CN112754511A - CT image intracranial thrombus detection and property classification method based on deep learning - Google Patents
CT image intracranial thrombus detection and property classification method based on deep learning Download PDFInfo
- Publication number
- CN112754511A CN112754511A CN202110076042.3A CN202110076042A CN112754511A CN 112754511 A CN112754511 A CN 112754511A CN 202110076042 A CN202110076042 A CN 202110076042A CN 112754511 A CN112754511 A CN 112754511A
- Authority
- CN
- China
- Prior art keywords
- thrombus
- image
- intracranial
- property
- inputting
- 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
Links
- 238000000034 method Methods 0.000 title claims abstract description 24
- 238000001514 detection method Methods 0.000 title claims abstract description 19
- 238000013135 deep learning Methods 0.000 title claims abstract description 15
- 201000001429 Intracranial Thrombosis Diseases 0.000 title claims abstract description 14
- 208000007536 Thrombosis Diseases 0.000 claims abstract description 101
- 230000011218 segmentation Effects 0.000 claims abstract description 27
- 238000007917 intracranial administration Methods 0.000 claims abstract description 18
- 208000006011 Stroke Diseases 0.000 claims abstract description 17
- 230000035699 permeability Effects 0.000 claims abstract description 10
- 238000007781 pre-processing Methods 0.000 claims abstract description 6
- 210000003625 skull Anatomy 0.000 claims abstract description 5
- 239000013598 vector Substances 0.000 claims description 10
- 238000011176 pooling Methods 0.000 claims description 8
- 238000013434 data augmentation Methods 0.000 claims description 6
- 230000006870 function Effects 0.000 claims description 5
- 101100456045 Schizosaccharomyces pombe (strain 972 / ATCC 24843) map3 gene Proteins 0.000 claims description 4
- 238000010606 normalization Methods 0.000 claims description 4
- 238000012549 training Methods 0.000 claims description 4
- 206010014513 Embolism arterial Diseases 0.000 claims description 3
- 208000031481 Pathologic Constriction Diseases 0.000 claims description 3
- 238000011065 in-situ storage Methods 0.000 claims description 3
- 238000002372 labelling Methods 0.000 claims description 3
- 230000036262 stenosis Effects 0.000 claims description 3
- 208000037804 stenosis Diseases 0.000 claims description 3
- PXFBZOLANLWPMH-UHFFFAOYSA-N 16-Epiaffinine Natural products C1C(C2=CC=CC=C2N2)=C2C(=O)CC2C(=CC)CN(C)C1C2CO PXFBZOLANLWPMH-UHFFFAOYSA-N 0.000 claims description 2
- 238000004364 calculation method Methods 0.000 claims description 2
- 238000012937 correction Methods 0.000 claims description 2
- 238000000605 extraction Methods 0.000 claims description 2
- 230000004927 fusion Effects 0.000 claims description 2
- 230000001131 transforming effect Effects 0.000 claims description 2
- 238000012935 Averaging Methods 0.000 claims 1
- 238000003745 diagnosis Methods 0.000 abstract description 4
- 230000003321 amplification Effects 0.000 abstract 1
- 230000007547 defect Effects 0.000 abstract 1
- 238000003199 nucleic acid amplification method Methods 0.000 abstract 1
- 238000012216 screening Methods 0.000 abstract 1
- 238000010968 computed tomography angiography Methods 0.000 description 16
- 208000032382 Ischaemic stroke Diseases 0.000 description 5
- 238000012545 processing Methods 0.000 description 3
- 102000056548 Member 3 Solute Carrier Family 12 Human genes 0.000 description 2
- 108091006623 SLC12A3 Proteins 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 2
- 238000002591 computed tomography Methods 0.000 description 2
- 230000006378 damage Effects 0.000 description 2
- 201000010099 disease Diseases 0.000 description 2
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 2
- 241000282414 Homo sapiens Species 0.000 description 1
- 230000004075 alteration Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 210000004204 blood vessel Anatomy 0.000 description 1
- 210000004958 brain cell Anatomy 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 239000003814 drug Substances 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 230000002427 irreversible effect Effects 0.000 description 1
- 208000037906 ischaemic injury Diseases 0.000 description 1
- 230000035515 penetration Effects 0.000 description 1
