CN115861394B - Medical image processing method and device, storage medium and electronic equipment - Google Patents
Medical image processing method and device, storage medium and electronic equipment Download PDFInfo
- Publication number
- CN115861394B CN115861394B CN202310175697.5A CN202310175697A CN115861394B CN 115861394 B CN115861394 B CN 115861394B CN 202310175697 A CN202310175697 A CN 202310175697A CN 115861394 B CN115861394 B CN 115861394B
- Authority
- CN
- China
- Prior art keywords
- sample image
- medical
- target
- medical sample
- image
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Landscapes
- Measuring And Recording Apparatus For Diagnosis (AREA)
Abstract
The application discloses a medical image processing method and device, a storage medium and electronic equipment, and relates to the technical field of image processing, wherein the method comprises the following steps: acquiring a target medical image of a first modality; registering the target medical image of the first modality through a target registration network to obtain a registered target medical image of a second modality, wherein the target registration network is trained through a target evaluator, the target evaluator is trained by a first medical sample image pair, the first medical sample image pair is composed of a first medical sample image and a second medical sample image, and the first medical sample image and the second medical sample image are images of the same modality in absolute terms. According to the method and the device, the problem that in the related art, the accuracy of registration of the registration network to the image is low due to the fact that the registration network is trained through existing multi-mode data is solved.
Description
Technical Field
The present invention relates to the field of image processing technologies, and in particular, to a medical image processing method and apparatus, a storage medium, and an electronic device.
Background
The purpose of multi-modal image registration is to align two images (moving (M) and target (T) images) with different distributions by distorting the space of the deformation field. One major challenge in multi-modality image registration is that the acquired images using different imaging machines or different acquisition parameters may differ in multiple aspects. Because the reconstruction and acquisition methods are significantly different, there is no simple one-to-one mapping between the different imaging modes. From the perspective of measuring similarity, the mainstream unsupervised multi-modality registration methods can be divided into two categories: registration based on similarity operators and registration based on image-to-image transformations.
The similarity operator based registration is a loss function using a multi-modal similarity operator as registration, e.g., cross-correlation (NCC), mutual Information (MI), and modality independent neighborhood descriptor (bond). The similarity operator is based on a priori mathematical knowledge of the researcher and is carefully designed and continuously improved over a long period of time. They can be applied to conventional registration processes and neural network registration. The similarity operator has several limitations: it is not possible to design a similarity operator that can maintain high accuracy for all data from the various imaging modalities. Secondly, the upper limit that these operators can reach cannot be estimated, so it is difficult to find the improvement direction.
Registration based on image-to-image conversion first converts a multi-modality image to a single-modality image using a generation countermeasure network (GAN), and then evaluates the error of each pixel in space using Mean Absolute Error (MAE) or Mean Square Error (MSE). The registration based on image-to-image conversion skillfully avoids the complex multi-modal problem and reduces the difficulty of registration to a certain extent. However, it has significant drawbacks: GAN-based methods require training using existing multimodal data. If the trained model encounters invisible data, it will not work, which greatly limits its applicable scenarios.
Aiming at the problem that the accuracy of registration of the registration network to the image is low due to the fact that the registration network is trained through the existing multi-mode data in the related art, no effective solution is proposed at present.
Disclosure of Invention
The main purpose of the application is to provide a medical image processing method and device, a storage medium and an electronic device, so as to solve the problem that in the related art, the accuracy of registration of the registration network to the image is low due to the fact that the registration network is trained through existing multi-mode data.
In order to achieve the above object, according to one aspect of the present application, there is provided a medical image processing method. The method comprises the following steps: acquiring a target medical image of a first modality; registering the target medical image of the first modality through a target registration network to obtain a registered target medical image of a second modality, wherein the target registration network is trained through a target evaluator, the target evaluator is trained by a first medical sample image pair, the first medical sample image pair is composed of a first medical sample image and a second medical sample image, and the first medical sample image and the second medical sample image are images of the same modality in absolute terms.
