CN112907563B - CTA full-automatic collateral circulation scoring method and system based on deep learning - Google Patents
CTA full-automatic collateral circulation scoring method and system based on deep learning Download PDFInfo
- Publication number
- CN112907563B CN112907563B CN202110286102.4A CN202110286102A CN112907563B CN 112907563 B CN112907563 B CN 112907563B CN 202110286102 A CN202110286102 A CN 202110286102A CN 112907563 B CN112907563 B CN 112907563B
- Authority
- CN
- China
- Prior art keywords
- cta
- brain
- image
- scoring
- map
- 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
- 206010069729 Collateral circulation Diseases 0.000 title claims abstract description 48
- 238000013135 deep learning Methods 0.000 title claims abstract description 24
- 238000013077 scoring method Methods 0.000 title claims abstract description 20
- 210000004556 brain Anatomy 0.000 claims abstract description 95
- 238000010968 computed tomography angiography Methods 0.000 claims abstract description 92
- 210000004204 blood vessel Anatomy 0.000 claims abstract description 51
- 239000013598 vector Substances 0.000 claims abstract description 39
- 238000000034 method Methods 0.000 claims abstract description 21
- 230000011218 segmentation Effects 0.000 claims abstract description 21
- 238000010969 brain CTA Methods 0.000 claims abstract description 13
- 230000017531 blood circulation Effects 0.000 claims abstract description 12
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 10
- 238000013145 classification model Methods 0.000 claims abstract description 8
- 230000000873 masking effect Effects 0.000 claims abstract description 5
- 238000010606 normalization Methods 0.000 claims description 21
- 238000012549 training Methods 0.000 claims description 19
- 210000001367 artery Anatomy 0.000 claims description 13
- 230000002490 cerebral effect Effects 0.000 claims description 12
- 238000012360 testing method Methods 0.000 claims description 10
- 210000001627 cerebral artery Anatomy 0.000 claims description 8
- 230000008569 process Effects 0.000 claims description 8
- 230000009466 transformation Effects 0.000 claims description 8
- 230000003044 adaptive effect Effects 0.000 claims description 7
- 238000012937 correction Methods 0.000 claims description 7
- 230000002792 vascular Effects 0.000 claims description 7
- 230000004913 activation Effects 0.000 claims description 6
- 238000007637 random forest analysis Methods 0.000 claims description 6
- 210000003625 skull Anatomy 0.000 claims description 6
- 230000001131 transforming effect Effects 0.000 claims description 6
- 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 5
- 210000004720 cerebrum Anatomy 0.000 claims description 5
- 238000005192 partition Methods 0.000 claims description 5
- 238000002224 dissection Methods 0.000 claims description 4
- 238000012216 screening Methods 0.000 claims description 4
- 238000005728 strengthening Methods 0.000 claims description 4
- 230000036541 health Effects 0.000 claims description 3
- 238000011176 pooling Methods 0.000 claims description 3
- 230000035945 sensitivity Effects 0.000 claims description 3
- 230000006835 compression Effects 0.000 claims description 2
- 238000007906 compression Methods 0.000 claims description 2
- 230000000694 effects Effects 0.000 claims 1
- 238000011156 evaluation Methods 0.000 abstract description 4
- 230000007246 mechanism Effects 0.000 abstract description 3
- 238000012545 processing Methods 0.000 abstract description 2
- 230000000302 ischemic effect Effects 0.000 description 10
- 208000006011 Stroke Diseases 0.000 description 9
- 208000032382 Ischaemic stroke Diseases 0.000 description 8
- 230000004087 circulation Effects 0.000 description 7
- 238000010586 diagram Methods 0.000 description 7
- 206010008190 Cerebrovascular accident Diseases 0.000 description 6
- 230000006870 function Effects 0.000 description 6
- 239000003146 anticoagulant agent Substances 0.000 description 4
- 230000036770 blood supply Effects 0.000 description 4
- 238000002560 therapeutic procedure Methods 0.000 description 4
- 230000002537 thrombolytic effect Effects 0.000 description 4
- 238000011282 treatment Methods 0.000 description 4
- 230000008901 benefit Effects 0.000 description 3
- 239000008280 blood Substances 0.000 description 3
- 210000004369 blood Anatomy 0.000 description 3
- 230000002008 hemorrhagic effect Effects 0.000 description 3
- 238000007917 intracranial administration Methods 0.000 description 3
- 210000003657 middle cerebral artery Anatomy 0.000 description 3
- 238000012163 sequencing technique Methods 0.000 description 3
- 210000001519 tissue Anatomy 0.000 description 3
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 2
- 206010053648 Vascular occlusion Diseases 0.000 description 2
- 230000001154 acute effect Effects 0.000 description 2
- 230000003872 anastomosis Effects 0.000 description 2
- 238000013459 approach Methods 0.000 description 2
- 230000003727 cerebral blood flow Effects 0.000 description 2
- 206010008118 cerebral infarction Diseases 0.000 description 2
- 208000026106 cerebrovascular disease Diseases 0.000 description 2
- 239000002872 contrast media Substances 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000018109 developmental process Effects 0.000 description 2
- 238000003384 imaging method Methods 0.000 description 2
- 238000012905 input function Methods 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 210000002569 neuron Anatomy 0.000 description 2
- 230000010412 perfusion Effects 0.000 description 2
- 210000003388 posterior cerebral artery Anatomy 0.000 description 2
- 230000010410 reperfusion Effects 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 208000024891 symptom Diseases 0.000 description 2
- 208000021331 vascular occlusion disease Diseases 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- 206010056292 Androgen-Insensitivity Syndrome Diseases 0.000 description 1
- 201000001320 Atherosclerosis Diseases 0.000 description 1
- 241000036569 Carp sprivivirus Species 0.000 description 1
- 208000005189 Embolism Diseases 0.000 description 1
- 208000016988 Hemorrhagic Stroke Diseases 0.000 description 1
- 206010061216 Infarction Diseases 0.000 description 1
- 206010033799 Paralysis Diseases 0.000 description 1
- 208000007536 Thrombosis Diseases 0.000 description 1
- 206010057469 Vascular stenosis Diseases 0.000 description 1
- 210000002551 anterior cerebral artery Anatomy 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000000740 bleeding effect Effects 0.000 description 1
- 230000008081 blood perfusion Effects 0.000 description 1
- 230000011157 brain segmentation Effects 0.000 description 1
- 210000005013 brain tissue Anatomy 0.000 description 1
- 238000004422 calculation algorithm Methods 0.000 description 1
- 210000000269 carotid artery external Anatomy 0.000 description 1
- 210000004004 carotid artery internal Anatomy 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000001447 compensatory effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 238000002059 diagnostic imaging Methods 0.000 description 1
- 230000008034 disappearance Effects 0.000 description 1
- 201000010099 disease Diseases 0.000 description 1
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 1
- 238000004880 explosion Methods 0.000 description 1
- 239000003527 fibrinolytic agent Substances 0.000 description 1
- 208000034783 hypoesthesia Diseases 0.000 description 1
- 238000003709 image segmentation Methods 0.000 description 1