- 230000008092 positive effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 230000001732 thrombotic effect Effects 0.000 description 1
Images
Classifications
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B6/00—Apparatus or devices for radiation diagnosis; Apparatus or devices for radiation diagnosis combined with radiation therapy equipment
- A61B6/02—Arrangements for diagnosis sequentially in different planes; Stereoscopic radiation diagnosis
- A61B6/03—Computed tomography [CT]
- A61B6/032—Transmission computed tomography [CT]
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B6/00—Apparatus or devices for radiation diagnosis; Apparatus or devices for radiation diagnosis combined with radiation therapy equipment
- A61B6/50—Apparatus or devices for radiation diagnosis; Apparatus or devices for radiation diagnosis combined with radiation therapy equipment specially adapted for specific body parts; specially adapted for specific clinical applications
- A61B6/501—Apparatus or devices for radiation diagnosis; Apparatus or devices for radiation diagnosis combined with radiation therapy equipment specially adapted for specific body parts; specially adapted for specific clinical applications for diagnosis of the head, e.g. neuroimaging or craniography
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B6/00—Apparatus or devices for radiation diagnosis; Apparatus or devices for radiation diagnosis combined with radiation therapy equipment
- A61B6/52—Devices using data or image processing specially adapted for radiation diagnosis
- A61B6/5211—Devices using data or image processing specially adapted for radiation diagnosis involving processing of medical diagnostic data
-
- 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
-
- 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
- G16H30/00—ICT specially adapted for the handling or processing of medical images
- G16H30/40—ICT specially adapted for the handling or processing of medical images for processing medical images, e.g. editing
-
- 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
- G06T2207/20081—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)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Medical Informatics (AREA)
- General Health & Medical Sciences (AREA)
- Physics & Mathematics (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Radiology & Medical Imaging (AREA)
- Public Health (AREA)
- Heart & Thoracic Surgery (AREA)
- Veterinary Medicine (AREA)
- Animal Behavior & Ethology (AREA)
- Theoretical Computer Science (AREA)
- Surgery (AREA)
- Molecular Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Biophysics (AREA)
- High Energy & Nuclear Physics (AREA)
- Optics & Photonics (AREA)
- Pathology (AREA)
- Biomedical Technology (AREA)
- General Physics & Mathematics (AREA)
- Pulmonology (AREA)
- Quality & Reliability (AREA)
- Primary Health Care (AREA)
- Epidemiology (AREA)
- Neurology (AREA)
- Neurosurgery (AREA)
- Dentistry (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Apparatus For Radiation Diagnosis (AREA)
Abstract
The invention discloses a CT image intracranial thrombus detection and property classification method based on deep learning. Preprocessing collected skull CTA data of a stroke patient and data amplification, identifying the intracranial data by utilizing a UNet-based improved segmentation network, determining a thrombus position and giving uncertainty estimation, and judging the thrombus permeability by utilizing a classification network. The method can quickly detect the potential thrombus position of the stroke patient according to the CT image, judge the thrombus permeability, provide a reliability reference for thrombus detection, provide a basis for a doctor to clinically and quickly diagnose and determine a thrombus taking treatment scheme, make up the defect that the doctor clinically and manually identifies the position and the property of the thrombus, and can be used as a quick screening tool to improve the diagnosis and treatment efficiency.
Description
Technical Field
The invention belongs to the field of medical image processing and clinical medicine intersection, relates to an auxiliary detection method for thrombus of stroke patients, and particularly relates to a CT image intracranial thrombus detection and property classification method based on deep learning.
