CN112330674A - Self-adaptive variable-scale convolution kernel method based on brain MRI (magnetic resonance imaging) three-dimensional image confidence - Google Patents

Self-adaptive variable-scale convolution kernel method based on brain MRI (magnetic resonance imaging) three-dimensional image confidence Download PDF

Info

Publication number
CN112330674A
CN112330674A CN202011441419.2A CN202011441419A CN112330674A CN 112330674 A CN112330674 A CN 112330674A CN 202011441419 A CN202011441419 A CN 202011441419A CN 112330674 A CN112330674 A CN 112330674A
Authority
CN
China
Prior art keywords
convolution kernel
encoding direction
confidence
calculating
scale
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.)
Granted
Application number
CN202011441419.2A
Other languages
Chinese (zh)
Other versions
CN112330674B (en
Inventor
吴佳胜
胡凯
郑翡
陈炜峰
王丽华
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Nanjing University of Information Science and Technology
Original Assignee
Nanjing University of Information Science and Technology
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Nanjing University of Information Science and Technology filed Critical Nanjing University of Information Science and Technology
Publication of CN112330674A publication Critical patent/CN112330674A/en
Application granted granted Critical
Publication of CN112330674B publication Critical patent/CN112330674B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G06T7/0012—Biomedical image inspection
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B60—VEHICLES IN GENERAL
    • B60L—PROPULSION OF ELECTRICALLY-PROPELLED VEHICLES; SUPPLYING ELECTRIC POWER FOR AUXILIARY EQUIPMENT OF ELECTRICALLY-PROPELLED VEHICLES; ELECTRODYNAMIC BRAKE SYSTEMS FOR VEHICLES IN GENERAL; MAGNETIC SUSPENSION OR LEVITATION FOR VEHICLES; MONITORING OPERATING VARIABLES OF ELECTRICALLY-PROPELLED VEHICLES; ELECTRIC SAFETY DEVICES FOR ELECTRICALLY-PROPELLED VEHICLES
    • B60L3/00—Electric devices on electrically-propelled vehicles for safety purposes; Monitoring operating variables, e.g. speed, deceleration or energy consumption
    • B60L3/0023—Detecting, eliminating, remedying or compensating for drive train abnormalities, e.g. failures within the drive train
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B60—VEHICLES IN GENERAL
    • B60L—PROPULSION OF ELECTRICALLY-PROPELLED VEHICLES; SUPPLYING ELECTRIC POWER FOR AUXILIARY EQUIPMENT OF ELECTRICALLY-PROPELLED VEHICLES; ELECTRODYNAMIC BRAKE SYSTEMS FOR VEHICLES IN GENERAL; MAGNETIC SUSPENSION OR LEVITATION FOR VEHICLES; MONITORING OPERATING VARIABLES OF ELECTRICALLY-PROPELLED VEHICLES; ELECTRIC SAFETY DEVICES FOR ELECTRICALLY-PROPELLED VEHICLES
    • B60L3/00—Electric devices on electrically-propelled vehicles for safety purposes; Monitoring operating variables, e.g. speed, deceleration or energy consumption
    • B60L3/12—Recording operating variables ; Monitoring of operating variables
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/082—Learning methods modifying the architecture, e.g. adding, deleting or silencing nodes or connections
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T9/00—Image coding
    • 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
    • G06T2207/10088—Magnetic resonance imaging [MRI]
    • 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
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
    • Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
    • Y02A90/30—Assessment of water resources

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • General Physics & Mathematics (AREA)
  • General Health & Medical Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • General Engineering & Computer Science (AREA)
  • Evolutionary Computation (AREA)
  • Mechanical Engineering (AREA)
  • Transportation (AREA)
  • Power Engineering (AREA)
  • Sustainable Energy (AREA)
  • Artificial Intelligence (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Sustainable Development (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Software Systems (AREA)
  • Mathematical Physics (AREA)
  • Medical Informatics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
  • Multimedia (AREA)
  • Radiology & Medical Imaging (AREA)
  • Quality & Reliability (AREA)
  • Magnetic Resonance Imaging Apparatus (AREA)

Abstract

The invention discloses a self-adaptive variable-scale convolution kernel method based on confidence coefficient of brain MRI three-dimensional image, belonging to the technical field of information and automatic control, which automatically changes the size of a convolution kernel in real time to better extract features. Better segmentation of MRI tumor images.