- 230000007574 infarction Effects 0.000 description 1
- 208000014674 injury Diseases 0.000 description 1
- 208000020658 intracerebral hemorrhage Diseases 0.000 description 1
- 238000001990 intravenous administration Methods 0.000 description 1
- 238000002372 labelling Methods 0.000 description 1
- 238000012417 linear regression Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 230000037361 pathway Effects 0.000 description 1
- 230000035515 penetration Effects 0.000 description 1
- 210000003446 pia mater Anatomy 0.000 description 1
- 238000012805 post-processing Methods 0.000 description 1
- 238000011002 quantification Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000011268 retreatment Methods 0.000 description 1
- 230000004083 survival effect Effects 0.000 description 1
- 229960000103 thrombolytic agent Drugs 0.000 description 1
- 230000008733 trauma Effects 0.000 description 1
- 238000011269 treatment regimen Methods 0.000 description 1
- 238000010200 validation analysis Methods 0.000 description 1
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
- 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
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T9/00—Image coding
- G06T9/002—Image coding using neural networks
-
- 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
-
- 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
- G06T2207/30101—Blood vessel; Artery; Vein; Vascular
-
- 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
- G06T2207/30101—Blood vessel; Artery; Vein; Vascular
- G06T2207/30104—Vascular flow; Blood flow; Perfusion
-
- 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
- G06T2207/30204—Marker
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Biomedical Technology (AREA)
- Software Systems (AREA)
- Bioinformatics & Computational Biology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Evolutionary Biology (AREA)
- Medical Informatics (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Radiology & Medical Imaging (AREA)
- Quality & Reliability (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
- Apparatus For Radiation Diagnosis (AREA)
Abstract
The application provides a CTA full-automatic collateral circulation scoring method and system based on deep learning, and relates to the technical field of medical image processing, wherein the method comprises the following steps: step S1: masking and normalizing the brain CTA image; step S2: performing brain region division on the CTA image, acquiring a related anatomical map and a functional map, and acquiring a mask map of each brain region of the weighted brain; step S3: performing blood vessel segmentation on the mask and the normalized CTA image; step S4: quantitatively calculating scoring characteristics based on the vessel segmentation result; step S5: measuring scoring characteristics of the wall thickness of the blood vessel based on the convolutional neural network; step S6: and constructing a multi-label grading classification model, and grading the calculated feature vectors in a classification way. The application can realize a collateral circulation blood vessel grading mechanism based on a blood flow compensation way, improves the segmentation precision of small blood vessels, and can lead a grading strategy to have wider application range and more objective evaluation threshold.
Description
Technical Field
The application relates to the technical field of medical image processing, in particular to a CTA full-automatic collateral circulation scoring method and system based on deep learning.
Background
Cerebral stroke is currently the leading cause of acquired disability in patients and is one of the leading causes of death worldwide. Most cerebral apoplexy is ischemic apoplexy, mainly cerebral infarction, and the cause of the cerebral apoplexy is that intracranial and extracranial vascular atherosclerosis causes vascular stenosis and occlusion to be the most common. Cerebral apoplexy is a common clinical disease, has acute onset and obvious symptoms, and mainly takes paralysis and numbness of limbs as the main manifestations. Many patients, despite normative treatment, have difficulty avoiding more or less sequelae, which is an important threat to human health. The cerebral apoplexy has high treatment cost, very influences the quality of life and brings great economic pressure to patients and society.
In ischemic cerebrovascular disease, arterial blockage can result in reduced perfusion of the blood supply area of the blocked vessel. If the blood flow perfusion rate is less than 10 ml/(100 g.min), the nerve cells are irreversibly damaged within a few minutes. If only a portion of cerebral blood flow (Cerebral Blood Flow, CBF) is reduced, the nerve cells will stop working, but still remain structurally intact, potentially being considered as a rescueable tissue (ischemic penumbra), which makes them the target for thrombolytic therapy. The cerebral collateral circulation refers to the process that when the blood supply artery of the brain is severely stenosed or blocked, blood reaches ischemic tissues through other branch blood vessels, so that the ischemic areas can obtain different degrees of blood perfusion compensation. Studies have shown that patients with good collateral circulation have larger ischemic penumbra, smaller cerebral infarct size, better thrombolysis and retreatment results. Thus, accurate assessment of cerebral collateral circulation is an important condition for the correct treatment of patients with acute ischemic stroke (Acute Ischaemic Stroke, AIS).
CT Angiography (CTA) is the most common technique for evaluating the intracranial vascular system, has the advantages of rapid imaging, no trauma, simple post-processing of images, and the like, can clearly determine the specific position of vascular occlusion, and is the preferred imaging examination of emergency patients. With this technique, the passage of contrast agent through the brain can be recorded by acquiring dynamic images. To accurately assess the patient's cerebral collateral circulation, arterial input function (Arterial Input Funct ion, AIF) and venous output function (Venous Output Function, VOF) in CTA images need to be measured, as concentration-Time curves (Area Under the Concentration-Time, AUC) for arterial and venous voxels, respectively, containing 100% blood.
Current research in collateral circulation scoring systems has focused mainly on two areas: a rough visual scoring and a quantified threshold scoring system.
The Alberta cerebral stroke early stage CT score (Alberta Stroke Program Early CT Score, aspcts) is a hierarchical system for evaluating the ischemic change of acute ischemic stroke CT, and can evaluate the early focus of ischemic stroke, which is currently used in a plurality of clinical trials for intravascular mechanical thrombus removal. However, since the ASPECT score is manually determined, the ASPECT score is susceptible to subjective factors of the evaluator.
Maas scoring the diseased side branch vessels were compared to normal side branch vessels based on the CTA original image (CTA-SI) and scored. The scoring method is used for evaluating the presence or absence of collateral blood vessels in the collateral fissure area and the pia mater bulge area and grading. But in fact about one quarter of patients on CTA were unable to measure substantial collateral circulation through collateral and leptomeningeal bulge vessels.