Technical Field
Stroke is one of the main diseases causing disability and death of human beings, Acute Ischemic Stroke (AIS) accounts for about 80% of all strokes, and rapid vessel opening and ischemic injury area saving are important targets for AIS treatment. At present, clinical tests relatively consistently consider that intravascular treatment mainly based on mechanical thrombus removal can bring clear benefits in screened patients with acute ischemic stroke of large blood vessels. In clinical treatment, the longer the patient's disease, the greater the likelihood of irreversible damage and other damage to the brain cells, and even death. The thrombus position is quickly determined, the thrombus permeability is judged, the thrombus taking operation scheme can be determined for judging the condition of the stroke patient, and precious treatment time is won for the patient.
In clinical practice, in order to acquire images of the skull of a patient in time, CT scanning is generally adopted to examine the patient, and mainly includes non-contrast Computed Tomography (NCCT, also called CT flat scan) and Computed Tomography Angiography (CTA). At present, the thrombus position of a stroke patient is determined mainly by manually judging a CTA image by a clinician, and the possible thrombus position can be quickly identified for the physician by utilizing artificial intelligence auxiliary judgment, so that the physician is helped to determine the thrombus more quickly, and the shunting and diagnosis decision of the patient are helped to be accelerated.
CTA does not directly image the thrombotic nature of the patient as Magnetic Resonance (MRI). But because the CT examination only needs a few minutes, compared with the examination time of more than half an hour of MRI, the method has the characteristic of high examination speed. The method promotes the relevant research of judging the thrombus property on the basis of the CT image, and scholars at home and abroad put forward some physiological index judgment bases for assisting the CT image and relevant methods for judging by utilizing the characteristics of different image formation of the thrombus on NCCT and CTA images. However, these methods rely on manual statistics and processing, have not yet formed practical computer-aided tools, have large manual errors, and also have the problem of being slow in processing in time. At present, although a computer platform for CT image stroke analysis exists, a computer-aided function for judging the position and the property of the thrombus based on the CT image is not available temporarily.
At present, the AIS clinical first aid needs a computer tool which can automatically and quickly locate thrombus and determine the nature of the thrombus by CT examination and assist doctors in making decisions on treatment schemes.
Disclosure of Invention
In order to solve the technical problems, the invention provides a CT image intracranial thrombus detection and property classification method based on deep learning by combining an artificial intelligence algorithm based on deep learning. The invention provides a judgment reference for the permeability of thrombus while identifying the position of the thrombus of the stroke patient by utilizing the CT image, thereby achieving the purposes of assisting a clinician to quickly judge the thrombus and determining a thrombus taking scheme.
The technical scheme adopted by the invention is as follows: a CT image intracranial thrombus detection and property classification method based on deep learning. The method is characterized by comprising the following steps:
step 1: preprocessing the collected CTA image of the head of the patient and amplifying the data to obtain an intracranial image, and simultaneously performing the same operation on the corresponding thrombus label to ensure that the thrombus label is the same as the actual thrombus position; wherein the CTA image of the head of the patient is derived from clinical scanning data in dicom format of the stroke patient, and the thrombus label is obtained by labeling the thrombus by a professional doctor;
step 2: inputting the intracranial image obtained in the step 1 into a thrombus segmentation network improved based on UNet, and acquiring identification characteristics under supervision training of a thrombus label to obtain a thrombus segmentation result and uncertainty estimation;
and step 3: combining the thrombus segmentation result obtained in the step 2 with an intracranial image as an attention map to obtain a thrombus property distinguishing area, and inputting the thrombus property distinguishing area into a thrombus property classification network based on ResNet-50 so as to obtain thrombus penetration properties (including in-situ stenosis, occlusion and arterial embolism).
Compared with the existing detection method, the method has the following advantages and positive effects:
(1) the invention fully automatically identifies the thrombus of the stroke patient based on the CT image, does not need the MRI examination result and clinical data, can shorten the diagnosis time and save the examination and treatment cost.
(2) The invention provides result reliability reference through uncertainty estimation while automatically determining the position of the thrombus.
(3) The invention can automatically determine the position of the thrombus by utilizing the CT image and make reference judgment on the permeability of the thrombus, thereby providing more reference information for doctors and improving the treatment efficiency.
Drawings
FIG. 1: block diagram of an embodiment of the present invention.
