WO2022016396A1 - 医学图像的处理方法、处理装置及计算机可读存储介质 - Google Patents
医学图像的处理方法、处理装置及计算机可读存储介质 Download PDFInfo
- Publication number
- WO2022016396A1 WO2022016396A1 PCT/CN2020/103371 CN2020103371W WO2022016396A1 WO 2022016396 A1 WO2022016396 A1 WO 2022016396A1 CN 2020103371 W CN2020103371 W CN 2020103371W WO 2022016396 A1 WO2022016396 A1 WO 2022016396A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- pixel
- image
- processed
- information
- medical image
- Prior art date
Links
- 238000012545 processing Methods 0.000 title claims abstract description 16
- 238000003672 processing method Methods 0.000 title claims abstract description 15
- 238000000034 method Methods 0.000 claims abstract description 31
- 238000000605 extraction Methods 0.000 claims abstract description 17
- 238000013527 convolutional neural network Methods 0.000 claims description 37
- 230000003044 adaptive effect Effects 0.000 claims description 15
- 230000004913 activation Effects 0.000 claims description 12
- 238000004590 computer program Methods 0.000 claims description 12
- 230000006870 function Effects 0.000 claims description 12
- 230000008569 process Effects 0.000 claims description 10
- 238000010606 normalization Methods 0.000 claims description 9
- 238000012549 training Methods 0.000 claims description 8
- 238000001035 drying Methods 0.000 claims description 5
- 238000010586 diagram Methods 0.000 description 8
- 238000005516 engineering process Methods 0.000 description 7
- 238000002595 magnetic resonance imaging Methods 0.000 description 5
- 230000002708 enhancing effect Effects 0.000 description 3
- 238000003384 imaging method Methods 0.000 description 3
- 210000004556 brain Anatomy 0.000 description 2
- 238000002591 computed tomography Methods 0.000 description 2
- 238000002059 diagnostic imaging Methods 0.000 description 2
- 238000001914 filtration Methods 0.000 description 2
- 230000004927 fusion Effects 0.000 description 2
- 238000002604 ultrasonography Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 125000004435 hydrogen atom Chemical group [H]* 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000012805 post-processing Methods 0.000 description 1
- 230000005855 radiation Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000012827 research and development Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
- 210000004872 soft tissue Anatomy 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
- G06V10/75—Organisation of the matching processes, e.g. simultaneous or sequential comparisons of image or video features; Coarse-fine approaches, e.g. multi-scale approaches; using context analysis; Selection of dictionaries
- G06V10/751—Comparing pixel values or logical combinations thereof, or feature values having positional relevance, e.g. template matching
-
- 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
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/70—Denoising; Smoothing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/774—Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting
- G06V10/7753—Incorporation of unlabelled data, e.g. multiple instance learning [MIL]
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H30/00—ICT specially adapted for the handling or processing of medical images
- G16H30/40—ICT specially adapted for the handling or processing of medical images for processing medical images, e.g. editing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/03—Recognition of patterns in medical or anatomical images
Definitions
- the present application relates to the technical field of image processing, and in particular, to a medical image processing method, a processing device, and a computer-readable storage medium.
- Magnetic resonance imaging is an examination method that converts the signals generated by radio frequency electromagnetic waves and hydrogen protons in the human body into an image after processing.
- Magnetic resonance imaging technology has the advantages of safety without radiation, high soft tissue contrast, and many functional imaging methods. It is one of the most powerful and complex technologies in medical imaging technology at present.
- magnetic resonance imaging technology has the disadvantage of slower data acquisition speed (longer scan time). For example, in dynamic imaging, it is easy to cause low image resolution and motion artifacts, which restrict the wide application of magnetic resonance imaging technology in clinical practice. If the MRI scan time is shortened, the MR reconstruction image may have artifacts due to the insufficient amount of acquired data. Therefore, research and development of post-processing technology for MR image denoising has important scientific significance and broad application prospects for the current medical diagnosis field.
- Embodiments of the present application provide a medical image processing method, a processing device, and a computer-readable storage medium, so as to solve the problem of poor medical image quality.