The Tan score is based on the collateral vessel filling condition score of the cerebral middle arterial infarction ischemic area, and the score is: 3 (side branch filling up to 100%), 2 (side branch filling up to 50% -100%), 1 (0 < side branch filling up to < 50%), 0 (no side branch blood supply). The branch circulation is poor in 0-1, and the branch circulation is good in 2-3. The scoring threshold for this score was determined by clinical studies and was only applicable to patients who were either intravenously injected with rt-PA thrombolytic agent or who did not receive thrombolytic therapy. For those patients who use other stroke-in treatment regimens, different thresholds may exist, and thus the scoring system has a limit in terms of repeatability.
Visual scoring systems can be affected by subjective interpretation, resulting in discrepancies between and within observers. The scoring threshold of the scoring system quantifying the evaluation criteria is susceptible to differences between different samples, while the automatic scoring system may facilitate objective and repeatable assessment of brain collateral status through deep learning.
Collateral circulation scoring depends on the difference between the occlusion of the middle cerebral artery (Middle Cerebral Artery, MCA) region and the contralateral arterial tree. Vessel segmentation is thus an essential step in the implementation step. Currently, convolutional neural networks (Convolutional Neural Networks, CNN) exhibit excellent performance in many medical vessel image segmentation tasks, which makes the implementation of fully automatic collateral circulation scoring possible. Based on a deep learning method, the difference between the arterial tree of the affected side and the healthy side is quantified, and the efficacy and the efficiency of the intravascular treatment diagnosis opinion of cerebral apoplexy can be effectively improved.
The current document Su Jiahang, et al, "Automatic Collateral Scoring From 3D CTA Images," IEEE transactions on medical imaging 39.6.6 (2020), wherein Su et al studied this direction and proposed using 3D U-net for vessel segmentation and machine learning classifiers to achieve reasoning about scoring thresholds in the Tan scoring system. However, this approach assumes that the cause of ischemic stroke is a large vascular embolism in the brain hemisphere, and thus the scoring system is not applicable to patients not conforming to this assumption, and the patient population was studied to account for 14-30% of the ischemic stroke population.
Disclosure of Invention
Aiming at the defects in the prior art, the application aims to provide a CTA full-automatic collateral circulation scoring method and system based on deep learning.
According to the CTA full-automatic collateral circulation scoring method and system based on deep learning, the scheme is as follows:
in a first aspect, a CTA full-automatic collateral circulation scoring method based on deep learning is provided, the method comprising:
step S1: acquiring a brain CTA image and a corresponding binary mask data set, acquiring a brain artery probability density map and a brain hemisphere map, and performing mask and normalization pretreatment on the brain CTA image;
step S2: performing brain region division on the CTA image, acquiring a related anatomical map and a functional map, and acquiring a mask map of each brain region of the weighted brain;
step S3: performing blood vessel segmentation on the mask and the normalized CTA image;
step S4: quantitatively calculating scoring characteristics based on the vessel segmentation result;
step S5: measuring scoring characteristics of the wall thickness of the blood vessel based on the convolutional neural network;
step S6: and constructing a multi-label grading classification model, and grading the calculated feature vectors in a classification way.
Preferably, the step S1 includes:
skull dissection and midline correction are carried out on the data set;
registering the manually labeled dataset using continuous affine and deformable symmetric differential isomorphic image registration;
carrying out space normalization and obtaining a brain artery probability density map through B spline interpolation transformation;
defining and manually marking a hemispheric brain map in a CTA image space;
correspondingly transforming the binary brain mask, the cerebral artery probability density map and the hemispherical map into each CTA image space for masking;
CTA images were normalized using maximum-minimum normalization.
Preferably, the step S2 includes:
performing anatomical atlas brain region division on the CTA image;
performing functional atlas brain region division on the CTA image;
performing weight assignment on each brain region;
and obtaining a CTA image after weighted brain region division.
Preferably, the step S3 includes:
acquiring a preprocessed CTA image;
expanding each single-pixel central line marked manually by using 3X 3 square structural elements;
inputting the training image into a network model, the model comprising an encoder and a decoder;
the encoder processes the image, each step of operation comprising two 3 x 3 convolutions to extract features and one 2 x 2 pooling with downsampling;
the decoder processes the image, each step comprising a 3 x 3 deconvolution to upsample, splicing feature graphs with the same size from the compression path, and finally carrying out convolution twice;
measuring the performance of the network model by using dice scores, and training;
inputting the test image into a network model;
the network model outputs the segmentation result of the test CTA image.
Preferably, the step S4 includes:
generating a patient side binary drawing and a non-patient side binary drawing based on clinical information and a cerebral hemisphere map;
calculating and weighting the ratio of the blood vessel volume in the volume aspect and the maximum phase time:
wherein p represents a voxel in the brain, H Y Is the brain of the affected side, H N Is for strengthening brain, W (D) i ) Weight value, T, representing three-level vessel partition Y Time phase time representing maximum volume value obtained from affected side, T N Representing the time phase time when the health side obtains the maximum volume value;
calculating a blood flow rate ratio in terms of flow rate;
wherein V is Y And V N Representing the vascular flow rates on the affected side and healthy side, respectively.
Preferably, the step S5 includes:
inputting CTA images of newly built collateral circulation blood vessels, and cutting the images into 16 patches;
each patch is activated by a batch normalization and parameterization correction linear unit after each convolution layer through 5 convolution layers and 1 full connection layer;
all the connection layers of the patch are connected and are connected through an additional all the connection layer;
output through the linear activation output layer;
calculating an adaptive learning rate by using an adaptive moment estimation optimizer, and training a model;
inputting a collateral circulation CTA image to be measured;
outputting the measured value r of the thickness of the collateral circulation blood vessel s ；
Preferably, the step S6 includes:
fusing the characteristic values to construct a characteristic vector f;
calculating the median value of each image feature vector f;
inputting a median value of the feature vector f, and determining a classification threshold according to a Tan scoring standard and clinical experience;
training a support vector classifier, and determining an optimal threshold value of the median of the feature vector f;
inputting a complete feature vector f, and training a random forest classifier;
training the ordered regression of the input complete feature vector;
performing integrated learning on a plurality of models;
testing a scoring threshold value obtained by ensemble learning;
screening a scoring threshold range with high accuracy and sensitivity;
and determining a cerebral collateral grading standard, and grading the collateral circulation CTA image in a classification way.
In a second aspect, a CTA fully automatic collateral circulation scoring system based on deep learning is provided, the system comprising:
module M1: acquiring a brain CTA image and a corresponding binary mask data set, acquiring a brain artery probability density map and a brain hemisphere map, and performing mask and normalization pretreatment on the brain CTA image;
module M2: performing brain region division on the CTA image, acquiring a related anatomical map and a functional map, and acquiring a mask map of each brain region of the weighted brain;
module M3: performing blood vessel segmentation on the mask and the normalized CTA image;
module M4: quantitatively calculating scoring characteristics based on the vessel segmentation result;
module M5: measuring scoring characteristics of the wall thickness of the blood vessel based on the convolutional neural network;
module M6: and constructing a multi-label grading classification model, and grading the calculated feature vectors in a classification way.