Description

Self-adaptive variable-scale convolution kernel method based on brain MRI (magnetic resonance imaging) three-dimensional image confidence
Technical Field
The invention belongs to the technical field of information and automatic control, and particularly relates to a self-adaptive variable-scale convolution kernel method based on brain MRI (magnetic resonance imaging) three-dimensional image confidence.
Background
Brain tumors have a high mortality rate and, once diagnosed, the patient typically has two or three years of life. Magnetic Resonance Imaging (MRI) is widely used for the imaging judgment of various systems of the human body, and doctors can diagnose the disease condition through magnetic resonance images so as to discover and treat the disease condition as early as possible.
The method has the advantages that the brain MRI tumor image can be detected quickly, automatically and accurately, and the method plays a vital role in tumor diagnosis. Although many deep learning networks such as Full Convolution Network (FCN) can detect tumor regions in MRI brain images, since the size and number of convolution layers in such networks are fixed, extracting features by such an invariant convolution kernel inevitably loses some detailed information after passing through a plurality of convolution layers or extracts much useless information, which wastes computational resources. The method cannot achieve rapidness, accuracy and rapidness and accuracy in automatically segmenting the brain MRI tumor, and even can influence a doctor to make wrong judgment.
Disclosure of Invention
The invention aims to solve the technical problem of an adaptive variable-scale convolution kernel method based on brain MRI three-dimensional image confidence. Because the gray scale ranges of images obtained by different patients and different instruments are different, the confidence degrees of MRI image target objects are also different, and the method provided by the invention can automatically adjust the scale of the convolution kernel in the depth network according to the confidence degree of MRI image tumors.
The invention adopts the following technical scheme for solving the technical problems:
a self-adaptive variable-scale convolution kernel method based on brain MRI three-dimensional image confidence coefficient comprises two parts of calculation of target region confidence coefficient in an MRI image and calculation of a self-adaptive variable-scale convolution kernel;
the calculation of the confidence of the target region in the MRI image is specifically as follows:
step 1: calculating the SNR as follows:
step 1.1, on the image uniformity layer, measuring the measurement interest region MROI of the image central signal region from the x, y and z directions respectively to obtain the signal intensity Sx,Sy,Sz；
Step 1.2, respectively measuring frequency coding, phase coding and layer selection coding direction area of 100mm from four corner background areas around the die body2Standard deviation of ROI signal intensity
Figure BDA0002822394630000011
Further calculating the signal-to-noise ratio (SNR);
Figure BDA0002822394630000021
Figure BDA0002822394630000022
Figure BDA0002822394630000023
wherein, when the layer thickness d is 10mm, k is 1; when d is<When the thickness of the film is 10mm,
Figure BDA0002822394630000024
step 2, calculating the uniformity of a main magnetic field;
step 2.1, calculating the frequency, phase and slice selection linear gradient field G respectivelyx,GyAnd Gz(ii) a The method comprises the following specific steps:
Figure BDA0002822394630000025
Figure BDA0002822394630000026
Figure BDA0002822394630000027
wherein γ is gyromagnetic ratio, FOVx,FOVyAnd FOVzThe effective Field of view (BW) in the frequency encoding direction, the phase encoding direction and the slice selection encoding direction, respectivelyx,BWyAnd BWzRespectively selecting the receiving bandwidth in the encoding direction along the frequency encoding direction, the phase encoding direction and the layer;
step 2.2, calculating the local inhomogeneous magnetic field delta B in the frequency encoding direction, the phase encoding direction and the layer selection encoding direction0Amount of image deformation caused by (x, y, z):
x'＝x+ΔB0x(x,y,z)/Gx
y'＝y+ΔB0y(x,y,z)/Gy
z'＝z+ΔB0z(x,y,z)/Gz
step 2.3, for the uniformity of a certain spherical phantom magnetic field, respectively calculating the main magnetic field uniformity along the frequency coding, phase coding and layer selection coding directions by the following formulas:
Figure BDA0002822394630000028
Figure BDA0002822394630000031
Figure BDA0002822394630000032
wherein, BWx1，BWy1，BWz1Respectively, a smaller reception bandwidth; BW (Bandwidth)x2，BWy2，BWz2Are respectively larger receiving bandwidths, x'1-x'2，y'1-y'2，z'1-z'2Respectively obtaining the deformed displacement difference values of the images in the respective coding directions by 2 times of scanning in a smaller bandwidth and a higher bandwidth; b is0(T) is the main magnetic field strength;