- the present application provides a medical image processing method, the method includes: acquiring an image to be processed. Feature extraction is performed on the image to be processed to obtain a corresponding feature image. Use the first information and second information of each pixel in the feature image to re-determine the pixel value of each pixel in the image to be processed, and then process the image to be processed, where the first information is the pixel adjacent to the target pixel information, and the second information is pixel information that is not adjacent to the target pixel and is similar to the target pixel.
- the present application also provides a medical image processing device, including a memory, a processor, and a computer program stored in the memory and running on the processor. The steps of the processing method.
- the present application also provides a computer-readable storage medium, where a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, the steps of the above medical image processing method are implemented.
- the beneficial effects of the present application are: the present application re-acquires the pixel value corresponding to each pixel by combining the adjacent pixel information of each pixel of the medical image and the pixel information that is not adjacent to the pixel but similar to the pixel, thereby making the processing
- the resulting medical images can better retain image details and improve image quality.
- FIG. 1 is a schematic flowchart of an embodiment of a medical image processing method provided by the present application.
- FIG. 2 is a schematic structural diagram of an embodiment of a convolutional neural network provided by the present application.
- FIG. 3 is a schematic structural diagram of another embodiment of a convolutional neural network provided by the present application.
- FIG. 4 is a schematic structural diagram of an embodiment of a medical image processing apparatus provided by the present application.
- FIG. 5 is a schematic structural diagram of an embodiment of a computer-readable storage medium provided by the present application.
- first, second, third, etc. may be used in this application to describe various information, such information should not be limited by these terms. These terms are only used to distinguish the same type of information from each other.
- first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information without departing from the scope of the present application.
- word "if” as used herein can be interpreted as "at the time of” or "when” or "in response to determining.”
- FIG. 1 is a schematic flowchart of an embodiment of a medical image processing method provided by the present application.
- the medical image processing method includes the following steps:
- the images to be processed in this embodiment may be some common medical images, such as magnetic resonance images, CT (Computed Tomography) images, X-ray images, B-scan ultrasonography (B-scan ultrasonography) images, etc. etc., for example, a magnetic resonance image of the human brain.
- CT Computer Tomography
- X-ray images X-ray images
- B-scan ultrasonography B-scan ultrasonography
- the above enumerated magnetic resonance image of the human brain is only an exemplary description, and in actual implementation, it may also be a medical image of other parts of the target, which is not limited in this application.
- S2 Perform feature extraction on the image to be processed to obtain a corresponding feature image.
- the feature extraction of the image to be processed may be performed in a convolution manner.
- a plurality of convolution kernels of different sizes are used to convolve the image to be processed, so as to output a multi-channel feature image of the image to be processed.
- the output channel number is 3, and each pixel of the feature image includes a three-dimensional Feature vector.
- S3 Re-determine the pixel value of each pixel in the image to be processed by using the first information and the second information of each pixel in the feature image, and then process the image to be processed, wherein the first information is adjacent to the target pixel
- the second information is pixel information that is not adjacent to the target pixel and is similar to the target pixel.
- the inventors of the present application have found through long-term research that most of the prior art de-drying methods for medical images only use pixel information in the image adjacent to the target pixel to de-dry the image, such as bilinear filtering methods and median filtering methods.
- pixel information that is not adjacent but similar to the target pixel is not considered.
- the pixels corresponding to these pixel information are far away from the target pixel in spatial distance, due to a large amount of information redundancy in the image, some pixels exist and the target pixel. Pixels are similar.
- the inventor of the present application combines the first information of the pixels in the image, that is, the pixel information adjacent to the target pixel, and the second information, that is, the pixel information that is not adjacent but similar to the target pixel. It makes the de-drying image clearer and improves the image quality.
- step S3 is specifically implemented through the following steps:
- S31 Acquire the pixels in the neighborhood with the radius r around the target pixel in the feature image and form the first pixel set, and obtain the pixels similar to the target pixel and located outside the neighborhood to form the second pixel set, where r is a preset neighborhood
- the feature image corresponding to the to-be-processed image whose size is L ⁇ H ⁇ W is set as The feature vector of any pixel i in feature space H l is Then the Euclidean distance between pixel i and pixel j in the feature space is:
- effective information can be obtained from the first information and the second information, that is, the first pixel and the second pixel obtained in step S32 are given weighted values, for example, the second pixel is more similar to the target pixel and the corresponding second contribution rate bigger.