Detailed Description
In order to facilitate the understanding and implementation of the present invention for those of ordinary skill in the art, the present invention is further described in detail with reference to the accompanying drawings and the implementation examples, it is to be understood that the implementation examples described herein are only for the purpose of illustration and explanation and are not to be construed as limiting the present invention.
A CT image intracranial thrombus detection and property classification method based on deep learning is a thrombus detection and property classification method aiming at stroke patients. Firstly, preprocessing and data augmentation are carried out on a CTA image of the head of a stroke patient; then, inputting the processed image into a thrombus segmentation network improved based on UNet, and determining the position of the thrombus; and finally, combining the thrombus segmentation result as an attention map with an intracranial image to obtain a thrombus property distinguishing area, and inputting the thrombus property distinguishing area into a thrombus property classification network to obtain the thrombus permeability.
Referring to fig. 1, the method for detecting intracranial thrombus and classifying properties of CT images based on deep learning provided by the present invention includes the following steps:
step 1: preprocessing the collected CTA image of the head of the patient and amplifying the data to obtain an intracranial image, and simultaneously performing the same operation on the corresponding thrombus label to ensure that the thrombus label is the same as the actual thrombus position; wherein the CTA image of the head of the patient is derived from clinical scanning data in dicom format of the stroke patient, and the thrombus label is obtained by labeling the thrombus by a professional doctor;
in this embodiment, the specific implementation of step 1 includes the following substeps:
step 1.1: removing the anatomical portion below the C1 vertebra from the patient's head CTA image sequence to obtain an image of the skull portion;
step 1.2: carrying out data augmentation treatment on the CTA intracranial image obtained in the step 1.1; the specific mode is to randomly combine the following operations: randomly rotating any angle between-90 degrees and 90 degrees, randomly horizontally turning, randomly affine transforming, and keeping the random scaling and cutting of the original size, thereby obtaining more groups of intracranial image data by expansion;
step 1.3: the data augmentation operation used on the CTA image in step 1.2 is applied to the corresponding thrombus label with the same parameters, ensuring that the thrombus label is the same as the actual thrombus position.
Step 2: inputting the intracranial image obtained in the step 1 into a thrombus segmentation network improved based on UNet, and acquiring identification characteristics under supervision training of a thrombus label to obtain a thrombus segmentation result and uncertainty estimation;
in this embodiment, the specific implementation of step 2 includes the following substeps:
step 2.1: inputting the CTA intracranial images with the size of 512 multiplied by 512 obtained in the step 1 into a thrombus segmentation network improved based on the UNet in the size of 256 multiplied by 256, firstly obtaining a characteristic map e _ out1 through an encoder block1 of the UNet, and then inputting the e _ out1 into an encoderlock 2 through a pooling layer to obtain e _ out 2; then, inputting the e _ out2 into an encoderlock 3 through a pooling layer to obtain e _ out3, and performing the same operation layer by layer to obtain e _ out4 and e _ out 5;
step 2.2: e _ out5 obtained in the step 2.1 passes through a bottleneck layer formed by a mixing pooling module to obtain d _ out 5;
step 2.3: d _ out5 obtained in step 2.2 is up-sampled to obtain d _ in4, d _ in4 and e _ out4 are used as the input of attention gate attention _ gate4 to obtain attention _ map4, and then attintionmap 4 and d _ in4 are input into decoderlock 4 after interleaving and splicing to obtain d _ out 4; d _ out4 is subjected to upsampling to obtain d _ in3, d _ in3 and e _ out3 are used as the input of an attention gate attention _ gate3 to obtain an attention _ map3, and then the attention map3 and d _ in3 are subjected to staggered splicing and input into decoderlock 3 to obtain d _ out3, and the same operation is carried out layer by layer to obtain d _ out2 and d _ out 1;
step 2.4: d _ out1 obtaining a group of prediction results through a group convolution module, obtaining uncertainty estimation by calculating variance of the prediction results, and obtaining original segmentation result s by calculating averageo(ii) a Wherein the encoder block and the decoder block consist of 1 × 1 group convolution + group normalization + leakyrelu +3 × 3 group convolution + group normalization + leakyrelu;
step 2.5: the original segmentation result s obtained in step 2.4 is usedoInputting the direction field into a direction field module to obtain a predicted direction field
Will be provided with
And soInputting the data into a feature correction and fusion module to obtain a final segmentation result sf。
The present embodiment trains the UNet-based improved thrombus segmentation network in step 2 by using the following loss function:
sgtground Truth, loss representing segmentation resultBCEAnd lossdiceRespectively, normalized binary cross entropy and DiThe loss of ce is a loss of the ce,
the calculation method is as follows:
wherein the symbol |2The expression of the euclidean distance,<>representing the included angle of the vector; DF is the group Truth of the direction field; Ω represents a full set of pixels; p is a certain pixel in Ω; DF (p) denotes the DF vector at p;
represented at p
Vector quantity; w (p) is the pixel weight, calculated as:
wherein, CiRepresents a pixel of class i, | CiI represents the number of i-th pixels, NclsAnd (4) representing the category number (specifically including thrombus and background).
