Deep learning-based cerebral infarction classification method | |
| CN115984257A (en) | Multi-modal medical image fusion method based on multi-scale transform | |
| CN116503684A (en) | Model training method and device, electronic equipment and storage medium | |
| CN113177923B (en) | Medical image content identification method, electronic equipment and storage medium | |
| JP2015510408A (en) | Bioimaging grid | |
| Lin et al. | MCI Conversion Prediction Based on Transfer Learning | |
| CN114937171B (en) | Deep learning-based Alzheimer's classification system | |
| CN114445643B (en) | Fmri brain function connection data feature extraction method based on significant sparse strong correlation | |
| CN116934741B (en) | Method and device for acquiring composition and quantitative parameters of one-stop type blood vessel wall | |
| Bueno Gómez et al. | Automatic MR Spinal Cord Segmentation by Hybrid Residual Attention-Aware Convolutional Neural Networks and Learning Rate Optimization on Real World Data | |
| Kaur et al. | A New Deep Learning Framework for Accurate Intracranial Brain Hemorrhage Detection and Classification Using Real-Time Collected NCCT Images |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |