CN111402218A - Cerebral hemorrhage detection method and device - Google Patents

Cerebral hemorrhage detection method and device Download PDF

Info

Publication number
CN111402218A
CN111402218A CN202010164910.9A CN202010164910A CN111402218A CN 111402218 A CN111402218 A CN 111402218A CN 202010164910 A CN202010164910 A CN 202010164910A CN 111402218 A CN111402218 A CN 111402218A
Authority
CN
China
Prior art keywords
cerebral hemorrhage
brain
detection model
bleeding
image
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.)
Pending
Application number
CN202010164910.9A
Other languages
Chinese (zh)
Inventor
梁孔明
韩凯
俞益洲
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Beijing Shenrui Bolian Technology Co Ltd
Shenzhen Deepwise Bolian Technology Co Ltd
Original Assignee
Beijing Shenrui Bolian Technology Co Ltd
Shenzhen Deepwise Bolian Technology Co Ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Beijing Shenrui Bolian Technology Co Ltd, Shenzhen Deepwise Bolian Technology Co Ltd filed Critical Beijing Shenrui Bolian Technology Co Ltd
Priority to CN202010164910.9A priority Critical patent/CN111402218A/en
Publication of CN111402218A publication Critical patent/CN111402218A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • 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
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/044—Recurrent networks, e.g. Hopfield networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/047—Probabilistic or stochastic networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/60—Analysis of geometric attributes
    • G06T7/62—Analysis of geometric attributes of area, perimeter, diameter or volume
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/70—Determining position or orientation of objects or cameras
    • G06T7/73—Determining position or orientation of objects or cameras 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
    • G06T2207/10—Image acquisition modality
    • G06T2207/10072—Tomographic images
    • G06T2207/10081—Computed x-ray tomography [CT]
    • 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
    • G06T2207/20076—Probabilistic image processing
    • 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
    • G06T2207/20081—Training; Learning
    • 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
    • G06T2207/20084—Artificial neural networks [ANN]
    • 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
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30004—Biomedical image processing
    • G06T2207/30016—Brain

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • General Engineering & Computer Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • Health & Medical Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Mathematical Physics (AREA)
  • Computing Systems (AREA)
  • Molecular Biology (AREA)
  • Computational Linguistics (AREA)
  • Biomedical Technology (AREA)
  • Software Systems (AREA)
  • Biophysics (AREA)
  • Probability & Statistics with Applications (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Evolutionary Biology (AREA)
  • Geometry (AREA)
  • Medical Informatics (AREA)
  • Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
  • Radiology & Medical Imaging (AREA)
  • Quality & Reliability (AREA)
  • Apparatus For Radiation Diagnosis (AREA)

Abstract

The invention provides a cerebral hemorrhage detection method and a device, wherein the method comprises the following steps: acquiring a plurality of confirmed cerebral hemorrhage brain CT images; constructing a cerebral hemorrhage detection model based on a 3D segmentation network; training a cerebral hemorrhage detection model through a cerebral hemorrhage brain CT image; inputting a brain CT image to be detected, predicting bleeding probability of each voxel point of the brain CT image to be detected through the trained brain bleeding detection model, and realizing detection, positioning and volume measurement of the brain bleeding according to a prediction result. The invention can automatically detect, position and measure the volume of the cerebral hemorrhage, and has higher efficiency and accuracy.