Preferably, the module M1 comprises:
skull dissection and midline correction are carried out on the data set;
registering the manually labeled dataset using continuous affine and deformable symmetric differential isomorphic image registration;
carrying out space normalization and obtaining a brain artery probability density map through B spline interpolation transformation;
defining and manually marking a hemispheric brain map in a CTA image space;
correspondingly transforming the binary brain mask, the cerebral artery probability density map and the hemispherical map into each CTA image space for masking;
CTA images were normalized using maximum-minimum normalization.
Preferably, the module M2 comprises:
performing anatomical atlas brain region division on the CTA image;
performing functional atlas brain region division on the CTA image;
performing weight assignment on each brain region;
and obtaining a CTA image after weighted brain region division.
Compared with the prior art, the application has the following beneficial effects:
1. according to the application, the brain functional area is divided, and different areas are weighted, so that a collateral circulation blood vessel grading mechanism based on a blood flow compensation way is realized, and the segmentation precision of small blood vessels is improved;
2. according to the application, the blood vessel volume, the weighted volume and the blood vessel flow velocity are taken as characteristics to obtain the volume and flow velocity ratio of the affected side to the healthy side in each period, so that the scoring strategy has a wider application range and a more objective evaluation threshold.
Drawings
Other features, objects and advantages of the present application will become more apparent upon reading of the detailed description of non-limiting embodiments, given with reference to the accompanying drawings in which:
FIG. 1 is a flow chart of a brain collateral circulation scoring method based on deep learning according to an embodiment of the application;
FIG. 2 is a schematic diagram of a U-net network according to an embodiment of the present application;
fig. 3 is a flow chart illustrating a method for scoring a cerebral collateral circulation according to an embodiment of the present application.
Detailed Description
The present application will be described in detail with reference to specific examples. The following examples will assist those skilled in the art in further understanding the present application, but are not intended to limit the application in any way. It should be noted that variations and modifications could be made by those skilled in the art without departing from the inventive concept. These are all within the scope of the present application.
The embodiment of the application provides a CTA full-automatic collateral circulation scoring method based on deep learning, which is shown by referring to FIG. 1 and specifically comprises the following steps:
step S1: the method comprises the steps of acquiring brain CTA images and corresponding binary mask data sets, acquiring brain artery probability density maps and brain hemisphere maps, and performing mask and normalization pretreatment on the brain CTA images. Specifically, the manually labeled dataset is first skull stripped and midline corrected using continuous affine and deformable symmetric differential isomorphic image registration. Secondly, carrying out space normalization, obtaining a brain artery probability density map through B spline interpolation transformation, defining a brain hemisphere map which is marked manually in a CTA image space, correspondingly transforming a binary brain mask, the brain artery probability density map and the hemisphere map into each CTA image space to mask, and finally normalizing the CTA image by using maximum and minimum value normalization. The B-spline interpolation described in this embodiment is very widely applied, where the function properties are also symmetrical, and the coefficient matrix is easily obtained by matrix inversion.
Step S2: brain regions were segmented from the CTA images, automated Anatomical Labeling (AAL) anatomical atlas, eichhoff-Zilles (EZ) anatomical atlas, talairach and Tournoux (TT) anatomical atlas, harvard-Oxford (HO) anatomical atlas, and CC200 and CC400 functional atlas were obtained, and a weighted brain region mask map was obtained. Specifically, anatomical atlas brain region division is performed on a CTA image, functional atlas brain region division is performed on the CTA image, weight assignment is performed on each brain region, and finally the CTA image after weighted brain region division is obtained.
Referring to fig. 2 and 3, step S3: and performing blood vessel segmentation on the mask and the normalized CTA image.
Wherein, fig. 2 is a schematic structural diagram of a U-net network model in the present application, the U-net network model includes: the encoder and the decoder are mainly convolution layers, the decoder is an up-sampling layer, and the feature map obtained by each convolution layer is connected to the corresponding up-sampling layer, and the standard Dice fraction is used as a loss function. The root mean square propagation optimizer was chosen with an initial learning rate of 0.1, halving it every 10 epochs, which was set to 600. The annotation dataset was randomly split into training and validation datasets, with a ratio of 3:1.
Specifically, a preprocessed CTA image is acquired first, and expanding each single-pixel central line marked by manpower by using 3X 3 square structural elements, and inputting the training image into a network model. The encoder processes the image, each step of operation comprising two 3 x 3 convolutions to extract features and one 2 x 2 pooling with Downsampling (Downsampling); the decoder processes the image, each step comprising a 3 x 3 deconvolution to upsample (Upsampling), concatenating the same-sized Feature maps (Feature Map) from the compressed path, and finally performing the convolution twice. And measuring the performance of the network model by using the dice fraction, training, inputting the test image into the network model, and finally outputting the segmentation result of the test CTA image by the network model.
Step S4: scoring features are quantitatively calculated based on the vessel segmentation results. The method comprises the following steps:
based on the clinical information and the cerebral hemisphere map, a patient side binary drawing and a non-patient side binary drawing are generated.
Calculating and weighting the ratio of the blood vessel volume in the volume aspect and the maximum phase time:
wherein p represents a voxel in the brain, H Y Is the brain of the affected side, H N Is for strengthening brain, W (D) i ) Weight value, T, representing three-level vessel partition Y Representative ofTime phase time when the affected side obtains maximum volume value, T N Representing the time phase at which the healthy side obtained the maximum volume value.
Calculating a blood flow rate ratio in terms of flow rate:
wherein V is Y And V N Representing the vascular flow rates on the affected side and healthy side, respectively.
Step S5: the scoring characteristics of the vessel wall thickness are measured based on a convolutional neural network. Specifically, a CTA image of a newly built collateral circulation blood vessel is input first, the image is cut into 16 patches, each patch is subjected to Batch Normalization (BN) and parameterized modified linear unit (PReLU) activation after passing through 5 convolution layers and 1 full connection layer, the full connection layers of each patch are connected, the full connection layers are output through an additional full connection layer and a linear activation output layer, an adaptive moment estimation (Adam) optimizer is used for calculating the self-adaptive learning rate, and a model is trained. Inputting a collateral circulation CTA image to be measured, and finally outputting a measured value r of the collateral circulation blood vessel thickness s 。
Step S6: and constructing a multi-label grading classification model, and grading the calculated feature vectors in a classification way. The method comprises the following steps: and fusing the characteristic values, constructing a characteristic vector f, calculating the median value of the characteristic vector f of each image, inputting the median value of the characteristic vector f, and determining a classification threshold according to the Tan scoring standard and clinical experience. Training a support vector classifier, determining an optimal threshold value of a median value of a feature vector f, inputting the complete feature vector f, training a random forest classifier, training ordered regression of the complete feature vector, performing ensemble learning on four models, testing scoring thresholds obtained by ensemble learning, screening a scoring threshold value range with high accuracy and sensitivity, finally determining a brain collateral scoring standard, and classifying and scoring collateral circulation CTA images.