And step 3: combining the thrombus segmentation result obtained in the step 2 with an intracranial image as an attention map to obtain a thrombus property distinguishing area, and inputting the thrombus property distinguishing area into a thrombus property classification network based on ResNet-50 so as to obtain thrombus permeability properties (including in-situ stenosis, occlusion and arterial embolism);
in this embodiment, the specific implementation of step 3 includes the following substeps:
step 3.1: the final segmentation result s obtained in the step 2fCombining the attention map with the original CTA image of the patient head, and cutting out a thrombus property distinguishing area with the size of 64 multiplied by 64;
step 3.2: inputting the thrombus property distinguishing region obtained in the step 3.1 into a ResNet-50 feature extraction part to obtain a classification feature map;
step 3.3: performing global average pooling on the classification feature map obtained in the step 3.2 to obtain feature vectors;
step 3.4: obtaining the classification probability p by using Softmax for the feature vector obtained in the step 3.3iWherein p isiRepresenting the probability that the thrombus is a type i thrombus.
This example trains the ResNet-50 based classification network of thrombus properties in step 3 using a standard cross entropy loss function.
The method uses collected skull CTA data of a stroke patient to carry out preprocessing and data augmentation, utilizes a UNet-based improved segmentation network to identify the intracranial data, determines the position of thrombus and gives uncertainty estimation, and utilizes a classification network to judge the thrombus permeability. The method provided by the invention is used as computer-aided judgment to complete the detection of the thrombus permeability on the basis of the detection of the position of the thrombus of the stroke patient based on the CT image, can shorten the diagnosis time, saves the inspection and treatment cost, and improves the treatment efficiency.
It should be understood that parts of the specification not set forth in detail are well within the prior art.
It should be understood that the above description of the preferred embodiments is given for clarity and not for any purpose of limitation, and that various changes, substitutions and alterations can be made herein without departing from the spirit and scope of the invention as defined by the appended claims.
Claims (6)
1. A CT image intracranial thrombus detection and property classification method based on deep learning is characterized by comprising the following steps:
step 1: preprocessing the collected CTA image of the head of the patient and amplifying the data to obtain an intracranial image, and simultaneously performing the same operation on the corresponding thrombus label to ensure that the thrombus label is the same as the actual thrombus position; wherein the CTA image of the head of the patient is derived from clinical scanning data in dicom format of the stroke patient, and the thrombus label is obtained by labeling the thrombus by a professional doctor;
step 2: inputting the intracranial image obtained in the step 1 into a thrombus segmentation network improved based on UNet, and acquiring identification characteristics under supervision training of a thrombus label to obtain a thrombus segmentation result and uncertainty estimation;
and step 3: and (3) combining the thrombus segmentation result obtained in the step (2) with an intracranial image as an attention map to obtain a thrombus property distinguishing area, and inputting the thrombus property distinguishing area into a thrombus property classification network based on ResNet-50 so as to obtain thrombus permeability properties including in-situ stenosis, occlusion and arterial embolism.