Further, before registering the target medical image of the first modality through a target registration network, the method further comprises: acquiring the first medical sample image and the second medical sample image, and calculating to obtain a real error map between the first medical sample image and the second medical sample image; performing image enhancement processing on the first medical sample image to obtain a processed first medical sample image; determining the processed first medical sample image, the second medical sample image and the true error map as a first medical sample image pair; training an initial evaluator according to the first medical sample image to obtain the target evaluator.
Further, acquiring the first medical sample image and the second medical sample image comprises: acquiring an original medical sample image; performing first random spatial transformation on the original medical sample image to obtain the first medical sample image; and performing second random spatial transformation on the original medical sample image to obtain the second medical sample image.
Further, performing image enhancement processing on the first medical sample image to obtain a processed first medical sample image, where the processing includes: normalizing the pixel points on the first medical sample image to a preset numerical range to obtain a normalized first medical sample image; randomly generating a plurality of control points in the preset numerical range, and dividing the normalized first medical sample image according to the plurality of control points to obtain a plurality of segmented first medical sample images, wherein the first control point is the starting point of the preset numerical range, and the second control point is the end point of the preset numerical range; numbering the plurality of segmented first medical sample images and randomly scrambling the order of the numbering; and mapping the plurality of segmented first medical sample images according to the randomly disturbed numbering sequence to obtain the processed first medical sample image.
Further, training an initial evaluator in accordance with the first medical sample image, the obtaining the target evaluator comprises: performing error evaluation on the processed first medical sample image and the processed second medical sample image through the initial evaluator to obtain a prediction error map; calculating a loss value according to the prediction error map and the real error map; training the initial evaluator according to the loss value to obtain the target evaluator.
Further, after training an initial evaluator from the first medical sample image to obtain the target evaluator, the method further comprises: acquiring a second medical sample image pair, wherein the second medical sample image pair is composed of a third medical sample image of the first modality and a fourth medical sample image of the second modality; training an initial registration network according to the second medical sample image pair and the target evaluator to obtain the target registration network.
Further, training an initial registration network in accordance with the second medical sample image pair and the target evaluator, the obtaining a target registration network comprising: inputting the second medical sample image pair into the initial registration network to obtain a deformation field from the third medical sample image to the fourth medical sample image, and registering the third medical sample image according to the deformation field to obtain a registered third medical sample image of a second mode; performing similarity evaluation on a third medical sample image of the second modality and a fourth medical sample image of the second modality according to the target evaluator to obtain an evaluation result; acquiring a smooth loss function of the deformation field; and training the initial registration network according to the evaluation result and the smooth loss function to obtain the target registration network.
Further, registering the target medical image of the first modality through a target registration network, and obtaining a registered target medical image of the second modality includes: inputting the target medical image of the first modality into the target registration network, and obtaining a target deformation field for registering the target medical image of the first modality to the target medical image of the second modality through the target registration network; registering the target medical image of the first modality according to the target deformation field to obtain a registered target medical image of the second modality.
In order to achieve the above object, according to another aspect of the present application, there is provided a medical image processing apparatus. The device comprises: a first acquisition unit configured to acquire a target medical image of a first modality; the registration unit is used for registering the target medical image of the first modality through a target registration network to obtain a target medical image of a registered second modality, wherein the target registration network is trained through a target evaluator, the target evaluator is trained by a first medical sample image pair, the first medical sample image pair consists of a first medical sample image and a second medical sample image, and the first medical sample image and the second medical sample image are images of the same modality.
Further, the apparatus further comprises: the second acquisition unit is used for acquiring the first medical sample image and the second medical sample image before registering the target medical image of the first modality through a target registration network, and calculating to obtain a real error map between the first medical sample image and the second medical sample image; the processing unit is used for carrying out image enhancement processing on the first medical sample image to obtain a processed first medical sample image; a confirmation unit for determining the processed first medical sample image, the second medical sample image and the true error map as a first medical sample image pair; and the first training unit is used for training the initial evaluator according to the first medical sample image to obtain the target evaluator.