Description

Cerebral hemorrhage detection method and device
Technical Field
The invention relates to the technical field of image processing and analysis, in particular to a cerebral hemorrhage detection method and a cerebral hemorrhage detection device.
Background
Cerebral hemorrhage (including hemorrhagic stroke and traumatic brain injury) is the main cause of disability and death of adults, and seriously harms the health of people. Patients with intracranial haemorrhages, often accompanied by severe neurological symptoms such as severe headache or loss of consciousness, require rapid and intensive medical diagnosis and treatment, wherein identification of the location, type and volume of the haemorrhage is a critical step in the treatment of the patient. The craniocerebral CT scanning is the primary examination method for judging the cerebral hemorrhage diseases, but even for trained experts, the diagnosis of the cerebral hemorrhage through the craniocerebral CT is complex and time-consuming.
Disclosure of Invention
The invention provides a method and a device for detecting cerebral hemorrhage, which can automatically detect, locate and measure the cerebral hemorrhage and have higher efficiency and accuracy.
The technical scheme adopted by the invention is as follows:
a cerebral hemorrhage detection method comprises the following steps: acquiring a plurality of confirmed cerebral hemorrhage brain CT images; constructing a cerebral hemorrhage detection model based on a 3D segmentation network; training the cerebral hemorrhage detection model through the cerebral hemorrhage brain CT image; inputting a brain CT image to be detected, predicting bleeding probability of each voxel point of the brain CT image to be detected through a trained cerebral bleeding detection model, and realizing detection, positioning and volume measurement of cerebral bleeding according to a prediction result.
After acquiring a plurality of CT images of the cerebral hemorrhage brain, the method further comprises the following steps: and carrying out standardization processing on the plurality of diagnosed cerebral hemorrhage brain CT images.
The cerebral hemorrhage detection model takes U-Net as a backbone network.
Wherein the cerebral hemorrhage detection model is constructed at multiple scales based on a 3D segmentation network.
The cerebral hemorrhage detection model comprises a plurality of basic modules and a global module.
Wherein the base module includes a 3D convolutional layer, a nonlinear activation unit, and an element addition layer.
Wherein the global module comprises a 3D convolutional layer, a matrix multiplication layer, an element addition layer and a Softmax activation unit.
Training the cerebral hemorrhage detection model by using cross entropy loss, multi-layer supervision loss and weighted loss.
A cerebral hemorrhage detecting device comprising: an acquisition module for acquiring a plurality of diagnosed cerebral hemorrhage brain CT images; a modeling module for constructing a cerebral hemorrhage detection model based on a 3D segmentation network; a training module for training the cerebral hemorrhage detection model through the cerebral hemorrhage brain CT image; and the detection module is used for predicting the bleeding probability of each voxel point of the input brain CT image to be detected through the trained brain bleeding detection model and realizing the detection, positioning and volume measurement of the brain bleeding according to the prediction result.
The cerebral hemorrhage detecting device further comprises: and the processing module is used for carrying out standardized processing on the plurality of diagnosed cerebral hemorrhage brain CT images.
The invention has the beneficial effects that:
the cerebral hemorrhage detection model is constructed based on the 3D segmentation network, and after the cerebral hemorrhage detection model is trained, the hemorrhage probability of each voxel point of the CT image of the brain to be detected is predicted through the trained cerebral hemorrhage detection model, so that the cerebral hemorrhage detection, positioning and volume measurement can be automatically performed through voxel-level analysis, and the cerebral hemorrhage detection method and the cerebral hemorrhage detection system have high efficiency and accuracy.
Drawings
FIG. 1 is a flow chart of a method for detecting cerebral hemorrhage according to an embodiment of the present invention;
FIG. 2 is a schematic overall structure diagram of a model for detecting cerebral hemorrhage according to an embodiment of the present invention;
FIG. 3 is a schematic structural diagram of a basic module in a cerebral hemorrhage detection model according to an embodiment of the present invention;
FIG. 4 is a schematic structural diagram of a global module in a cerebral hemorrhage detection model according to an embodiment of the present invention;
FIG. 5 is a diagram illustrating the visualization effect of the detection of cerebral hemorrhage according to one embodiment of the present invention;
fig. 6 is a block diagram of a cerebral hemorrhage detecting device according to an embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
As shown in fig. 1, the method for detecting cerebral hemorrhage according to the embodiment of the present invention includes the following steps:
and S1, acquiring a plurality of confirmed cerebral hemorrhage brain CT images.
In one embodiment of the present invention, a case in which a CT image is diagnosed as a patient with cerebral hemorrhage and clinical diagnosis of the case is consistent with the CT image diagnosis can be selected, CT image data of the brain of the patient can be obtained by PACS (Picture Archiving and Communication Systems), for example, a CT machine of a brand of 120kV, such as siemens 16 or 32, philips, etc., the data format is consistent with the Digital Imaging and communications in Medicine (DICOM, Digital Imaging and communications in Medicine) standard, the scanning layer thickness is 5mm, the examination method is a supine position, and the scanning range is skull base to skull top. The brain CT image data includes five types of parenchymal hemorrhage, ventricular hemorrhage, epidural hemorrhage, subdural hemorrhage, and subarachnoid hemorrhage. After the CT image of the brain of the patient is acquired, artificial labeling can be carried out, and the labeling process of all bleeding cases can comprise two stages. The first stage, marking the outline, category and position of the focus by a doctor according to the diagnosis report; and in the second stage, a subsidiary chief physician checks the focus, marks the missed focus, deletes the mis-marked focus, and corrects the contour, category and position of the focus. The label checked by senior medical doctors can be used as a gold standard for cerebral hemorrhage detection, positioning and volume measurement, and is used for training and result evaluation of the following models. In one embodiment of the present invention, the annotation data can be divided into a training set for training and optimal parameter selection of the model described below and a test set for evaluating the performance of the cerebral hemorrhage detection method of the embodiments of the present invention.