step 2.4, according to the above steps, calculating confidence degrees on x, y and z axes respectively, namely calculating confidence degrees of a K space frequency encoding direction, a phase encoding direction and a layer selection encoding direction of the corresponding MRI; the confidence in the three directions is defined as:
Px＝βSNRx+(1-β)ΔB0x(ppm)
Py＝βSNRy+(1-β)ΔB0y(ppm)
Pz＝βSNRz+(1-β)ΔB0z(ppm)
wherein β is a weight parameter set artificially, and β is 0.5;
the calculation of the self-adaptive variable-scale convolution kernel specifically comprises the following steps:
step 3, solving the size of a specific multi-scale convolution kernel according to the calculated confidence coefficient of the target object;
step 3.1, finding out the relation between the confidence coefficient of the target object and the scale of the convolution kernel; wherein the size of the three-dimensional convolution kernel can be represented as w, h, d; wherein the initial values of w, h and d are all 3 and are marked as w0，h0，d0(ii) a Since the confidence value is a decimal between 0 and 1, it is not suitable for direct use as a convolution kernel scale(ii) a The size of the three-dimensional convolution kernel is determined by the following equation:
Figure BDA0002822394630000033
rounding up the fractional parts appearing in the three formulas;
Figure BDA0002822394630000034
based on the two groups of formulas, the three-dimensional convolution kernel has self-adaptive capacity to various different target objects, namely the three-dimensional convolution kernel can automatically adjust the scale of the three-dimensional convolution kernel according to the confidence degrees of different target object spaces and the probability of the occurrence of the target objects.
Compared with the prior art, the invention adopting the technical scheme has the following technical effects:
the invention relates to a self-adaptive variable-scale convolution kernel method based on confidence coefficient of brain MRI (magnetic resonance imaging) image, which can automatically change the scale of a convolution kernel in real time according to the confidence coefficient of a tumor region in the MRI image to better extract features, provide a large-scale convolution kernel to extract features if the confidence coefficient of the tumor region is high, and otherwise use a small convolution kernel to extract the features, so that computing resources are used at key positions, the problem of different gray scale ranges of MRI images obtained by different patients and different instruments is solved, and the size of the convolution kernel is changed in real time according to the confidence coefficient of a three-dimensional image pixel, so that the MRI tumor image can be better segmented.
Drawings
FIG. 1 is a schematic representation of an MRI three-dimensional image uniformity layer of an embodiment of the present invention;
FIG. 2 is a schematic diagram of an adaptive variable scale convolution kernel in accordance with an embodiment of the present invention.
Detailed Description
The technical scheme of the invention is further explained in detail by combining the attached drawings:
the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The invention provides a self-adaptive variable-scale convolution kernel method based on brain MRI three-dimensional image confidence based on the existing artificial intelligence technology.
Because different patients and MRI images obtained by different instruments have different qualities, the confidence degrees of target objects in the images are different, and three-dimensional convolution kernels with different scales are required for better extracting key features.
The technical solution for realizing the purpose of the invention is as follows: an adaptive variable-scale convolution kernel method based on brain MRI three-dimensional image confidence coefficient is particularly used for a method capable of changing the size of a convolution kernel in real time according to the value of the confidence coefficient of an interested region in an MRI three-dimensional image.
A first module: the confidence of the MRI image is calculated as follows:
step 1: the signal-to-noise ratio (SNR) is calculated.
Step 1.1: on the image uniformity layer as shown in fig. 1, the measurement region of interest (MROI) of the image center signal region is measured from x, y, z directions, respectively (at least 75% of a centered regular geometric region), resulting in signal intensity: sx,Sy,Sz；
Step 1.1.1: then, the frequency code and the phase code are respectively measured from the background areas of four corners of the periphery of the die body, and the area of the layer selection code direction is 100mm2Standard deviation of region of interest (ROI) signal intensity