Further, the second acquisition unit includes: the first acquisition module is used for acquiring an original medical sample image; the first transformation module is used for carrying out first random space transformation on the original medical sample image to obtain the first medical sample image; and the second transformation module is used for carrying out second random space transformation on the original medical sample image to obtain the second medical sample image.
Further, the processing unit includes: the normalization module normalizes the pixel points on the first medical sample image to a preset numerical range to obtain a normalized first medical sample image; the dividing module is used for randomly generating a plurality of control points in the preset numerical range, dividing the normalized first medical sample image according to the plurality of control points to obtain a plurality of segmented first medical sample images, wherein the first control point is the starting point of the preset numerical range, and the second control point is the end point of the preset numerical range; the numbering module is used for numbering the first medical sample images of the plurality of segments and randomly scrambling the sequence of the numbering; and the mapping module is used for mapping the plurality of segmented first medical sample images according to the number sequence which is randomly disturbed to obtain the processed first medical sample image.
Further, the first training unit includes: the first evaluation module is used for carrying out error evaluation on the processed first medical sample image and the processed second medical sample image through the initial evaluator to obtain a prediction error map; the calculation module is used for calculating a loss value according to the prediction error map and the real error map; and the first training module is used for training the initial evaluator according to the loss value to obtain the target evaluator.
Further, the apparatus further comprises: a third obtaining unit, configured to obtain a second medical sample image pair after training an initial evaluator according to the first medical sample image pair to obtain the target evaluator, where the second medical sample image pair is composed of a third medical sample image of a first modality and a fourth medical sample image of a second modality; and the second training unit is used for training the initial registration network according to the second medical sample image pair and the target evaluator to obtain the target registration network.
Further, the second training unit includes: the first registration module is used for inputting the second medical sample image pair into the initial registration network to obtain a deformation field from the third medical sample image to the fourth medical sample image, and registering the third medical sample image according to the deformation field to obtain a registered third medical sample image of the second modality; the second evaluation module is used for evaluating the similarity between the third medical sample image of the second modality and the fourth medical sample image of the second modality according to the target evaluator to obtain an evaluation result; the second acquisition module is used for acquiring the smooth loss function of the deformation field; and the second training module is used for training the initial registration network according to the evaluation result and the smooth loss function to obtain the target registration network.
Further, the registration unit includes: the input module is used for inputting the target medical image of the first modality into the target registration network, and obtaining a target deformation field for registering the target medical image of the first modality to the target medical image of the second modality through the target registration network; and the second configuration module is used for registering the target medical image of the first modality according to the target deformation field to obtain a registered target medical image of the second modality.
In order to achieve the above object, according to another aspect of the present application, there is also provided a computer-readable storage medium storing a program, wherein the program, when run, controls a device in which the storage medium is located to execute the method of processing a medical image as described in any one of the above.
To achieve the above object, according to one aspect of the present application, there is provided an electronic device including one or more processors and a memory for storing a processing method of the one or more processors for implementing the medical image according to any one of the above.
Through the application, the following steps are adopted: acquiring a target medical image of a first modality; registering the target medical image of the first modality through a target registration network to obtain a target medical image of a registered second modality, wherein the target registration network is trained through a target evaluator, the target evaluator is trained by a first medical sample image pair, the first medical sample image pair consists of a first medical sample image and a second medical sample image, and the first medical sample image and the second medical sample image are images of the same absolute modality, so that the problem that the registration accuracy of the registration network to the images is lower due to the fact that the registration network is trained through the existing multi-modality data in the related art is solved. In the scheme, the target evaluator is obtained through the image training of the absolute same modality, so that the target evaluator only focuses on the difference of the spatial positions and does not focus on the difference of multi-modality distribution caused by different image acquisition mechanisms, and the configuration effect of the registration network can be effectively improved by optimizing and training the registration network through the target evaluator, and the effect of improving the accuracy of the registration network on the image registration is further achieved.