- the pixel value of the target pixel can be re-determined by using the above two types of pixels.
- the pixel values of all the first pixels and the second pixels can be averaged as the pixel value of the target pixel, that is, all the first contribution rates and the second contribution rates are equal.
- the first contribution rate and the second contribution rate may also be acquired by using a preset convolutional neural network M.
- the medical image processing method provided in this embodiment firstly performs feature extraction on the image to be processed to obtain a corresponding feature image, and then uses the adjacent pixel information of each pixel in the feature image and the non-adjacent but similar pixel information to re-determine the to-be-processed image
- the pixel value of each pixel corresponding to in the prior art solves the problem that the quality of the obtained image is poor due to the loss of detail information caused by only using the pixel information adjacent to each pixel to de-dry the image in the prior art.
- FIG. 2 is a schematic structural diagram of an embodiment of a convolutional neural network provided by the present application.
- a preset convolutional neural network M is established in this embodiment, and the preset convolutional neural network M includes at least one feature extraction layer 1, at least one desiccation layer 4, at least one batch of normalization layers.
- a normalization layer 5 and at least one activation layer 2 are connected to each feature extraction layer 1 and each de-drying layer 4 .
- the above-mentioned preset convolutional neural network M may further include at least one adaptive residual unit 70 connected in series, wherein each adaptive residual unit 70 includes three denoising layers 4, three activation layers 2, 3 batch normalization layers 5 and 1 adaptive skip connection unit 6. Therefore, the preset convolutional neural network M includes K+1 adaptive skip connection units 6 , where K represents the number of adaptive residual units 70 included in the preset convolutional neural network M.
- the input and output are directly connected through the adaptive skip connection unit 6 to retain more detailed information of the input image, enhance the feature extraction of the preset convolutional neural network M, and increase the training process of the preset convolutional neural network M The convergence speed of , thereby greatly enhancing the learning ability of the preset convolutional neural network M.
- the adaptive skip connection unit 6 of this embodiment can directly use unit mapping to connect the input and output.
- Fig. 3 is a schematic structural diagram of another embodiment of the convolutional neural network provided by the present application.
- the preset convolutional neural network M may include 6 feature extraction layers 1, 13 desiccation layers 4, 18 activation layers 2, 12 batch normalization layers 5, a fusion layer 3 and 4 adaptive skip connections Unit 6.
- the preset convolutional neural network M further includes 3 adaptive residual units 70 connected in series, and each adaptive residual unit 70 includes 3 denoising layers 4 , 3 activation layers 2 , and 3 batch normalization layers Layer 5 and an adaptive skip connection unit 6.
- the feature extraction layer 1 is used to receive the image to be processed and perform feature extraction on it to obtain the corresponding feature image.
- the fusion layer 3 is used to fuse the images output by the multiple activation layers 2 and input them into the desiccation layer 4.
- the desiccation layer 4 is used to re-determine the pixel value of each pixel in the image to be processed by using the first information and the second information of each pixel in the feature image, and then process the image to be processed, wherein the first information is Pixel information adjacent to the target pixel, and the second information is pixel information not adjacent to the target pixel but similar to the target pixel.
- the introduction of the adaptive skip connection unit 6 directly connects the input and output, thereby retaining more detailed information of the input image, enhancing the feature extraction of the preset convolutional neural network M, and enhancing the training process of the preset convolutional neural network M. convergence speed.
- the desiccation layer 4 is used to re-determine the pixel value of each pixel in the image to be processed by using the first information and the second information of each pixel in the feature image, and then process the image to be processed, wherein the first The first information is the pixel information adjacent to the target pixel, and the second information is the pixel information that is not adjacent to the target pixel but similar to the target pixel.
- the following formula can be used to express:
- ⁇ represents a nonlinear activation function, such as the Sigmoid function, the relu function or the Hanh function.