Figure BDA0002822394630000051
The signal-to-noise ratio (SNR) is calculated from the following equation:
Figure BDA0002822394630000052
Figure BDA0002822394630000053
Figure BDA0002822394630000054
wherein when the layer thickness d is 10mm, k is 1; when d is<When the thickness of the film is 10mm,
Figure BDA0002822394630000055
step 1.2: the main magnetic field uniformity is calculated.
Step 1.2.1: calculating the frequency, phase and slice-select linear gradient fields G separatelyx,GyAnd Gz；
Figure BDA0002822394630000056
Figure BDA0002822394630000057
Figure BDA0002822394630000058
Where γ is the gyromagnetic ratio, FOVx,FOVyAnd FOVzThe effective Field of view (BW) in the frequency encoding direction, the phase encoding direction and the slice selection encoding direction, respectivelyx,BWyAnd BWzThe reception bandwidths in the frequency encoding direction, the phase encoding direction and the slice selection encoding direction, respectively.
Step 1.2.2: calculating the magnetic field due to local inhomogeneity Delta B in the frequency encoding direction, the phase encoding direction and the slice selection encoding direction0(x, y, z) induced image distortion amount
x'＝x+ΔB0x(x,y,z)/Gx
y'＝y+ΔB0y(x,y,z)/Gy
z'＝z+ΔB0z(x,y,z)/Gz
Step 1.2.3: for the uniformity of a certain spherical phantom magnetic field, the main magnetic field uniformity along the frequency coding, phase coding and layer selection coding directions is respectively calculated by the following formula:
Figure BDA0002822394630000061
Figure BDA0002822394630000062
Figure BDA0002822394630000063
wherein BWx1，BWy1，BWz1Respectively, a smaller reception bandwidth. BW (Bandwidth)x2，BWy2，BWz2Are respectively larger receiving bandwidths (x'1-x'2)，(y'1-y'2)，(z'1-z'2) The displacement difference of the deformation of the image in the respective coding direction is obtained by 2 times of scanning in a smaller bandwidth and a higher bandwidth respectively. B is0(T) is the main magnetic field strength.
Step 1.2.4: and respectively calculating the confidence degrees on the x, y and z axes according to the steps, namely calculating the confidence degrees of the K space frequency encoding direction, the phase encoding direction and the layer selection encoding direction of the corresponding MRI. This patent defines the confidence in three directions as:
Px＝βSNRx+(1-β)ΔB0x(ppm)
Py＝βSNRy+(1-β)ΔB0y(ppm)
Pz＝βSNRz+(1-β)ΔB0z(ppm)
where β is a weight parameter set manually, and β is set to 0.5 in this patent.
And a second module: the calculation of the self-adaptive variable-scale convolution kernel specifically comprises the following steps:
step 1: and calculating the size of a specific multi-scale convolution kernel according to the confidence coefficient of the target object calculated in the first module. First, the relationship between the confidence of the object and the scale of the convolution kernel is found. The size of the three-dimensional convolution kernel can be expressed as (w, h, d), and the initial values of w, h and d are all 3 and are denoted as w0，h0，d0. Since the confidence value is a fraction between 0 and 1, it is not suitable as a direct convolution kernel scale. The size of the three-dimensional convolution kernel is determined by the following equation:
Figure BDA0002822394630000064
for the fractional part appearing in the above three equations, rounding-up processing is performed.
Figure BDA0002822394630000065
Based on the two groups of formulas, the three-dimensional convolution kernel has self-adaptive capacity to various different target objects, namely the three-dimensional convolution kernel can automatically adjust the scale of the three-dimensional convolution kernel according to the confidence degrees of different target object spaces and the probability of the occurrence of the target objects.
It will be understood by those skilled in the art that, unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. It will be further understood that terms, such as those defined in commonly used dictionaries, should be interpreted as having a meaning that is consistent with their meaning in the context of the prior art and will not be interpreted in an idealized or overly formal sense unless expressly so defined herein.
The above embodiments are only for illustrating the technical idea of the present invention, and the protection scope of the present invention is not limited thereby, and any modifications made on the basis of the technical scheme according to the technical idea of the present invention fall within the protection scope of the present invention. While the embodiments of the present invention have been described in detail, the present invention is not limited to the above embodiments, and various changes can be made without departing from the spirit of the present invention within the knowledge of those skilled in the art.