2. The deep learning-based CT image intracranial thrombus detection and property classification method according to claim 1, wherein the detailed implementation of the step 1 is as follows:
step 1.1: removing the anatomical portion below the C1 vertebra from the patient's head CTA image sequence to obtain an image of the skull portion;
step 1.2: carrying out data augmentation treatment on the CTA intracranial image obtained in the step 1.1; the specific mode is to randomly combine the following operations: randomly rotating any angle between-90 degrees and 90 degrees, randomly horizontally turning, randomly affine transforming, and keeping the random scaling and cutting of the original size, thereby obtaining more groups of intracranial image data by expansion;
step 1.3: the data augmentation operation used on the CTA image in step 1.2 is applied to the corresponding thrombus label with the same parameters, ensuring that the thrombus label is the same as the actual thrombus position.
3. The deep learning-based CT image intracranial thrombus detection and property classification method according to claim 1, wherein the step 2 is realized by the following steps:
step 2.1: inputting the CTA intracranial images with the size of 512 multiplied by 512 obtained in the step 1 into a thrombus segmentation network improved based on the UNet in the size of 256 multiplied by 256, firstly obtaining a characteristic map e _ out1 through encoderlock 1 of the UNet, and then inputting e _ out1 into encoderlock 2 through a pooling layer to obtain e _ out 2; then, inputting the e _ out2 into an encoderlock 3 through a pooling layer to obtain e _ out3, and performing the same operation layer by layer to obtain e _ out4 and e _ out 5;
step 2.2: e _ out5 obtained in the step 2.1 passes through a bottleneck layer formed by a mixing pooling module to obtain d _ out 5;
step 2.3: d _ out5 obtained in step 2.2 is up-sampled to obtain d _ in4, d _ in4 and e _ out4 are used as the input of attention gate attention _ gate4 to obtain attention _ map4, and then attintionmap 4 and d _ in4 are input into decoderlock 4 after interleaving and splicing to obtain d _ out 4; d _ out4 is up-sampled to obtain d _ in3, d _ in3 and e _ out3 are used as the input of an attention gate attribute _ gate3 to obtain attribute _ map3, then attribute map3 and d _ in3 are input into decoderlock 3 to obtain d _ out3, and the same operation is carried out layer by layer to obtain d _ out2 and d _ tlout;
step 2.4: d _ outl is processed by a group convolution module to obtain a group of prediction results, variance of the prediction results is calculated to obtain uncertainty estimation, and averaging is carried out to obtain an original segmentation result so(ii) a Wherein the encoder block and the decoder block consist of 1 × 1 group convolution + group normalization + leakyrelu +3 × 3 group convolution + group normalization + leakyrelu;
step 2.5: the original segmentation result s obtained in step 2.4 is usedoInputting the direction field into a direction field module to obtain a predicted direction field
Will be provided with
And soInputting the data into a feature correction and fusion module to obtain a final segmentation result sf。
4. The deep learning-based CT image intracranial thrombus detection and property classification method according to claim 3, wherein: training the UNet-based improved thrombus segmentation network in step 2 by using the following loss function:
sgtground Truth, loss representing segmentation resultBCEAnd lossdiceRespectively the standard binary cross entropy and the Dice loss,
the calculation method is as follows:
wherein the symbol | | | purple2The expression of the euclidean distance,<>representing the included angle of the vector; DF is the GroudTruth of the direction field; Ω represents a full set of pixels; p is a certain pixel in Ω; DF (p) denotes the DF vector at p;
represented at p
Vector quantity; w (p) is the pixel weight, calculated as:
wherein, CiRepresents a pixel of class i, | CiI represents the number of i-th pixels, NclsThe number of representative categories specifically includes thrombus and background.