In an embodiment of the present invention, after acquiring a plurality of diagnosed cerebral hemorrhage brain CT images, the plurality of diagnosed cerebral hemorrhage brain CT images may be further standardized. Specifically, the raw CT data of the 5mm thick layer may be automatically normalized, including performing gray level normalization, three-dimensional correction transformation, and the like by using the window width and window level.
And S2, constructing a cerebral hemorrhage detection model based on the 3D segmentation network.
Compared with a two-dimensional image, the brain CT image is three-dimensional data containing spatial domain changes, and the information of the upper layer and the lower layer can effectively avoid misdiagnosis of artifacts and calcifications generated by a scanning device as bleeding, so that the embodiment of the invention adopts a 3D full-volume deep segmentation network to better model the context information of the tomography CT.
The cerebral hemorrhage detection model of the embodiment of the invention takes U-Net as a backbone network, the overall structure of the model is shown in fig. 2 and comprises a plurality of basic modules and a global module, wherein as shown in fig. 3, the basic modules comprise a 3D convolutional layer CONV, a nonlinear activation unit SFM and an element addition layer ADD, as shown in fig. 4, the global module comprises a 3D convolutional layer CONV, a matrix multiplication layer MU L, an element addition layer ADD and a Softmax activation unit SFM.
The network consists of two paths of coding and decoding, wherein the coding path inputs the characteristics of each scale into a module B L OCK, and extracts corresponding high-level semantics through pooling layers or step convolution.
According to the embodiment of the invention, the context capture capability of the area to be predicted can be effectively enhanced by modeling the relation between different position characteristics.
And S3, training a cerebral hemorrhage detection model through the cerebral hemorrhage brain CT image.
Specifically, the training set obtained in step S1 may be input into the cerebral hemorrhage detection model constructed in step S2, and the cerebral hemorrhage detection model may be trained. The loss function used for training also introduces multilayer supervision loss and weighting loss on the basis of cross entropy loss, and ensures that the model can more quickly converge and learn the input training sample more effectively. In one embodiment of the present invention, the learning rate in the hyper-parameters associated with model training is 0.001, and the epoch number of model iterations is 100.
And S4, inputting the brain CT image to be detected, predicting the bleeding probability of each voxel point of the brain CT image to be detected through the trained brain bleeding detection model, and realizing the detection, positioning and volume measurement of the brain bleeding according to the prediction result.
Because the cerebral hemorrhage detection model is constructed based on the 3D segmentation network, the cerebral hemorrhage probability of each individual pixel point can be predicted for one input cerebral CT image to be detected. In one embodiment of the present invention, if the bleeding probability of a certain voxel point is greater than a given threshold, for example greater than 0.5, it can be determined that the voxel point bleeds, otherwise the voxel point is normal.
In an embodiment of the present invention, if it is determined by prediction that bleeding voxel points are adjacent to each other and constitute one or more regions, it may be determined that the brain CT image to be detected is a brain CT image of a patient with cerebral hemorrhage, so as to detect cerebral hemorrhage of the patient. In addition, by determining the positions of the bleeding individual simple points and the normal individual simple points, the division boundary of the bleeding, that is, the position where the bleeding is obtained can be identified, and the parenchyma, the ventricle, the epidural, the subdural and the subarachnoid hemorrhage can be distinguished. The bleeding volume can be calculated from the number of voxel points per area constituted by the bleeding voxel points and the volume of each voxel point.
In order to verify the effect of the embodiment of the present invention, the method for detecting cerebral hemorrhage according to the embodiment of the present invention is evaluated by the test set, specifically, a Dice coefficient may be used to measure a doctor labeling gold standard a and a detection result B of the model according to the embodiment of the present invention, and the specific form is as follows:
Figure BDA0002407095560000051
where | a · B | is the number of overlapping voxels of the gold standard a and the detection result B of the model of the embodiment of the present invention, | a | is the number of gold standard voxels, and | B | is the number of voxels predicted by the model of the embodiment of the present invention.
In the verification process, the Dice coefficient is respectively calculated for each type of bleeding to evaluate the segmentation performance of the model for various types of bleeding, and the experimental result is shown in table 1.
TABLE 1
Type of bleeding Essence of the material Ventricle Epidural space Under the hard film Subarachnoid space
Dice coefficient 0.79 0.68 0.47 0.53 0.37
As can be seen from table 1, the method of the embodiment of the present invention has a good detection effect on cerebral hemorrhage, and can perform relatively accurate detection on subarachnoid hemorrhage whose contour is difficult to identify by a doctor. The visual effect diagram for detecting cerebral hemorrhage according to one embodiment of the present invention is shown in fig. 5, and can accurately segment the boundary of the focus, position the hemorrhage focus, and split the focus according to the position of the focus, where the focus marked by the solid line on the left side in the diagram is ventricular hemorrhage, and the focus marked by the dotted line on the right side in the diagram is substantial hemorrhage.
In summary, according to the method for detecting cerebral hemorrhage of the embodiment of the present invention, the cerebral hemorrhage detection model is constructed based on the 3D segmentation network, and after the cerebral hemorrhage detection model is trained, the hemorrhage probability of each voxel point of the CT image of the brain to be detected is predicted by the trained cerebral hemorrhage detection model, so that the detection, the positioning and the volume measurement of the cerebral hemorrhage can be automatically performed by the analysis at the voxel level, and the method has high efficiency and accuracy.