Claims (1)

1. An adaptive variable-scale convolution kernel method based on brain MRI three-dimensional image confidence coefficient is characterized in that: the method comprises two parts of calculation of confidence coefficient of a target region in an MRI image and calculation of an adaptive variable-scale convolution kernel;
the calculation of the confidence of the target region in the MRI image is specifically as follows:
step 1: calculating the SNR as follows:
step 1.1, on the image uniformity layer, measuring the measurement interest region MROI of the image central signal region from the x, y and z directions respectively to obtain the signal intensity Sx,Sy,Sz；
Step 1.2, respectively measuring frequency coding, phase coding and layer selection coding direction area of 100mm from four corner background areas around the die body2Standard deviation of ROI signal intensity
Figure FDA0002822394620000011
Further calculating the signal-to-noise ratio (SNR);
Figure FDA0002822394620000012
Figure FDA0002822394620000013
Figure FDA0002822394620000014
wherein, when the layer thickness d is 10mm, k is 1; when d is<When the thickness of the film is 10mm,
Figure FDA0002822394620000015
step 2, calculating the uniformity of a main magnetic field;
step 2.1, calculating the frequency, phase and slice selection linear gradient field G respectivelyx,GyAnd Gz(ii) a The method comprises the following specific steps:
Figure FDA0002822394620000016
Figure FDA0002822394620000017
Figure FDA0002822394620000018
wherein γ is gyromagnetic ratio, FOVx,FOVyAnd FOVzEffective Field of view, BW, in the frequency encoding direction, the phase encoding direction and the slice selection encoding direction, respectivelyx,BWyAnd BWzRespectively selecting the receiving bandwidth in the encoding direction along the frequency encoding direction, the phase encoding direction and the layer;
step 2.2, calculating the local inhomogeneous magnetic field delta B in the frequency encoding direction, the phase encoding direction and the layer selection encoding direction0Amount of image deformation caused by (x, y, z):
x′＝x+ΔB0x(x,y,z)/Gx
y′＝y+ΔB0y(x,y,z)/Gy
z′＝z+ΔB0z(x,y,z)/Gz
step 2.3, for the uniformity of a certain spherical phantom magnetic field, respectively calculating the main magnetic field uniformity along the frequency coding, phase coding and layer selection coding directions by the following formulas:
Figure FDA0002822394620000021
Figure FDA0002822394620000022
Figure FDA0002822394620000023
wherein, BWx1，BWy1，BWz1Respectively, a smaller reception bandwidth; BW (Bandwidth)x2，BWy2，BWz2Are respectively larger receiving bandwidths, x'1-x′2，y′1-y′2，z′1-z′2Respectively obtaining the deformed displacement difference values of the images in the respective coding directions by 2 times of scanning in a smaller bandwidth and a higher bandwidth; b is0(T) is the main magnetic field strength;
step 2.4, according to the above steps, calculating confidence degrees on x, y and z axes respectively, namely calculating confidence degrees of a K space frequency encoding direction, a phase encoding direction and a layer selection encoding direction of the corresponding MRI; the confidence in the three directions is defined as:
Px＝βSNRx+(1-β)ΔB0x(ppm)
Py＝βSNRy+(1-β)ΔB0y(ppm)
Pz＝βSNRz+(1-β)ΔB0z(ppm)
wherein β is a weight parameter set artificially, and β is 0.5;
the calculation of the self-adaptive variable-scale convolution kernel specifically comprises the following steps:
step 3, solving the size of a specific multi-scale convolution kernel according to the calculated confidence coefficient of the target object;
step (ii) of3.1, finding out the relation between the confidence coefficient of the target object and the scale of the convolution kernel; wherein the size of the three-dimensional convolution kernel can be represented as w, h, d; wherein the initial values of w, h and d are all 3 and are marked as w0，h0，d0(ii) a The size of the three-dimensional convolution kernel is further determined by the following equation:
Figure FDA0002822394620000031
rounding up the fractional parts appearing in the three formulas;
Figure FDA0002822394620000032
based on the two groups of formulas, the three-dimensional convolution kernel has self-adaptive capacity to various different target objects, namely the three-dimensional convolution kernel can automatically adjust the scale of the three-dimensional convolution kernel according to the confidence degrees of different target object spaces and the probability of the occurrence of the target objects.
CN202011441419.2A 2020-05-07 2020-12-08 Self-adaptive variable-scale convolution kernel method based on brain MRI three-dimensional image confidence coefficient Active CN112330674B (en)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN2020103769177 2020-05-07
CN202010376917 2020-05-07