Drawings
The accompanying drawings, which are included to provide a further understanding of the application, illustrate and explain the application and are not to be construed as limiting the application. In the drawings:
FIG. 1 is a flow chart of a method of processing medical images provided in accordance with an embodiment of the present application;
FIG. 2 is a schematic diagram of a training derived target evaluator provided in accordance with an embodiment of the present application;
FIG. 3 is a schematic diagram of a training-derived target registration network provided in accordance with an embodiment of the present application;
FIG. 4 is a schematic illustration of a medical image processing apparatus provided according to an embodiment of the present application;
fig. 5 is a schematic diagram of an electronic device provided according to an embodiment of the present application.
Detailed Description
It should be noted that, in the case of no conflict, the embodiments and features in the embodiments may be combined with each other. The present application will be described in detail below with reference to the accompanying drawings in conjunction with embodiments.
In order to make the present application solution better understood by those skilled in the art, the following description will be made in detail and with reference to the accompanying drawings in the embodiments of the present application, it is apparent that the described embodiments are only some embodiments of the present application, not all embodiments. All other embodiments, which can be made by one of ordinary skill in the art based on the embodiments herein without making any inventive effort, shall fall within the scope of the present application.
It should be noted that the terms "first," "second," and the like in the description and claims of the present application and the above figures are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used may be interchanged where appropriate in order to describe the embodiments of the present application described herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.
For convenience of description, the following will describe some terms or terms related to the embodiments of the present application:
the relative homomodality: for any one
If all satisfy specific and identical data distribution rules, i.e
. Then->
Referred to as relative co-modal data with respect to each other, such as MR (magnetic resonance imaging) images taken under different machine parameters.
Absolute co-modal: for any one 
And j+.k, if x j And x k Through only a specific spatial transformationφCan be obtained from each other, i.e.)>
Or->
. Then->
Known as absolute co-modal data with each other.
Multimodal image: let x be j And x k Is an image belonging to the absolute same mode
，φLet x in k Generates random displacements for each pixel of (a). At this time, x j And x k There is only a spatial difference between them. Next, some undefined noise ε is added to x k And (3) the following steps: />
. This noise gives x k ，x j It is possible to have a large gray scale difference even on the same organ tissue. At this time->
And x k The relationship becomes a multi-modal image relationship.
It should be noted that, related information (including, but not limited to, user equipment information, user personal information, etc.) and data (including, but not limited to, data for presentation, analyzed data, medical image data, etc.) related to the present disclosure are information and data authorized by a user or sufficiently authorized by each party. For example, an interface is provided between the system and the relevant user or institution, before acquiring the relevant information, the system needs to send an acquisition request to the user or institution through the interface, and acquire the relevant information after receiving the consent information fed back by the user or institution.
The invention will be described with reference to preferred embodiments, and FIG. 1 is a flowchart of a method for processing medical images according to an embodiment of the present application, as shown in FIG. 1, the method comprising the steps of:
step S101, acquiring a target medical image of a first modality;
specifically, a target medical image of a first modality is acquired, for example, the target medical image of the first modality being an MR medical image.
Step S102, registering a target medical image of a first modality through a target registration network to obtain a registered target medical image of a second modality, wherein the target registration network is trained through a target evaluator, the target evaluator is trained by a first medical sample image pair, the first medical sample image pair consists of a first medical sample image and a second medical sample image, and the first medical sample image and the second medical sample image are images of the same modality in absolute terms.
Specifically, the obtained target medical image of the first modality is input into a target registration network, the target medical image of the first modality is registered by using the target registration network, and a registered target medical image of the second modality is obtained, for example, the medical image of the second modality is a CT medical image.
The target registration network is trained by a target evaluator. The target evaluator is trained by adopting images with the same absolute mode, so that the target evaluator only focuses on the difference of the spatial positions and does not focus on the difference of multi-mode distribution caused by different image acquisition mechanisms, and further the target evaluator can evaluate the difference between the two images more accurately. The registration accuracy of the target registration network can be effectively improved by correspondingly training the registration network through the target evaluator.