- F l [d l (i, j)] represents the second contribution rate, and the second contribution rate is determined by the Euclidean distance d l (i, j) between the second pixel and the target pixel.
- W l represents the first contribution rate
- b l represents the bias
- represents the second pixel Indicates the image after drying.
- unprocessed medical images and corresponding processed medical images can be acquired as training data sets.
- the preset convolutional neural network M is trained with the unprocessed medical images as input data and the corresponding processed medical images as ground truth labels.
- the unprocessed medical image is a medical image obtained under a scanning condition less than a set time period
- the corresponding processed medical image represents a medical image obtained under a scanning condition greater than the set time period.
- the set time period may refer to a time period determined by multiple medical imaging experiments. When the scanning time is longer than this time period, the obtained medical image quality is good, otherwise, the obtained corresponding medical image quality is poor.
- the preset convolutional neural network M is trained using the mean square error loss function.
- the formula for calculating MSE is as follows:
- J( ⁇ ) is the mean square error loss function
- m is the total number of input unprocessed medical image pixels
- h ⁇ ( xi ) is the pixel value of the image pixel i output by the preset convolutional neural network M
- x i is the pixel value of the ith pixel in the input unprocessed image
- y i represents the pixel value of the ith pixel in the processed medical image corresponding to the unprocessed medical image.
- the model parameters ⁇ of the preset convolutional neural network M are updated by using the gradient back-propagation algorithm, so that the loss function tends to be minimized.
- the gradient descent algorithm needs to be used when the gradient back-propagation algorithm is performed to train network parameters ⁇ such as the first contribution rate and the second contribution rate.
- Commonly used gradient descent algorithms include batch gradient descent, stochastic gradient descent, mini-batch gradient descent, and so on.
- the gradient descent algorithm used is stochastic gradient descent.
- other gradient descent algorithms can also be used, depending on the specific situation, which is not specifically limited here.
- the partial derivative of the mean square loss function MSE is obtained for the parameter ⁇ that needs to be trained, and the gradient of the loss function is:
- the parameter ⁇ in the convolutional layer can be updated by the following formula:
- the parameters ⁇ that need to be determined include the first contribution rate W l of the desiccation layer, the second contribution rate F l [d l (i,j)], the bias parameter b l and the weight parameter W and bias term b of other convolutional layers in the preset convolutional neural network M.
- the image to be processed is input into the preset convolutional neural network M, and the preset convolutional neural network M is used to perform feature extraction of the image to be processed, so as to obtain the corresponding feature image and use the first feature of each pixel in the feature image.
- the first information and the second information re-determine the pixel value of each pixel corresponding to the to-be-processed image, so as to further process the to-be-processed image.
- optimal model parameters can be obtained, thereby improving the effective processing of information by the network, thereby improving the quality of the output image.
- FIG. 4 is a schematic structural diagram of an embodiment of a medical image processing apparatus provided by the present application.
- the processing apparatus 100 includes a memory 110 and a processor 120, wherein the memory 110 is used for storing computer programs, and the processor 120 is used for A computer program is executed to implement the steps of the medical image processing method provided by the present application.
- the processor 120 may be a central processing unit (CPU), or an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present application.
- CPU central processing unit
- ASIC application specific integrated circuit
- Memory 110 is used for executable instructions.
- the memory 110 may include high-speed RAM memory 110, or may include non-volatile memory 110, such as at least one disk memory.
- Memory 110 may also be a memory array.
- the storage 110 may also be divided into blocks, and the blocks may be combined into virtual volumes according to certain rules.
- the instructions stored in the memory 110 can be executed by the processor 120 to enable the processor 120 to perform the method of medical image processing in any of the above-described method embodiments.
- FIG. 5 is a schematic structural diagram of an embodiment of a computer-readable storage medium provided by the present application.
- a computer program 201 is stored on the computer-readable storage medium 200, and when the computer program 201 is executed by the processor, the steps of the medical image processing method provided in this application are implemented.