5. The deep learning-based CT image intracranial thrombus detection and property classification method according to claim 1, wherein the step 3 is implemented specifically as follows:
step 3.1: will be described in detail2 final segmentation result sfCombining the attention map with the original CTA image of the patient head, and cutting out a thrombus property distinguishing area with the size of 64 multiplied by 64;
step 3.2: inputting the thrombus property distinguishing region obtained in the step 3.1 into a ResNet-50 feature extraction part to obtain a classification feature map;
step 3.3: performing global average pooling on the classification feature map obtained in the step 3.2 to obtain feature vectors;
step 3.4: obtaining the classification probability p by using Softmax for the feature vector obtained in the step 3.3iWherein p isiRepresenting the probability that the thrombus is a type i thrombus.
6. The deep learning based CT image intracranial thrombus detection and property classification method according to any one of claims 1-5, wherein: the ResNet-50 based classification network of thrombus properties described in step 3 was trained using a standard cross entropy loss function.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110076042.3A CN112754511A (en) | 2021-01-20 | 2021-01-20 | CT image intracranial thrombus detection and property classification method based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110076042.3A CN112754511A (en) | 2021-01-20 | 2021-01-20 | CT image intracranial thrombus detection and property classification method based on deep learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112754511A true CN112754511A (en) | 2021-05-07 |
Family
ID=75701759
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110076042.3A Pending CN112754511A (en) | 2021-01-20 | 2021-01-20 | CT image intracranial thrombus detection and property classification method based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112754511A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113223704A (en) * | 2021-05-20 | 2021-08-06 | 吉林大学 | Auxiliary diagnosis method for computed tomography aortic aneurysm based on deep learning |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103646135A (en) * | 2013-11-28 | 2014-03-19 | 哈尔滨医科大学 | Computer-assisted ultrasonic diagnosis method for left atrium/left auricle thrombus |
| KR20190037458A (en) * | 2017-09-29 | 2019-04-08 | 주식회사 인피니트헬스케어 | Computing system and method for identifying and visualizing cerebral thrombosis based on medical images |
| CN110808096A (en) * | 2019-10-30 | 2020-02-18 | 北京邮电大学 | Automatic heart lesion detection system based on convolutional neural network |
| CN110956092A (en) * | 2019-11-06 | 2020-04-03 | 江苏大学 | Intelligent metallographic detection and rating method and system based on deep learning |
| CN111242168A (en) * | 2019-12-31 | 2020-06-05 | 浙江工业大学 | Human skin image lesion classification method based on multi-scale attention features |
| WO2020138932A1 (en) * | 2018-12-24 | 2020-07-02 | 주식회사 제이엘케이인스펙션 | Machine learning-based method and system for classifying thrombi using gre image |
| CN111667458A (en) * | 2020-04-30 | 2020-09-15 | 杭州深睿博联科技有限公司 | Method and device for detecting early acute cerebral infarction in flat-scan CT |
| CN111798458A (en) * | 2020-06-15 | 2020-10-20 | 电子科技大学 | Interactive medical image segmentation method based on uncertainty guidance |
-
2021
- 2021-01-20 CN CN202110076042.3A patent/CN112754511A/en active Pending
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103646135A (en) * | 2013-11-28 | 2014-03-19 | 哈尔滨医科大学 | Computer-assisted ultrasonic diagnosis method for left atrium/left auricle thrombus |
| KR20190037458A (en) * | 2017-09-29 | 2019-04-08 | 주식회사 인피니트헬스케어 | Computing system and method for identifying and visualizing cerebral thrombosis based on medical images |
| WO2020138932A1 (en) * | 2018-12-24 | 2020-07-02 | 주식회사 제이엘케이인스펙션 | Machine learning-based method and system for classifying thrombi using gre image |
| CN110808096A (en) * | 2019-10-30 | 2020-02-18 | 北京邮电大学 | Automatic heart lesion detection system based on convolutional neural network |
| CN110956092A (en) * | 2019-11-06 | 2020-04-03 | 江苏大学 | Intelligent metallographic detection and rating method and system based on deep learning |