How to obtain the target evaluator is crucial, therefore, in the medical image processing method provided in the embodiment of the present application, before registering the target medical image of the first modality through the target registration network, the method further includes: acquiring a first medical sample image and a second medical sample image, and calculating to obtain a real error map between the first medical sample image and the second medical sample image; performing image enhancement processing on the first medical sample image to obtain a processed first medical sample image; determining the processed first medical sample image, the processed second medical sample image and the processed true error map as a first medical sample image pair; training the initial evaluator according to the first medical sample image to obtain a target evaluator.
Specifically, a first medical sample image and a second medical sample image of the same absolute modality are obtained, an absolute error is calculated on the first medical sample image and the second medical sample image, and then an error map with the same dimension as the first medical sample image can be accurately obtained, and the error map can be used as a real label between the first medical sample image and the second medical sample image. Since noise epsilon is added to the medical image at the time of generation of the actual medical image, in order to enable the evaluator to accurately evaluate the difference between the noisy first medical sample image and the second medical sample image, the image enhancement of the first medical sample image is implemented. By performing image enhancement on the first medical sample image, the diversity of the training samples can be greatly increased, and therefore, the accuracy of the evaluation of the evaluator can be improved.
The processed first medical sample image (i.e., the image-enhanced first medical sample image), the second medical sample image, and the true error map are determined as a first medical sample image pair. The initial evaluator is trained through the first medical sample image pair to obtain a target evaluator.
In the medical image processing method provided by the embodiment of the application, the following steps are adopted to acquire a first medical sample image and a second medical sample image: acquiring an original medical sample image; performing first random spatial transformation on the original medical sample image to obtain a first medical sample image; and performing second random spatial transformation on the original medical sample image to obtain a second medical sample image.
Specifically, an original medical sample image is acquired, and a first random spatial variation is added on the original medical sample image
Obtaining a first medical sample image, which is in the original stateAdding a second random spatial variation to the medical sample image>
A second medical sample image is obtained and random spatial variation operations include, but are not limited to, rotation of the whole, translation, scaling, random deformation displacement of each pixel, and the like. Because the first medical sample image and the second medical sample image are from the same original medical sample image, only spatial errors are satisfied between them and the absolute co-modal definition is satisfied.
How to perform image enhancement to increase the diversity of training samples is important, so in the medical image processing method provided in the embodiment of the present application, performing image enhancement processing on a first medical sample image, to obtain a processed first medical sample image includes: normalizing pixel points on the first medical sample image to a preset numerical range to obtain a normalized first medical sample image; randomly generating a plurality of control points in a preset numerical range, and dividing the normalized first medical sample image according to the plurality of control points to obtain a plurality of segmented first medical sample images, wherein the first control point is the starting point of the preset numerical range, and the second control point is the end point of the preset numerical range; numbering the first medical sample images of the plurality of segments, and randomly disturbing the sequence of the numbering; and mapping the plurality of segmented first medical sample images according to the randomly disturbed numbering sequence to obtain a processed first medical sample image.
Specifically, pixel points on the first medical sample image are normalized to be within a preset numerical range, and a normalized first medical sample image is obtained. For example, pixels on the first medical sample image are normalized to between [ -1,1 ]. A plurality of control points are randomly generated within a preset numerical range. For example, P0 to PN control points are randomly generated between [ -1,1], where P0, PN are endpoints-1 and 1, and the locations of the remaining Pn points are completely random. The normalized first medical sample image is divided into a plurality of segmented first medical sample images through the control points, the segmented first medical sample images are numbered according to the sequence from small to large, the numbers are randomly disordered, and the disordered sequence is remapped to obtain the processed first medical sample image. For example, pixel x in the range of (Pi, pi+1) is remapped to the range of (Pj, pj+1):
wherein x' is the pixel point after mapping, the pixel point before x mapping,p i for the i-th control point,p j for the j-th control point,p i+1 for the (i + 1) th control point,p j+1 for the j+1th control point, the above formula is to randomly map the pixels x in the range of (Pi, pi+1) to the range of (Pj, pj+1), so as to realize the image enhancement of the first medical sample image.