- the computer storage medium 200 can be any available medium or data storage device that can be accessed by a computer, including but not limited to magnetic storage (eg, floppy disk, hard disk, magnetic tape, magneto-optical disk (MO), etc.), optical storage (eg, CD, DVD, etc.) , BD, HVD, etc.), as well as semiconductor memory (eg, ROM, EPROM, EEPROM, non-volatile memory 110 (NANDFLASH), solid state disk (SSD)), and the like.
- magnetic storage eg, floppy disk, hard disk, magnetic tape, magneto-optical disk (MO), etc.
- optical storage eg, CD, DVD, etc.
- BD magneto-optical disk
- HVD etc.
- semiconductor memory eg, ROM, EPROM, EEPROM, non-volatile memory 110 (NANDFLASH), solid state disk (SSD)
- the solution of the present application can use the adjacent pixel information of each pixel in the feature image and the pixel information that is not adjacent to the pixel but similar to the pixel to re-determine the pixel value of each pixel corresponding to the image to be processed, so as to solve the problem that the prior art only uses and The pixel information adjacent to each pixel de-drys the image, and the detail information is lost, which leads to the problem of poor image quality.
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Software Systems (AREA)
- Computing Systems (AREA)
- Evolutionary Computation (AREA)
- Medical Informatics (AREA)
- Databases & Information Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Multimedia (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Computational Linguistics (AREA)
- Mathematical Physics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Radiology & Medical Imaging (AREA)
- Public Health (AREA)
- Primary Health Care (AREA)
- Epidemiology (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Image Analysis (AREA)
- Measuring And Recording Apparatus For Diagnosis (AREA)
Abstract
Description
Claims (10)
- 一种医学图像的处理方法,其特征在于,所述方法包括:获取待处理图像;对所述待处理图像进行特征提取,以得到对应的特征图像;利用所述特征图像中每个像素的第一信息和第二信息重新确定所述待处理图像中相对应的每个所述像素的像素值,进而对所述待处理图像进行处理,其中,所述第一信息为与目标像素邻近的像素信息,所述第二信息为与所述目标像素非邻近且与所述目标像素相似的像素信息。
- 根据权利要求1所述的方法,其特征在于,所述利用所述特征图像中每个像素的第一信息和第二信息重新确定所述待处理图像中相对应的每个所述像素的像素值,进而对所述待处理图像进行处理,包括:获取所述特征图像中的所述目标像素周围半径为r的邻域内的像素并组成第一像素集合,获取与所述目标像素相似且位于所述邻域外的像素并组成第二像素集合,其中r为预设邻域半径参数,r为正整数;获取所述第一像素集合中每一个第一像素对所述目标像素的第一贡献率,获取所述第二像素集合中每一个第二像素对所述目标像素的第二贡献率,其中,所有所述第一贡献率与所有所述第二贡献率之和为1;将所有所述第一像素的像素值和对应的所述第一贡献率相乘作为第一乘积,将所有所述第二像素的像素值和对应的所述第二贡献率相乘作为第二乘积,将所有的所述第一乘积与所有的所述第二乘积相加作为所述目标像素的像素值。
- 根据权利要求2所述的方法,其特征在于,所述获取与所述目标像素相似且位于所述邻域外的像素并组成第二像素集合包括:分别计算所述目标像素的特征向量与所述邻域外的每一个像素的特征向量的欧式距离,并获取其中N个欧式距离最小的像素并组成所述第二像素集合,在所述第二像素集合中,所述欧式距离越小,其对应的所述第二像素的所述第二贡献率越大;其中,N为正整数。
- 根据权利要求1所述的方法,其特征在于,所述方法还包括:将所述待处理图像输入至预设卷积神经网络,并利用所述预设卷积神经网络执行所述对所述待处理图像进行特征提取,以得到对应的特征图像,以及所述利用所述特征图像中每个像素的第一信息和第二信息重新确定所述待处理图像中相对应的每个所述像素的像素值,进而对所述待处理图像进行处理的步骤。
- 根据权利要求4所述的方法,其特征在于,所述方法还包括:建立所述预设卷积神经网络;其中,所述预设卷积神经网络包括至少一特征提取层、至少一去燥层、至少一批归一化层以及至少一激活层,其中,每个所述特征提取层后连接一所述激活层,每个所述去燥层后依次连接一所述批归一化层以及一所述激活层;获取未处理的所述医学图像和对应的处理后的所述医学图像作为训练数据集;以所述未处理的所述医学图像作为输入数据,且以所述对应的处理后的所述医学图像作为真值标签,对所述预设卷积神经网络进行训练。