The invention also provides a cerebral hemorrhage detection device corresponding to the cerebral hemorrhage detection method of the embodiment.
As shown in fig. 6, the cerebral hemorrhage detecting apparatus according to the embodiment of the present invention includes an obtaining module 10, a modeling module 20, a training module 30, and a detecting module 40. The acquisition module 10 is used for acquiring a plurality of diagnosed cerebral hemorrhage brain CT images; the modeling module 20 is used for constructing a cerebral hemorrhage detection model based on a 3D segmentation network; the training module 30 is used for training the cerebral hemorrhage detection model through the cerebral hemorrhage brain CT image; the detection module 40 is configured to predict a bleeding probability of each voxel point of the input brain CT image to be detected through the trained brain bleeding detection model, and implement detection, location, and volume measurement of brain bleeding according to a prediction result.
After the CT image of the cerebral hemorrhage brain is obtained and labeled, the CT image is divided into a training set and a testing set, and then input into the obtaining module 10 for subsequent retrieval. In addition, the cerebral hemorrhage detecting device according to the embodiment of the present invention may further include a processing module, which is capable of performing a standardized processing on the plurality of diagnosed cerebral hemorrhage brain CT images acquired by the acquiring module 10. Specifically, the processing module may automatically normalize the original CT data of the 5mm thick layer, including performing operations such as gray level normalization and three-dimensional correction transformation by using a window width and window level.
Compared with a two-dimensional image, the brain CT image is three-dimensional data including spatial domain changes, and the information of the upper layer and the lower layer can effectively avoid misdiagnosis of artifacts and calcifications generated by the scanning device as bleeding, so the modeling module 20 of the embodiment of the present invention adopts a 3D full-volume depth segmentation network to better model the context information of the tomographic CT.
The cerebral hemorrhage detection model of the embodiment of the invention takes U-Net as a backbone network, the overall structure of the model is shown in fig. 2 and comprises a plurality of basic modules and a global module, wherein as shown in fig. 3, the basic modules comprise a 3D convolutional layer CONV, a nonlinear activation unit SFM and an element addition layer ADD, as shown in fig. 4, the global module comprises a 3D convolutional layer CONV, a matrix multiplication layer MU L, an element addition layer ADD and a Softmax activation unit SFM.
The network consists of two paths of coding and decoding, wherein the coding path inputs the characteristics of each scale into a module B L OCK and extracts corresponding high-level semantics through pooling layers or step convolution.
According to the embodiment of the invention, the context capture capability of the area to be predicted can be effectively enhanced by modeling the relation between different position characteristics.
The training module 30 may input the training set acquired by the acquiring module 10 into the cerebral hemorrhage detection model constructed by the modeling module 20, and train the cerebral hemorrhage detection model. The loss function used for training also introduces multilayer supervision loss and weighting loss on the basis of cross entropy loss, and ensures that the model can more quickly converge and learn the input training sample more effectively. In one embodiment of the present invention, the learning rate in the hyper-parameters associated with model training is 0.001, and the epoch number of model iterations is 100.
Because the cerebral hemorrhage detection model is constructed based on the 3D segmentation network, the cerebral hemorrhage probability of each individual pixel point can be predicted for one input cerebral CT image to be detected. In one embodiment of the present invention, if the bleeding probability of a certain voxel point is greater than a given threshold, for example greater than 0.5, it can be determined that the voxel point bleeds, otherwise the voxel point is normal.
In an embodiment of the present invention, if it is determined by prediction that the bleeding voxel points are adjacent to each other and constitute one or more regions, the detection module 40 may determine that the brain CT image to be detected is a brain CT image of a patient with cerebral hemorrhage, so as to detect cerebral hemorrhage of the patient. In addition, by determining the positions of the bleeding individual simple points and the normal individual simple points, the division boundary of the bleeding, that is, the position where the bleeding is obtained can be identified, and the parenchyma, the ventricle, the epidural, the subdural and the subarachnoid hemorrhage can be distinguished. The bleeding volume can be calculated from the number of voxel points per area constituted by the bleeding voxel points and the volume of each voxel point.
According to the cerebral hemorrhage detection device provided by the embodiment of the invention, the cerebral hemorrhage detection model is constructed based on the 3D segmentation network, and after the cerebral hemorrhage detection model is trained, the hemorrhage probability of each voxel point of the CT image of the brain to be detected is predicted through the trained cerebral hemorrhage detection model, so that the detection, the positioning and the volume measurement of the cerebral hemorrhage can be automatically performed through the analysis of the voxel level, and the efficiency and the accuracy are higher.
In the description of the present invention, "a plurality" means two or more unless specifically defined otherwise.
In the description herein, references to the description of the term "one embodiment," "some embodiments," "an example," "a specific example," or "some examples," etc., mean that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the invention. In this specification, the schematic representations of the terms used above are not necessarily intended to refer to the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples. Furthermore, various embodiments or examples and features of different embodiments or examples described in this specification can be combined and combined by one skilled in the art without contradiction.
Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that changes, modifications, substitutions and alterations can be made in these embodiments without departing from the principles and spirit of the invention, the scope of which is defined in the appended claims and their equivalents.