Through the steps, the diversity of the first medical sample image is effectively improved, the first medical sample image of any mode can be obtained, and the practicability and the robustness of the target evaluator can be effectively improved.
Optionally, in the medical image processing method provided in the embodiment of the present application, training the initial evaluator according to the first medical sample image to obtain the target evaluator includes: performing error evaluation on the processed first medical sample image and the processed second medical sample image through an initial evaluator to obtain a prediction error map; calculating according to the prediction error map and the real error map to obtain a loss value; training the initial estimator according to the loss value to obtain a target estimator.
Specifically, the processed first medical sample image and the processed second medical sample image are input into an initial evaluator, a prediction error map is obtained through the initial evaluator, a prediction error map is calculated, then a loss value between the prediction error map and a real error map is calculated, the initial evaluator is trained through minimizing the loss value, and then a target evaluator is obtained. In an alternative embodiment, the loss function may be as follows:
Wherein, the liquid crystal display device comprises a liquid crystal display device,
for the prediction error map, |x1-x2| is the true error map described above. x1 is the first medical sample image, x2 is the second medical sample image, and x1+ε is the processed first medical sample image.
In an alternative embodiment, the target estimator may be obtained using a schematic diagram as shown in fig. 2, applying two random spatial transformations to one original image x, respectively
And->
And transformed x1 and x2 are obtained. Wherein the spatial transformation operation includes overall rotation, translation, scaling and random deformation displacement of each pixel. Because x1 and x2 are from the same image x, there is only a spatial error between them and the absolute co-modal definition is satisfied. Next, an absolute error calculation is performed on x1 and x2 to accurately obtain an error map of the same dimension as x and treat it as a label. A random noise epsilon is added to x1 to change the modal distribution relationship between x1 and x2, and the data preparation before the training is completed. And stacking the x1+epsilon and x2 as the input of the initial evaluator E, and finally carrying out loss calculation on the output predicted by the initial evaluator and the label manufactured before, so as to realize calculation on the initial evaluator and obtain the target evaluator.
After obtaining the target evaluator, training the target registration network by the target evaluator includes: acquiring a second medical sample image pair, wherein the second medical sample image pair consists of a third medical sample image of the first modality and a fourth medical sample image of the second modality; training the initial registration network according to the second medical sample image pair and the target evaluator to obtain a target registration network.
Training the initial registration network in accordance with the second medical sample image pair and the target evaluator, the obtaining the target registration network comprising: inputting the second medical sample image pair into an initial registration network to obtain a deformation field from a third medical sample image to a fourth medical sample image, and registering the third medical sample image according to the deformation field to obtain a registered third medical sample image of a second mode; performing similarity evaluation on a third medical sample image of the second mode and a fourth medical sample image of the second mode according to the target evaluator to obtain an evaluation result; acquiring a smooth loss function of the deformation field; training the initial registration network according to the evaluation result and the smooth loss function to obtain a target registration network.
Specifically, a second medical sample image pair is formed by a third medical sample image of a first modality and a fourth medical sample image of a second modality, then the second medical sample image pair is input into an initial registration network to obtain a deformation field from the third medical sample image to the fourth medical sample image, the third medical sample image is registered according to the deformation field to obtain a registered third medical sample image of the second modality, then the registered third medical sample image of the second modality and the fourth medical sample image are input into a target evaluator, and similarity evaluation is carried out on the third medical sample image of the second modality and the fourth medical sample image of the second modality by using the target evaluator to obtain an evaluation result. And finally, training the initial registration network through the evaluation result and the smoothing loss function to obtain the target registration network.