- 根据权利要求5所述的方法,其特征在于,所述卷积神经网络包括相互串联的至少一自适应残差单元,其中每个所述自适应残差单元包括3个所述去噪层、3个所述激活层、3个所述批归一化层以及1个自适应跳跃连接单元。
- 根据权利要求5所述的方法,其特征在于,所述未处理的所述医学图像为在小于设定时间段的扫描条件下获取的医学图像,所述对应的处理后的所述医学图像表示在大于所述设定时间段的扫描条件下获取的医学图像。
- 根据权利要求5所述的方法,其特征在于,所述以所述未处理的所述医学图像作为输入数据,且以所述对应的处理后的所述医学图像作为真值标签,对所述预设卷积神经网络进行训练,包括:利用均方误差损失函数训练所述卷积神经网络;利用梯度反向传播算法更新所述卷积神经网络的模型参数,以使得所述损失函数趋于最小化。
- 一种医学图像的处理装置,其特征在于,包括存储器、处理器以及存储在所述存储器中并可在所述处理器上运行的计算机程序,所述处理器执行所述计算机程序时实现如权利要求1至8任一项所述方法的步骤。
- 一种计算机可读存储介质,其特征在于,所述计算机可读存储介质存储有计算机程序,所述计算机程序被处理器执行时实现如权利要求1至8任一项所述方法的步骤。
Priority Applications (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
PCT/CN2020/103371 WO2022016396A1 (zh) | 2020-07-21 | 2020-07-21 | 医学图像的处理方法、处理装置及计算机可读存储介质 |
US17/739,121 US20220262498A1 (en) | 2020-07-21 | 2022-05-08 | Medical image processing method, processing apparatus, and computer-readable storage medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
PCT/CN2020/103371 WO2022016396A1 (zh) | 2020-07-21 | 2020-07-21 | 医学图像的处理方法、处理装置及计算机可读存储介质 |
Related Child Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
US17/739,121 Continuation US20220262498A1 (en) | 2020-07-21 | 2022-05-08 | Medical image processing method, processing apparatus, and computer-readable storage medium |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2022016396A1 true WO2022016396A1 (zh) | 2022-01-27 |
Family
ID=79728382
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/CN2020/103371 WO2022016396A1 (zh) | 2020-07-21 | 2020-07-21 | 医学图像的处理方法、处理装置及计算机可读存储介质 |
Country Status (2)
Country | Link |
---|---|
US (1) | US20220262498A1 (zh) |
WO (1) | WO2022016396A1 (zh) |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103927737A (zh) * | 2013-10-31 | 2014-07-16 | 王浩然 | 基于非局部均值的sar图像变化检测方法 |
US20170178341A1 (en) * | 2015-12-21 | 2017-06-22 | Uti Limited Partnership | Single Parameter Segmentation of Images |
CN111340739A (zh) * | 2020-03-24 | 2020-06-26 | 珠海嘉润医用影像科技有限公司 | 一种图像处理方法及*** |
CN111429370A (zh) * | 2020-03-23 | 2020-07-17 | 煤炭科学技术研究院有限公司 | 一种煤矿井下的图像增强方法、***及计算机存储介质 |
Family Cites Families (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20080310695A1 (en) * | 2003-09-04 | 2008-12-18 | Garnier Stephen J | Locally Adaptive Nonlinear Noise Reduction |
EP2380132B1 (en) * | 2008-12-25 | 2018-02-07 | Medic Vision - Imaging Solutions Ltd. | Denoising medical images |
CN111582027B (zh) * | 2020-04-01 | 2024-03-22 | 广州亚美智造科技有限公司 | 身份认证方法、装置、计算机设备和存储介质 |
-
2020
- 2020-07-21 WO PCT/CN2020/103371 patent/WO2022016396A1/zh active Application Filing
-
2022
- 2022-05-08 US US17/739,121 patent/US20220262498A1/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103927737A (zh) * | 2013-10-31 | 2014-07-16 | 王浩然 | 基于非局部均值的sar图像变化检测方法 |
US20170178341A1 (en) * | 2015-12-21 | 2017-06-22 | Uti Limited Partnership | Single Parameter Segmentation of Images |
CN111429370A (zh) * | 2020-03-23 | 2020-07-17 | 煤炭科学技术研究院有限公司 | 一种煤矿井下的图像增强方法、***及计算机存储介质 |
CN111340739A (zh) * | 2020-03-24 | 2020-06-26 | 珠海嘉润医用影像科技有限公司 | 一种图像处理方法及*** |