Next, the present application will be described in more detail.
The application provides a CTA full-automatic collateral circulation scoring method based on deep learning, which comprises the following steps:
referring to fig. 1 and 2, brain CTA images are preprocessed:
acquiring brain CTA images and corresponding binary mask datasets; identifying venous late stage of the arterial peak period and the venous peak period 2 days after the arterial peak period according to AIF and VOF, and performing skull stripping and midline correction; acquiring a cerebral artery probability density map M and a cerebral hemisphere map H; and correspondingly transforming the binary brain mask, the cerebral artery probability density map M and the hemispheric map H into each CTA image space to mask, and normalizing the CTA images by using maximum and minimum normalization.
Brain segmentation was performed on CTA images:
and (3) performing brain region division according to the anatomical map and the functional map by manually marking on CTA axial images (combined with coronal positions) of healthy people to obtain each brain region mask. And giving weight to each brain region according to the importance degree of each brain region, and carrying out weighted division on the CTA image by using the weighted brain region mask.
Referring to fig. 2 and 3, the mask and normalized CTA image are vessel segmented:
the vessels were segmented hierarchically using a classical 3D U-net network structure with depth 5, the model comprising an encoder and a decoder. Adding a residual error module in an encoder path, avoiding gradient explosion caused by the deep penetration of the encoder, and obtaining a deeper characteristic value; in the cascade decoder path, a depth supervision module is added to avoid information loss and gradient disappearance in each convolution layer. The input data of the model is a cube of 64 x 64 voxels, a classifier is used to output a vessel classification map U of three-dimensional voxels, the blood vessels are divided into class I side branches, class II side branches and class III side branches according to the blood flow compensatory pathway. Wherein the class I side branch is an anastomotic arterial ring (Willis ring of Willis) located at the bottom of the brain; the class II side branch comprises an intracranial pia artery anastomosis channel between the distal ends of anterior, middle and posterior cerebral arteries, between the posterior cerebral artery and the main cerebral artery, and an anastomosis channel between an external carotid artery branch and an internal carotid artery branch; class III side branches refer to newly formed capillaries around ischemic tissue.
Calculating scoring characteristics based on quantification of vessel segmentation results:
the artificial subjective scoring of the collateral brain circulation is based on comparing the number of visible blood vessels on the affected side and the non-affected side, and the fully automatic scoring method adopts a similar strategy: the blood vessel grading diagram U output by the U-net model is comprehensively used, and the corresponding cerebral artery probability density diagram M and the corresponding cerebral hemisphere diagram H are used for comparing and determining the suffering side and the non-suffering side of the brain. Based on the known affected hemisphere of clinical symptoms, based on the information and the brain hemisphere diagram H, an affected hemisphere binary diagram H is generated Y And non-affected side binary pattern H N . Based on the images, the specific numerical values of the blood vessel volume, the weighted volume, the phase-to-phase time ratio of the obtained volume maximum value, the blood vessel flow rate and the like are obtained, and the four ratios of the volumes and the flow rates of the affected side and the non-affected side in each period are calculated to be used as scoring characteristics, so that the problem that the ASPECTS needs to manually obtain the scoring characteristics is solved:
1. volume: the number of blood vessels on the affected side and the non-affected side are obviously different, and the graph H is drawn through secondary therapy on the affected side Y And non-affected side binary pattern H N The vessel volume ratio of each brain region (R1-R8) was obtained and graded. It is assumed that in the unaffected hemisphere the ratio of the volume of the blood vessel (the affected volume divided by the unaffected volume) should be a value between 0 (no blood vessels are visible on the affected side) and 1 (the same number of blood vessels are visible on both sides). Summing the maximum values of the blood vessel volumes of the affected side and the non-affected side, and weighting according to the weight value of the three-level blood vessel partition to obtain r respectively V And r w 。
Wherein p represents a voxel in the brain, H Y Is the brain of the affected side, H N Is for strengthening brain, W (D) i ) Weight value, T, representing three-level vessel partition Y Maximum volume value is obtained on the affected sideTime phase time of T N Representing the time phase at which the healthy side obtained the maximum volume value.
2. Phase: CTA requires accurate arterial phase, typically measured using triggering techniques or pre-emphasis. In the blood supply rich region, contrast agent is easy to accumulate, and the time of the maximum development volume of the affected side and the non-affected side is different due to the blood flow rate and other reasons, so the ratio r of the time T As a feature.
3. Flow rate: there is a significant difference in blood flow rates between the patient and non-patient sides. By means of the ratio r of blood flow rate of blood vessels on the affected side to blood flow rate of blood vessels on the non-affected side v As a feature.
Wherein V is Y And V N Representing the vascular flow rates on the affected side and healthy side, respectively.
Measuring a blood vessel wall thickness scoring characteristic based on a convolutional neural network:
reperfusion is important for survival of ischemic penumbra brain tissue, but reperfusion may induce hemorrhagic transformation. Post-thrombolytic hemorrhagic transformation (Hemorrhagic transformation, HT) refers to hemorrhagic stroke that occurs over a period of time after intravenous thrombolytic therapy in acute ischemic stroke patients. Because the newly built side branch circulation blood vessel wall is not sound in development, the side branch circulation blood vessel wall is impacted by blood to cause bleeding during recanalization. The application measures the thickness of the side branch circulation blood vessel wall through the convolution neural network and takes the thickness as one characteristic of the side branch circulation score.
The network cuts the picture into 16 patches, each patch passes through 5 convolution layers and 1 full connection layer, the full connection layers of each patch are connected and pass through the additional full connection layer, and finally output through the linear activation output layer. All layers except the output layer normalize each batch of data by parameterized modified linear unit (prime lu, parametric Rectified Linear Unit) activation and batch normalization (BN, batch Normalization) and calculate the adaptive learning rate for each parameter using an adaptive moment estimation (Adam) optimizer.