Claims (10)

1. A cerebral hemorrhage detection method is characterized by comprising the following steps:
acquiring a plurality of confirmed cerebral hemorrhage brain CT images;
constructing a cerebral hemorrhage detection model based on a 3D segmentation network;
training the cerebral hemorrhage detection model through the cerebral hemorrhage brain CT image;
inputting a brain CT image to be detected, predicting bleeding probability of each voxel point of the brain CT image to be detected through a trained cerebral bleeding detection model, and realizing detection, positioning and volume measurement of cerebral bleeding according to a prediction result.
2. The method for detecting cerebral hemorrhage according to claim 1, further comprising, after acquiring a plurality of confirmed cerebral hemorrhage brain CT images:
and carrying out standardization processing on the plurality of diagnosed cerebral hemorrhage brain CT images.
3. The method according to claim 1 or 2, wherein the cerebral hemorrhage detection model uses U-Net as a backbone network.
4. The method according to claim 3, wherein the cerebral hemorrhage detection model is constructed at a plurality of scales based on a 3D segmentation network.
5. The method according to claim 4, wherein the cerebral hemorrhage detection model comprises a plurality of basic modules and a global module.
6. The method of claim 5, wherein the base module comprises a 3D convolutional layer, a nonlinear activation unit, and an elemental additive layer.
7. The method of claim 5, wherein the global module comprises a 3D convolutional layer, a matrix multiplication layer, an element addition layer, and a Softmax activation unit.
8. The method of claim 1, wherein the brain hemorrhage detection model is trained using cross-entropy loss, multi-level supervised loss, and weighted loss.
9. A cerebral hemorrhage testing device, comprising:
an acquisition module for acquiring a plurality of diagnosed cerebral hemorrhage brain CT images;
a modeling module for constructing a cerebral hemorrhage detection model based on a 3D segmentation network;
a training module for training the cerebral hemorrhage detection model through the cerebral hemorrhage brain CT image;
and the detection module is used for predicting the bleeding probability of each voxel point of the input brain CT image to be detected through the trained brain bleeding detection model and realizing the detection, positioning and volume measurement of the brain bleeding according to the prediction result.
10. The cerebral hemorrhage detecting device according to claim 9, further comprising:
and the processing module is used for carrying out standardized processing on the plurality of diagnosed cerebral hemorrhage brain CT images.
CN202010164910.9A 2020-03-11 2020-03-11 Cerebral hemorrhage detection method and device Pending CN111402218A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010164910.9A CN111402218A (en) 2020-03-11 2020-03-11 Cerebral hemorrhage detection method and device