In an alternative embodiment, the target registration network may be trained using a schematic diagram as shown in fig. 3, with a real moving image x (registration image), a fixed image y (fixed image). Taking x and y as input of the registration network R and obtaining deformation field phi from x to y x2y Warping x with deformation fields to obtain registered image xophi x2y . Next, xO phi will be x2y And inputting the fixed image y into a target evaluator to obtain an error similarity evaluation result. Finally only the error is required to be minimizedAnd updating the weight of the registration network R.
In an alternative embodiment, the optimization function of the error similarity assessment result is as follows:
wherein E (xoR (x, y), y) is the error similarity described above.
In an alternative embodiment, the smoothing loss function is as follows:
where, v represents the gradient operator, and v R (x, y) is the degree of correlation between x and y.
Optionally, in the method for processing a medical image provided in the embodiment of the present application, registering the target medical image of the first modality through the target registration network, obtaining the registered target medical image of the second modality includes: inputting the target medical image of the first modality into a target registration network, and obtaining a target deformation field for registering the target medical image of the first modality to the target medical image of the second modality through the target registration network; registering the target medical image of the first modality according to the target deformation field to obtain a registered target medical image of the second modality.
Specifically, inputting a target medical image of a first modality into a target registration network, obtaining a target deformation field for registering the target medical image of the first modality to a target medical image of a second modality by using the target registration network, and registering the target medical image of the first modality by using the target deformation field to obtain a registered target medical image of the second modality.
According to the medical image processing method, the target medical image of the first mode is acquired; registering the target medical image of the first modality through a target registration network to obtain a target medical image of a registered second modality, wherein the target registration network is trained through a target evaluator, the target evaluator is trained by a first medical sample image pair, the first medical sample image pair consists of a first medical sample image and a second medical sample image, and the first medical sample image and the second medical sample image are images of the same absolute modality, so that the problem that the registration accuracy of the registration network to the images is lower due to the fact that the registration network is trained through the existing multi-modality data in the related art is solved. In the scheme, the target evaluator is obtained through the image training of the absolute same modality, so that the target evaluator only focuses on the difference of the spatial positions and does not focus on the difference of multi-modality distribution caused by different image acquisition mechanisms, and the configuration effect of the registration network can be effectively improved by optimizing and training the registration network through the target evaluator, and the effect of improving the accuracy of the registration network on the image registration is further achieved.
It should be noted that the steps illustrated in the flowcharts of the figures may be performed in a computer system such as a set of computer executable instructions, and that although a logical order is illustrated in the flowcharts, in some cases the steps illustrated or described may be performed in an order other than that illustrated herein.
The embodiment of the application also provides a medical image processing device, and it should be noted that the medical image processing device of the embodiment of the application can be used for executing the medical image processing method provided by the embodiment of the application. The following describes a medical image processing apparatus provided in an embodiment of the present application.
Fig. 4 is a schematic view of a medical image processing apparatus according to an embodiment of the present application. As shown in fig. 4, the apparatus includes: a first acquisition unit 401 and a registration unit 402.
A first acquisition unit 401 for acquiring a target medical image of a first modality;
the registration unit 402 is configured to register the target medical image of the first modality through a target registration network, and obtain a target medical image of a registered second modality, where the target registration network is trained by a target evaluator, and the target evaluator is trained by a first medical sample image pair, and the first medical sample image pair is composed of a first medical sample image and a second medical sample image, and the first medical sample image and the second medical sample image are images of an absolute same modality.
According to the medical image processing device provided by the embodiment of the application, the target medical image of the first modality is acquired through the first acquisition unit 401; the registration unit 402 registers the target medical image of the first modality through a target registration network to obtain a target medical image of a registered second modality, wherein the target registration network is obtained through training of a target evaluator, the target evaluator is obtained through training of a first medical sample image pair, the first medical sample image pair is composed of a first medical sample image and a second medical sample image, the first medical sample image and the second medical sample image are images of the same absolute modality, and the problem that in the related art, the registration accuracy of the registration network to the images is low due to the fact that the registration network is trained through existing multi-modality data is solved. In the scheme, the target evaluator is obtained through the image training of the absolute same modality, so that the target evaluator only focuses on the difference of the spatial positions and does not focus on the difference of multi-modality distribution caused by different image acquisition mechanisms, and the configuration effect of the registration network can be effectively improved by optimizing and training the registration network through the target evaluator, and the effect of improving the accuracy of the registration network on the image registration is further achieved.