Based on the network, the wall thickness of the blood vessel thinner than the voxels can be accurately measured, thereby realizing the measurement of the wall thickness of the collateral circulation blood vessel, and the thickness is characterized by r s 。
Multi-label scoring classification model:
using the eigenvector f= [ r ] V ,r w ,r T ,r v ，r s ]As input predictive Tan scores (0, 1,2, 3), four methods for determining score type thresholds are proposed for integrated learning, namely, clinically defined thresholds, support vector classifiers, random forest classifiers and ordered regression classifiers, and threshold screening is performed according to test results of each classifier after training is completed.
1. Clinical definition threshold: the median value of the feature vector f is used as input to determine the classification threshold using a clinical definition of the brain collateral circulation score. The side branch states are ranked according to Tan scoring criteria based on general clinical knowledge and experience in order to provide a reliable side branch scoring reference standard for the present application.
2. Support vector classifier: the median value of the feature vector f is input, and an optimal threshold value of the median value of the feature vector f is determined using a Support Vector Classifier (SVC) with a linear kernel. The feature vector median classification is maximized using the spacing of SVCs. The generalization capability of the learning machine is improved by seeking to minimize the structuring risk, and the minimization of experience risk and confidence range is realized, so that the aim of obtaining good statistical rules under the condition of less statistical sample size is fulfilled.
3. Random forest classifier: random forests are an algorithm that integrates multiple trees through the idea of ensemble learning, which has the advantage of being able to process input samples with high-dimensional features, without dimension reduction, and to evaluate the importance of each feature on classification problems, thus using the complete feature vector f as input.
4. Ordered regression classifier: the feature vector generated by the present application is essentially a sequenced variable, i.e., between a continuous variable and a constant class variable, separated into different classes in relative order at the measurement level, but not continuous. If a multi-classification model is used on the sequencing variables, the inherent ordering of the data may be disregarded, resulting in a loss of ordering information, such that the statistics lose statistical efficiency by missing ordering information. If a common linear regression model is used, then treating the sequenced variables as continuous variables results in artificial information expansion. Thus, for sequencing variables, a corresponding model needs to be employed to fit the two-way nature of the sequencing variables, the most common approach being an ordered regression model. Ordered classification predictions are thus made with ordered regression of the input complete feature vectors.
The embodiment of the application provides a deep learning-based CTA full-automatic collateral circulation scoring method, wherein brain functional areas are divided, different areas are weighted, a collateral circulation blood vessel grading mechanism based on a blood flow compensation way is realized, and the segmentation precision of small blood vessels is improved; meanwhile, the blood vessel volume, the weighted volume and the blood vessel flow velocity are taken as characteristics to obtain the volume and flow velocity ratio of the affected side to the healthy side in each period, so that the scoring strategy has wider application range and more objective evaluation threshold.
Those skilled in the art will appreciate that the application provides a system and its individual devices, modules, units, etc. that can be implemented entirely by logic programming of method steps, in addition to being implemented as pure computer readable program code, in the form of logic gates, switches, application specific integrated circuits, programmable logic controllers, embedded microcontrollers, etc. Therefore, the system and various devices, modules and units thereof provided by the application can be regarded as a hardware component, and the devices, modules and units for realizing various functions included in the system can also be regarded as structures in the hardware component; means, modules, and units for implementing the various functions may also be considered as either software modules for implementing the methods or structures within hardware components.
The foregoing describes specific embodiments of the present application. It is to be understood that the application is not limited to the particular embodiments described above, and that various changes or modifications may be made by those skilled in the art within the scope of the appended claims without affecting the spirit of the application. The embodiments of the application and the features of the embodiments may be combined with each other arbitrarily without conflict.
Claims (10)
1. A deep learning-based CTA full-automatic collateral circulation scoring method is characterized by comprising the following steps:
step S1: acquiring a brain CTA image and a corresponding binary mask data set, acquiring a brain artery probability density map and a brain hemisphere map, and performing mask and normalization pretreatment on the brain CTA image;
step S2: performing brain region division on the CTA image, acquiring a related anatomical map and a functional map, and acquiring a mask map of each brain region of the weighted brain;
step S3: performing blood vessel segmentation on the mask and the normalized CTA image;
step S4: quantitatively calculating scoring characteristics based on the vessel segmentation result;
step S5: measuring scoring characteristics of the wall thickness of the blood vessel based on the convolutional neural network;
step S6: and constructing a multi-label scoring classification model, calculating the feature vector of each CTA image, and classifying and scoring the feature vector.
2. The deep learning-based CTA fully automatic collateral circulation scoring method of claim 1, wherein said step S1 comprises:
step S1.1: skull dissection and midline correction are carried out on the data set;
step S1.2: registering the manually labeled dataset using continuous affine and deformable symmetric differential isomorphic image registration;
step S1.3: carrying out space normalization and obtaining a brain artery probability density map through B spline interpolation transformation;
step S1.4: defining and manually marking a hemispheric brain map in a CTA image space;
step S1.5: correspondingly transforming the binary brain mask, the cerebral artery probability density map and the hemispherical map into each CTA image space for masking;
step S1.6: CTA images were normalized using maximum-minimum normalization.
3. The deep learning-based CTA fully automatic collateral circulation scoring method of claim 1, wherein said step S2 comprises:
step S2.1: performing anatomical atlas brain region division on the CTA image;
step S2.2: performing functional atlas brain region division on the CTA image;
step S2.3: performing weight assignment on each brain region;
step S2.4: and obtaining a CTA image after weighted brain region division.
4. The deep learning-based CTA fully automatic collateral circulation scoring method of claim 1, wherein said step S3 comprises:
step S3.1: acquiring a preprocessed CTA image;
step S3.2: expanding each single-pixel central line marked manually by using 3X 3 square structural elements;
step S3.3: inputting the training image into a network model, the model comprising an encoder and a decoder;
step S3.4: the encoder processes the image, each step of operation comprising two 3 x 3 convolutions to extract features and one 2 x 2 pooling with downsampling;
step S3.5: the decoder processes the image, each step comprising a 3 x 3 deconvolution to upsample, splicing feature graphs with the same size from the compression path, and finally carrying out convolution twice;
step S3.6: measuring the performance of the network model by using dice scores, and training;
step S3.7: inputting the test image into a network model;
step S3.8: the network model outputs the segmentation result of the test CTA image.