Also Published As
Publication number | Publication date |
---|---|
US20220262498A1 (en) | 2022-08-18 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
WO2021077997A1 (zh) | 图像去噪的多生成器生成对抗网络学习方法 | |
Liu et al. | Deep iterative reconstruction estimation (DIRE): approximate iterative reconstruction estimation for low dose CT imaging | |
Yuan et al. | SIPID: A deep learning framework for sinogram interpolation and image denoising in low-dose CT reconstruction | |
CN110223255B (zh) | 一种基于残差编解码网络的低剂量ct图像去噪递归方法 | |
CN112837244B (zh) | 一种基于渐进式生成对抗网络的低剂量ct图像降噪及去伪影方法 | |
WO2022121100A1 (zh) | 一种基于darts网络的多模态医学图像融合方法 | |
CN108038840B (zh) | 一种图像处理方法、装置、图像处理设备及存储介质 | |
CN114863225A (zh) | 图像处理模型训练方法、生成方法、装置、设备及介质 | |
Li et al. | Learning non-local perfusion textures for high-quality computed tomography perfusion imaging | |
Liu et al. | Low-dose CT noise reduction based on local total variation and improved wavelet residual CNN | |
Luo et al. | Ultra-dense denoising network: application to cardiac catheter-based X-ray procedures | |
Bhonsle et al. | Reduction of Gaussian noise from Computed Tomography Images using Optimized Bilateral Filter by Enhanced Grasshopper Algorithm | |
WO2020007026A1 (zh) | 分割模型训练方法、装置及计算机可读存储介质 | |
CN111860836A (zh) | 一种自监督学习方法及应用 | |
Mangalagiri et al. | Toward generating synthetic CT volumes using a 3D-conditional generative adversarial network | |
Gholizadeh-Ansari et al. | Low-dose CT denoising using edge detection layer and perceptual loss | |
WO2022016396A1 (zh) | 医学图像的处理方法、处理装置及计算机可读存储介质 | |
US20220292641A1 (en) | Dynamic imaging and motion artifact reduction through deep learning | |
US20240233091A9 (en) | Generalizable Image-Based Training Framework for Artificial Intelligence-Based Noise and Artifact Reduction in Medical Images | |
Kang et al. | Gradient extraction based multiscale dense cross network for LDCT denoising | |
Tänzer et al. | Faster diffusion cardiac MRI with deep learning-based breath hold reduction | |
CN112053292B (zh) | 医学图像的处理方法、处理装置及计算机可读存储介质 | |
WO2021189383A1 (zh) | 生成高能ct图像模型的训练及生成方法、设备、存储介质 | |
Li et al. | Dual-domain fusion deep convolutional neural network for low-dose CT denoising | |
Liu et al. | Improving diagnostic accuracy of reduced-dose studies with full-dose noise-to-noise learning in cardiac spect |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 20946343 Country of ref document: EP Kind code of ref document: A1 |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 20946343 Country of ref document: EP Kind code of ref document: A1 |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 20946343 Country of ref document: EP Kind code of ref document: A1 |
|
32PN | Ep: public notification in the ep bulletin as address of the adressee cannot be established |
Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 07/07/2023) |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 20946343 Country of ref document: EP Kind code of ref document: A1 |