| CN111242168A (en) * | 2019-12-31 | 2020-06-05 | 浙江工业大学 | Human skin image lesion classification method based on multi-scale attention features |
| CN111667458A (en) * | 2020-04-30 | 2020-09-15 | 杭州深睿博联科技有限公司 | Method and device for detecting early acute cerebral infarction in flat-scan CT |
| CN111798458A (en) * | 2020-06-15 | 2020-10-20 | 电子科技大学 | Interactive medical image segmentation method based on uncertainty guidance |
Non-Patent Citations (3)
| Title |
|---|
| F. CHENG ET AL: "Learning Directional Feature Maps for Cardiac MRI Segmentation", 《INTERNATIONAL CONFERENCE ON MEDICAL IMAGE COMPUTING AND COMPUTER-ASSISTED INTERVENTION 2020》 * |
| 祝恩等: "《自动指纹识别技术》", 31 May 2006 * |
| 魏凤芹: "基于深度学习的甲状腺超声图像自动分割方法研究", 《中国优秀硕士学位论文全文数据库医药卫生科技辑》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113223704A (en) * | 2021-05-20 | 2021-08-06 | 吉林大学 | Auxiliary diagnosis method for computed tomography aortic aneurysm based on deep learning |
| CN113223704B (en) * | 2021-05-20 | 2022-07-26 | 吉林大学 | Auxiliary diagnosis method for computed tomography aortic aneurysm based on deep learning |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Li et al. | Deep learning for hemorrhagic lesion detection and segmentation on brain CT images | |
| CN110310281B (en) | Mask-RCNN deep learning-based pulmonary nodule detection and segmentation method in virtual medical treatment | |
| Saikumar et al. | A novel implementation heart diagnosis system based on random forest machine learning technique. | |
| CN106056595B (en) | Based on the pernicious assistant diagnosis system of depth convolutional neural networks automatic identification Benign Thyroid Nodules | |
| CN113420826B (en) | Liver focus image processing system and image processing method | |
| Yao et al. | Automated hematoma segmentation and outcome prediction for patients with traumatic brain injury | |
| CN108665454A (en) | A kind of endoscopic image intelligent classification and irregular lesion region detection method | |
| CN111667478A (en) | Method and system for identifying carotid plaque through CTA-MRA cross-modal prediction | |
| Zhou et al. | A unified visual information preservation framework for self-supervised pre-training in medical image analysis | |
| CN113160120A (en) | Liver blood vessel segmentation method and system based on multi-mode fusion and deep learning | |
| CN113159223A (en) | Carotid artery ultrasonic image identification method based on self-supervision learning | |
| Peña-Solórzano et al. | Findings from machine learning in clinical medical imaging applications–Lessons for translation to the forensic setting | |
| Nayan et al. | A deep learning approach for brain tumor detection using magnetic resonance imaging | |
| CN112754511A (en) | CT image intracranial thrombus detection and property classification method based on deep learning | |
| Cao et al. | Segmentation of intracerebral hemorrhage based on improved U-Net | |
| CN209770401U (en) | Medical image rapid analysis processing system | |
| CN116934683A (en) | Method for assisting ultrasonic diagnosis of spleen wound by artificial intelligence | |
| CN114926486B (en) | Thyroid ultrasound image intelligent segmentation method based on multi-level improvement | |
| Cheng et al. | Automatic intracranial aneurysm segmentation based on spatial information fusion feature from 3D-RA using U-Net | |
| Wen et al. | A novel lesion segmentation algorithm based on U-Net network for Tuberculosis CT image | |
| Zhou et al. | Pcrlv2: A unified visual information preservation framework for self-supervised pre-training in medical image analysis | |
| Ashame et al. | Abnormality Detection in Eye Fundus Retina | |
| Khastavaneh et al. | Automated segmentation of abnormal tissues in medical images | |
| Cao et al. | Automatic lung segmentation algorithm on chest X-ray images based on fusion variational auto-encoder and three-terminal attention mechanism. Symmetry 2021; 13 (5): 814 | |
| Moalla et al. | Deep AngioKey: A Novel Approach for Objective Keyframe Extraction in Coronary Angiography Analysis |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210507 |
|
| RJ01 | Rejection of invention patent application after publication |