5. The deep learning-based CTA fully automatic collateral circulation scoring method of claim 1, wherein said step S4 comprises:
step S4.1: generating a patient side binary drawing and a non-patient side binary drawing based on clinical information and a cerebral hemisphere map;
step S4.2: calculating and weighting the ratio of the blood vessel volume in the volume aspect and the maximum phase time:
wherein,represents a voxel in the brain, < >>Is the brain on the affected side, is->Is effective in strengthening brain, and relieving side effects of%>Weight value representing a tertiary vessel partition, < ->Time phase representing the maximum volume value obtained from the affected side, +.>Representing the time phase time when the health side obtains the maximum volume value;
step S4.3: calculating a blood flow rate ratio in terms of flow rate;
wherein,and->Representing the vascular flow rates on the affected side and healthy side, respectively.
6. The deep learning-based CTA fully automatic collateral circulation scoring method of claim 1, wherein said step S5 comprises:
step S5.1: inputting CTA images of newly built collateral circulation blood vessels, and cutting the images into 16 patches;
step S5.2: each patch is activated by a batch normalization and parameterization correction linear unit after each convolution layer through 5 convolution layers and 1 full connection layer;
step S5.3: all the connection layers of the patch are connected and are connected through an additional all the connection layer;
step S5.3: output through the linear activation output layer;
step S5.4: calculating an adaptive learning rate by using an adaptive moment estimation optimizer, and training a model;
step S5.5: inputting a collateral circulation CTA image to be measured;
step S5.5: outputting the measured value of the thickness of the collateral circulation blood vessel。
7. The deep learning-based CTA fully automatic collateral circulation scoring method of claim 1, wherein said step S6 comprises:
step S6.1: fusing the characteristic values to construct a characteristic vector f;
step S6.2: calculating the median value of each image feature vector f;
step S6.3: inputting a median value of the feature vector f, and determining a classification threshold according to a Tan scoring standard and clinical experience;
step S6.4: training a support vector classifier, and determining an optimal threshold value of the median of the feature vector f;
step S6.5: inputting a complete feature vector f, and training a random forest classifier;
step S6.6: training the ordered regression of the input complete feature vector;
step S6.7: performing integrated learning on a plurality of models;
step S6.8: testing a scoring threshold value obtained by ensemble learning;
step S6.9: screening a scoring threshold range with high accuracy and sensitivity;
step S6.10: and determining a cerebral collateral grading standard, and grading the collateral circulation CTA image in a classification way.
8. A CTA full-automatic collateral circulation scoring system based on deep learning, comprising:
module M1: acquiring a brain CTA image and a corresponding binary mask data set, acquiring a brain artery probability density map and a brain hemisphere map, and performing mask and normalization pretreatment on the brain CTA image;
module M2: performing brain region division on the CTA image, acquiring a related anatomical map and a functional map, and acquiring a mask map of each brain region of the weighted brain;
module M3: performing blood vessel segmentation on the mask and the normalized CTA image;
module M4: quantitatively calculating scoring characteristics based on the vessel segmentation result;
module M5: measuring scoring characteristics of the wall thickness of the blood vessel based on the convolutional neural network;
module M6: and constructing a multi-label scoring classification model, calculating the feature vector of each CTA image, and classifying and scoring the feature vector.
9. The deep learning based CTA fully automatic collateral circulation scoring system of claim 8, wherein said module M1 comprises:
skull dissection and midline correction are carried out on the data set;
registering the manually labeled dataset using continuous affine and deformable symmetric differential isomorphic image registration;
carrying out space normalization and obtaining a brain artery probability density map through B spline interpolation transformation;
defining and manually marking a hemispheric brain map in a CTA image space;
correspondingly transforming the binary brain mask, the cerebral artery probability density map and the hemispherical map into each CTA image space for masking;
CTA images were normalized using maximum-minimum normalization.
10. The deep learning based CTA fully automatic collateral circulation scoring system of claim 8, wherein said module M2 comprises:
performing anatomical atlas brain region division on the CTA image;
performing functional atlas brain region division on the CTA image;
performing weight assignment on each brain region;
and obtaining a CTA image after weighted brain region division.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110286102.4A CN112907563B (en) | 2021-03-17 | 2021-03-17 | CTA full-automatic collateral circulation scoring method and system based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110286102.4A CN112907563B (en) | 2021-03-17 | 2021-03-17 | CTA full-automatic collateral circulation scoring method and system based on deep learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112907563A CN112907563A (en) | 2021-06-04 |
| CN112907563B true CN112907563B (en) | 2023-12-01 |
Family
ID=76105471
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110286102.4A Active CN112907563B (en) | 2021-03-17 | 2021-03-17 | CTA full-automatic collateral circulation scoring method and system based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112907563B (en) |
Families Citing this family (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113610800A (en) * | 2021-08-05 | 2021-11-05 | 西门子数字医疗科技(上海)有限公司 | Device for assessing collateral circulation, non-diagnostic method and electronic apparatus |
| CN113963001B (en) * | 2021-10-25 | 2023-05-23 | 中国人民解放军总医院第一医学中心 | Side branch circulation vessel quantification method based on multi-phase CTA |
| CN114119519A (en) * | 2021-11-16 | 2022-03-01 | 高峰 | Collateral circulation assessment method |
| CN114332043B (en) * | 2021-12-31 | 2022-12-13 | 深圳市铱硙医疗科技有限公司 | CT perfusion image-based collateral circulation measurement method, device, equipment and medium |
| CN114463288B (en) * | 2022-01-18 | 2023-01-10 | 深圳市铱硙医疗科技有限公司 | Brain medical image scoring method and device, computer equipment and storage medium |
| CN117291281B (en) * | 2023-08-29 | 2024-08-20 | 强联智创（北京）科技有限公司 | Method for training evaluation model for evaluating brain collateral circulation and related product |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP3316217A1 (en) * | 2016-10-31 | 2018-05-02 | Siemens Healthcare GmbH | Deep learning based bone removal in computed tomography angiography |