Publications (2)

Publication Number Publication Date
CN112330674A true CN112330674A (en) 2021-02-05
CN112330674B CN112330674B (en) 2023-06-30

Family

ID=74301865

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202011441419.2A Active CN112330674B (en) 2020-05-07 2020-12-08 Self-adaptive variable-scale convolution kernel method based on brain MRI three-dimensional image confidence coefficient

Country Status (1)

Country Link
CN (1) CN112330674B (en)

Citations (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107064838A (en) * 2017-04-25 2017-08-18 北京青檬艾柯科技有限公司 It is a kind of to form the magnet system configurations and measuring method for becoming gradient magnetostatic field
CN107240102A (en) * 2017-04-20 2017-10-10 合肥工业大学 Malignant tumour area of computer aided method of early diagnosis based on deep learning algorithm
CN107403201A (en) * 2017-08-11 2017-11-28 强深智能医疗科技（昆山）有限公司 Tumour radiotherapy target area and jeopardize that organ is intelligent, automation delineation method
US20180017655A1 (en) * 2016-07-18 2018-01-18 Siemens Healthcare Gmbh Method and apparatus for recording calibration data for a grappa magnetic resonance imaging algorithm
US20180218502A1 (en) * 2017-01-27 2018-08-02 Arterys Inc. Automated segmentation utilizing fully convolutional networks
CN109086770A (en) * 2018-07-25 2018-12-25 成都快眼科技有限公司 A kind of image, semantic dividing method and model based on accurate scale prediction
CN109872364A (en) * 2019-01-28 2019-06-11 腾讯科技（深圳）有限公司 Image-region localization method, device, storage medium and medical image processing equipment
CN109886929A (en) * 2019-01-24 2019-06-14 江苏大学 A kind of MRI tumour voxel detection method based on convolutional neural networks
CN109978838A (en) * 2019-03-08 2019-07-05 腾讯科技（深圳）有限公司 Image-region localization method, device and Medical Image Processing equipment
CN110415234A (en) * 2019-07-29 2019-11-05 北京航空航天大学 Brain tumor dividing method based on multi-parameter magnetic resonance imaging
CN111027547A (en) * 2019-12-06 2020-04-17 南京大学 Automatic detection method for multi-scale polymorphic target in two-dimensional image

Patent Citations (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20180017655A1 (en) * 2016-07-18 2018-01-18 Siemens Healthcare Gmbh Method and apparatus for recording calibration data for a grappa magnetic resonance imaging algorithm
US20180218502A1 (en) * 2017-01-27 2018-08-02 Arterys Inc. Automated segmentation utilizing fully convolutional networks
CN107240102A (en) * 2017-04-20 2017-10-10 合肥工业大学 Malignant tumour area of computer aided method of early diagnosis based on deep learning algorithm
CN107064838A (en) * 2017-04-25 2017-08-18 北京青檬艾柯科技有限公司 It is a kind of to form the magnet system configurations and measuring method for becoming gradient magnetostatic field
CN107403201A (en) * 2017-08-11 2017-11-28 强深智能医疗科技（昆山）有限公司 Tumour radiotherapy target area and jeopardize that organ is intelligent, automation delineation method
CN109086770A (en) * 2018-07-25 2018-12-25 成都快眼科技有限公司 A kind of image, semantic dividing method and model based on accurate scale prediction
CN109886929A (en) * 2019-01-24 2019-06-14 江苏大学 A kind of MRI tumour voxel detection method based on convolutional neural networks
CN109872364A (en) * 2019-01-28 2019-06-11 腾讯科技（深圳）有限公司 Image-region localization method, device, storage medium and medical image processing equipment
CN109978838A (en) * 2019-03-08 2019-07-05 腾讯科技（深圳）有限公司 Image-region localization method, device and Medical Image Processing equipment
CN110415234A (en) * 2019-07-29 2019-11-05 北京航空航天大学 Brain tumor dividing method based on multi-parameter magnetic resonance imaging
CN111027547A (en) * 2019-12-06 2020-04-17 南京大学 Automatic detection method for multi-scale polymorphic target in two-dimensional image