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010164910.9A CN111402218A (en) 2020-03-11 2020-03-11 Cerebral hemorrhage detection method and device

Publications (1)

Publication Number Publication Date
CN111402218A true CN111402218A (en) 2020-07-10

Family

ID=71430787

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010164910.9A Pending CN111402218A (en) 2020-03-11 2020-03-11 Cerebral hemorrhage detection method and device

Country Status (1)

Country Link
CN (1) CN111402218A (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114092446A (en) * 2021-11-23 2022-02-25 中国人民解放军总医院 Intracranial hemorrhage parameter acquisition method and device based on self-supervision learning and M-Net
CN114519701A (en) * 2022-01-27 2022-05-20 新瑞鹏宠物医疗集团有限公司 Ear hematoma detection method based on 3D image construction and related equipment
CN115187600A (en) * 2022-09-13 2022-10-14 杭州涿溪脑与智能研究所 Brain hemorrhage volume calculation method based on neural network

Citations (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104414636A (en) * 2013-08-23 2015-03-18 北京大学 Magnetic resonance image based cerebral micro-bleeding computer auxiliary detection system
CN109003299A (en) * 2018-07-05 2018-12-14 北京推想科技有限公司 A method of the calculating cerebral hemorrhage amount based on deep learning
CN109544534A (en) * 2018-11-26 2019-03-29 上海联影智能医疗科技有限公司 A kind of lesion image detection device, method and computer readable storage medium
CN109919961A (en) * 2019-02-22 2019-06-21 北京深睿博联科技有限责任公司 A kind of processing method and processing device for aneurysm region in encephalic CTA image
CN110120033A (en) * 2019-04-12 2019-08-13 天津大学 Based on improved U-Net neural network three-dimensional brain tumor image partition method
CN110211140A (en) * 2019-06-14 2019-09-06 重庆大学 Abdominal vascular dividing method based on 3D residual error U-Net and Weighted Loss Function
CN110598711A (en) * 2019-08-31 2019-12-20 华南理工大学 Target segmentation method combined with classification task
CN110705555A (en) * 2019-09-17 2020-01-17 中山大学 Abdomen multi-organ nuclear magnetic resonance image segmentation method, system and medium based on FCN
CN110738643A (en) * 2019-10-08 2020-01-31 上海联影智能医疗科技有限公司 Method for analyzing cerebral hemorrhage, computer device and storage medium
CN110874860A (en) * 2019-11-21 2020-03-10 哈尔滨工业大学 Target extraction method of symmetric supervision model based on mixed loss function