| CN108629784A (en) * | 2018-05-08 | 2018-10-09 | 上海嘉奥信息科技发展有限公司 | A kind of CT image intracranial vessel dividing methods and system based on deep learning |
| CN110934608A (en) * | 2019-10-31 | 2020-03-31 | 上海杏脉信息科技有限公司 | Cerebral apoplexy early-stage CTA image evaluation system and method and readable storage medium |
| CN111369525A (en) * | 2020-03-02 | 2020-07-03 | 联影智能医疗科技(北京)有限公司 | Image analysis method, apparatus and storage medium |
| CN111415361A (en) * | 2020-03-31 | 2020-07-14 | 浙江大学 | Method and device for estimating brain age of fetus and detecting abnormality based on deep learning |
| WO2020144374A1 (en) * | 2019-01-11 | 2020-07-16 | Pie Medical Imaging Bv | Methods and systems for dynamic coronary roadmapping |
| WO2020172188A1 (en) * | 2019-02-19 | 2020-08-27 | Cedars-Sinai Medical Center | Systems and methods for calcium-free computed tomography angiography |
| CN111986101A (en) * | 2020-07-09 | 2020-11-24 | 浙江工业大学 | Cerebrovascular map construction method |
| CN112101523A (en) * | 2020-08-24 | 2020-12-18 | 复旦大学附属华山医院 | Stroke risk screening method and system for CBCT (cone beam computed tomography) image cross-modal prediction CTA (computed tomography angiography) image based on deep learning |
| CN112419282A (en) * | 2020-11-26 | 2021-02-26 | 中国科学院苏州生物医学工程技术研究所 | Automatic detection method and system for aneurysm in brain medical image |
| US10950337B1 (en) * | 2018-11-29 | 2021-03-16 | Heuron Co., Ltd. | Aspect score estimating system and method |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10580131B2 (en) * | 2017-02-23 | 2020-03-03 | Zebra Medical Vision Ltd. | Convolutional neural network for segmentation of medical anatomical images |
| WO2018172990A1 (en) * | 2017-03-24 | 2018-09-27 | Pie Medical Imaging B.V. | Method and system for assessing vessel obstruction based on machine learning |
| US11436732B2 (en) * | 2019-03-12 | 2022-09-06 | The General Hospital Corporation | Automatic segmentation of acute ischemic stroke lesions in computed tomography data |
-
2021
- 2021-03-17 CN CN202110286102.4A patent/CN112907563B/en active Active
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP3316217A1 (en) * | 2016-10-31 | 2018-05-02 | Siemens Healthcare GmbH | Deep learning based bone removal in computed tomography angiography |
| CN108629784A (en) * | 2018-05-08 | 2018-10-09 | 上海嘉奥信息科技发展有限公司 | A kind of CT image intracranial vessel dividing methods and system based on deep learning |
| US10950337B1 (en) * | 2018-11-29 | 2021-03-16 | Heuron Co., Ltd. | Aspect score estimating system and method |
| WO2020144374A1 (en) * | 2019-01-11 | 2020-07-16 | Pie Medical Imaging Bv | Methods and systems for dynamic coronary roadmapping |
| WO2020172188A1 (en) * | 2019-02-19 | 2020-08-27 | Cedars-Sinai Medical Center | Systems and methods for calcium-free computed tomography angiography |
| CN110934608A (en) * | 2019-10-31 | 2020-03-31 | 上海杏脉信息科技有限公司 | Cerebral apoplexy early-stage CTA image evaluation system and method and readable storage medium |
| CN111369525A (en) * | 2020-03-02 | 2020-07-03 | 联影智能医疗科技(北京)有限公司 | Image analysis method, apparatus and storage medium |
| CN111415361A (en) * | 2020-03-31 | 2020-07-14 | 浙江大学 | Method and device for estimating brain age of fetus and detecting abnormality based on deep learning |
| CN111986101A (en) * | 2020-07-09 | 2020-11-24 | 浙江工业大学 | Cerebrovascular map construction method |
| CN112101523A (en) * | 2020-08-24 | 2020-12-18 | 复旦大学附属华山医院 | Stroke risk screening method and system for CBCT (cone beam computed tomography) image cross-modal prediction CTA (computed tomography angiography) image based on deep learning |
| CN112419282A (en) * | 2020-11-26 | 2021-02-26 | 中国科学院苏州生物医学工程技术研究所 | Automatic detection method and system for aneurysm in brain medical image |
Non-Patent Citations (3)
| Title |
|---|
| Fully automated detection and segmentation of intracranial aneurysms in subarachnoid hemorrhage on CTA using deep learning;Rahil Shahzad et.al;《nature》;全文 * |
| 基于深度学习的脑部CTA图像血管分割方法研究;刘国玮;《中国优秀硕士学位论文全文数据库（信息科技辑）》(第03期);全文 * |
| 基于深度学习的计算机体层摄影血管造影颈动脉斑块分割初步研究;吴秋雯等;《上海医学》;第43卷(第05期);280-283 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112907563A (en) | 2021-06-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112907563B (en) | CTA full-automatic collateral circulation scoring method and system based on deep learning | |
| US20220284584A1 (en) | Computerised tomography image processing | |
| Chen et al. | 3D intracranial artery segmentation using a convolutional autoencoder | |
| Francia et al. | Chaining a U-net with a residual U-net for retinal blood vessels segmentation | |
| CN114926477B (en) | Brain tumor multi-mode MRI image segmentation method based on deep learning | |
| CN110415234A (en) | Brain tumor dividing method based on multi-parameter magnetic resonance imaging | |
| Tursynova et al. | Deep learning-enabled brain stroke classification on computed tomography images | |
| Wang et al. | Automatic real-time CNN-based neonatal brain ventricles segmentation | |
| Šušteršič et al. | A deep learning model for automatic detection and classification of disc herniation in magnetic resonance images | |
| Savaş et al. | Comparison of deep learning models in carotid artery Intima-Media thickness ultrasound images: CAIMTUSNet | |
| CN116843667A (en) | Brain image classification and brain cognition scoring prediction method based on multitask learning | |
| Zhang et al. | Machine learning model comparison for automatic segmentation of intracoronary optical coherence tomography and plaque cap thickness quantification | |
| Gayathri et al. | Fetal brain segmentation using improved maximum entropy threshold | |
| CN117853505A (en) | Pancreatic CT image segmentation method based on Unet and U2net cascade convolutional neural network | |
| GB2599682A (en) | Aortic aneurysm growth rate prediction from geometric analysis | |
| Lu et al. | Data enhancement and deep learning for bone age assessment using the standards of skeletal maturity of hand and wrist for chinese | |
| AU3720700A (en) | 3-d shape measurements using statistical curvature analysis | |
| Abid et al. | Detection of lungs cancer through computed tomographic images using deep learning | |
| Veeturi et al. | Radiomics Features on MRI are Associated with High Risk Intracranial Aneurysms | |
| Apaydin et al. | Deep convolutional neural networks using U-net for automatic intervertebral disc segmentation in axial MRI | |
| Nugroho et al. | Thyroid nodules categorization based on margin features using deep learning | |
| Banerjee et al. | Segmentation of intracranial aneurysm remnant in MRA using dual-attention atrous net | |
| Mahyoub et al. | Brain Tumor Segmentation in Fluid-Attenuated Inversion Recovery Brain MRI using Residual Network Deep Learning Architectures | |
| Rajendran et al. | An Ensemble Deep Learning Network in Classifying the Early CT Slices of Ischemic Stroke Patients. | |
| Al Sariera et al. | Automated Cataract Detection and Classification Using Random Forest Classifier in Fundus 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 |