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
KONSTANTINOS KAMNITSAS等: "Efficient multi-scale 3D CNN with fully connected CRF for accurate brain lesion segmentation", 《MEDICAL IMAGE ANALYSIS》, vol. 36, pages 61 - 78, XP029882414, DOI: 10.1016/j.media.2016.10.004 *
李苹: "多模态视网膜图像血管分割及配准研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》, no. 1, pages 138 - 4604 *
杨珍等: "利用DCE-MRI结合改进卷积神经网络的MR图像自动分割与分类方法", 《重庆理工大学学报（自然科学）》, vol. 34, no. 2, pages 147 - 157 *

Also Published As

Publication number Publication date
CN112330674B (en) 2023-06-30

Similar Documents

Publication Publication Date Title
US20210225027A1 (en) Image region localization method, image region localization apparatus, and medical image processing device
US11645748B2 (en) Three-dimensional automatic location system for epileptogenic focus based on deep learning
Goshtasby et al. Similarity and dissimilarity measures
Thanki et al. Medical imaging and its objective quality assessment: an introduction
Chen et al. MRI brain tissue classification using unsupervised optimized extenics-based methods
CN105488781A (en) Dividing method based on CT image liver tumor focus
Sharan et al. Encoder modified U-net and feature pyramid network for multi-class segmentation of cardiac magnetic resonance images
CN103632345A (en) MRI image inhomogeneity correction method based on regularization
CN113628234B (en) Method for detecting salient polarization SAR image change based on comprehensive neighborhood information
Shobana et al. Brain tumor diagnosis from MRI feature analysis-A comparative study
Nain et al. Statistical shape analysis of brain structures using spherical wavelets
CN110458842B (en) Brain tumor segmentation method based on two-channel three-dimensional dense connection network
Li et al. 3-D inspection method for industrial product assembly based on single X-ray projections
Li et al. High‐fidelity fast volumetric brain MRI using synergistic wave‐controlled aliasing in parallel imaging and a hybrid denoising generative adversarial network (HDnGAN)
Abbas et al. Breast cancer image segmentation using morphological operations
CN116758087B (en) Lumbar vertebra CT bone window side recess gap detection method and device
CN111429404B (en) Imaging system and method for detecting cardiovascular and cerebrovascular vessels
Wang et al. A fuzzy C-means model based on the spatial structural information for brain MRI segmentation
CN112330674B (en) Self-adaptive variable-scale convolution kernel method based on brain MRI three-dimensional image confidence coefficient
Kr Ghosh et al. Development of intuitionistic fuzzy special embedded convolutional neural network for mammography enhancement
CN105631897B (en) Based on the film nuclear magnetic resonance image sequence motion method of estimation for singly drilling signal characteristic distance and cross-correlation transformation optical flow algorithm
CN114723937A (en) Method and system for classifying blood vessel surrounding gaps based on nuclear magnetic resonance image
Gao et al. Using the compressed sensing technique for lumbar vertebrae imaging: comparison with conventional parallel imaging
Morais et al. Fully automatic left ventricular myocardial strain estimation in 2D short-axis tagged magnetic resonance imaging
Shamsi Koshki et al. A level‐set method for inhomogeneous image segmentation with application to breast thermography 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