Patent Citations (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104414636A (en) * 2013-08-23 2015-03-18 北京大学 Magnetic resonance image based cerebral micro-bleeding computer auxiliary detection system
CN109003299A (en) * 2018-07-05 2018-12-14 北京推想科技有限公司 A method of the calculating cerebral hemorrhage amount based on deep learning
CN109544534A (en) * 2018-11-26 2019-03-29 上海联影智能医疗科技有限公司 A kind of lesion image detection device, method and computer readable storage medium
CN109919961A (en) * 2019-02-22 2019-06-21 北京深睿博联科技有限责任公司 A kind of processing method and processing device for aneurysm region in encephalic CTA image
CN110120033A (en) * 2019-04-12 2019-08-13 天津大学 Based on improved U-Net neural network three-dimensional brain tumor image partition method
CN110211140A (en) * 2019-06-14 2019-09-06 重庆大学 Abdominal vascular dividing method based on 3D residual error U-Net and Weighted Loss Function
CN110598711A (en) * 2019-08-31 2019-12-20 华南理工大学 Target segmentation method combined with classification task
CN110705555A (en) * 2019-09-17 2020-01-17 中山大学 Abdomen multi-organ nuclear magnetic resonance image segmentation method, system and medium based on FCN
CN110738643A (en) * 2019-10-08 2020-01-31 上海联影智能医疗科技有限公司 Method for analyzing cerebral hemorrhage, computer device and storage medium
CN110874860A (en) * 2019-11-21 2020-03-10 哈尔滨工业大学 Target extraction method of symmetric supervision model based on mixed loss function

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114092446A (en) * 2021-11-23 2022-02-25 中国人民解放军总医院 Intracranial hemorrhage parameter acquisition method and device based on self-supervision learning and M-Net
CN114092446B (en) * 2021-11-23 2024-07-16 中国人民解放军总医院 Intracranial hemorrhage parameter acquisition method and device based on self-supervision learning and M-Net
CN114519701A (en) * 2022-01-27 2022-05-20 新瑞鹏宠物医疗集团有限公司 Ear hematoma detection method based on 3D image construction and related equipment
CN115187600A (en) * 2022-09-13 2022-10-14 杭州涿溪脑与智能研究所 Brain hemorrhage volume calculation method based on neural network

Similar Documents

Publication Publication Date Title
KR101740464B1 (en) Method and system for diagnosis and prognosis of stroke and systme therefor
Majumdar et al. Detecting intracranial hemorrhage with deep learning
CN111402218A (en) Cerebral hemorrhage detection method and device
CN111445443B (en) Early acute cerebral infarction detection method and device
US11969265B2 (en) Neural network classification of osteolysis and synovitis near metal implants
CN111861989A (en) Method, system, terminal and storage medium for detecting midline of brain
US20230148996A1 (en) Lung ultrasound processing systems and methods
US20220004838A1 (en) Machine learning-based automated abnormality detection in medical images and presentation thereof
WO2021011775A1 (en) Systems and methods for generating classifying and quantitative analysis reports of aneurysms from medical image data
CN115274094A (en) Liver cancer imaging combined AFP diagnosis model construction method, system and device
CN112508943A (en) Breast tumor identification method based on ultrasonic image
CN111402219A (en) Old cerebral infarction detection method and device
Supriyanti et al. Coronal slices segmentation of mri images using active contour method on initial identification of alzheimer severity level based on clinical dementia rating (CDR)
WO2020099940A1 (en) Application of deep learning for medical imaging evaluation
Peddarapu et al. Liver Tumor Risk Prediction using Ensemble Methods
EP3965117A1 (en) Multi-modal computer-aided diagnosis systems and methods for prostate cancer
Mohammed et al. Esophageal Cancer Detection Using Feed-Forward Neural Network
Lee et al. Learning radiologist’s step-by-step skill for cervical spinal injury examination: Line drawing, prevertebral soft tissue thickness measurement, and swelling detection
Yildirim et al. DeepMSWeb: a web-based decision support system via deep learning for automatic detection of MS lesions
Gunasekara et al. A feasibility study for deep learning based automated brain tumor segmentation using magnetic resonance images
CN112541917B (en) CT image processing method for cerebral hemorrhage disease
JP7450410B2 (en) Medical information processing device, medical information processing method, and medical information processing program
Sun et al. Reliable detection of lymph nodes in whole pelvic for radiotherapy
Topçu Prediction of Covid-19 risk of a person by analyzing computed tomography images using convolutional neural networks
Cortes-Ferre et al. Deep Learning applied to Intracranial Hemorrhage Detection. Preprints 2021, 1, 0

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