Optionally, in the medical image processing apparatus provided in the embodiment of the present application, the apparatus further includes: the second acquisition unit is used for acquiring a first medical sample image and a second medical sample image before registering the target medical image of the first modality through the target registration network, and calculating to obtain a real error map between the first medical sample image and the second medical sample image; the processing unit is used for carrying out image enhancement processing on the first medical sample image to obtain a processed first medical sample image; a confirmation unit for determining the processed first medical sample image, second medical sample image and true error map as a first medical sample image pair; the first training unit is used for training the initial evaluator according to the first medical sample image to obtain the target evaluator.
Optionally, in the medical image processing apparatus provided in the embodiment of the present application, the second acquisition unit includes: the first acquisition module is used for acquiring an original medical sample image; the first transformation module is used for carrying out first random space transformation on the original medical sample image to obtain a first medical sample image; and the second transformation module is used for carrying out second random space transformation on the original medical sample image to obtain a second medical sample image.
Optionally, in the medical image processing apparatus provided in the embodiment of the present application, the processing unit includes: the normalization module normalizes pixel points on the first medical sample image to a preset numerical range to obtain a normalized first medical sample image; the dividing module is used for randomly generating a plurality of control points in a preset numerical range, dividing the normalized first medical sample image according to the plurality of control points to obtain a plurality of segmented first medical sample images, wherein the first control point is the starting point of the preset numerical range, and the second control point is the end point of the preset numerical range; the numbering module is used for numbering the first medical sample images of the plurality of segments and randomly scrambling the numbering sequence; the mapping module is used for mapping the plurality of segmented first medical sample images according to the number sequence which is randomly disturbed to obtain a processed first medical sample image.
Optionally, in the medical image processing apparatus provided in the embodiment of the present application, the first training unit includes: the first evaluation module is used for carrying out error evaluation on the processed first medical sample image and the processed second medical sample image through the initial evaluator to obtain a prediction error map; the calculation module is used for calculating a loss value according to the prediction error map and the real error map; and the first training module is used for training the initial evaluator according to the loss value to obtain the target evaluator.
Optionally, in the medical image processing apparatus provided in the embodiment of the present application, the apparatus further includes: a third obtaining unit, configured to obtain a second medical sample image pair after training the initial evaluator according to the first medical sample image to obtain the target evaluator, where the second medical sample image pair is composed of a third medical sample image of the first modality and a fourth medical sample image of the second modality; and the second training unit is used for training the initial registration network according to the second medical sample image pair and the target evaluator to obtain a target registration network.
Optionally, in the medical image processing apparatus provided in the embodiment of the present application, the second training unit includes: the first registration module is used for inputting the second medical sample image pair into the initial registration network to obtain a deformation field from the third medical sample image to the fourth medical sample image, and registering the third medical sample image according to the deformation field to obtain a registered third medical sample image of the second mode; the second evaluation module is used for evaluating the similarity between the third medical sample image of the second mode and the fourth medical sample image of the second mode according to the target evaluator to obtain an evaluation result; the second acquisition module is used for acquiring a smooth loss function of the deformation field; and the second training module is used for training the initial registration network according to the evaluation result and the smooth loss function to obtain a target registration network.
Optionally, in the medical image processing apparatus provided in the embodiment of the present application, the registration unit includes: the input module is used for inputting the target medical image of the first modality into a target registration network, and obtaining a target deformation field for registering the target medical image of the first modality to the target medical image of the second modality through the target registration network; and the second configuration module is used for registering the target medical image of the first modality according to the target deformation field to obtain a registered target medical image of the second modality.
The processing device for medical images comprises a processor and a memory, wherein the first acquisition unit 401 and the registration unit 402 are stored in the memory as program units, and the processor executes the program units stored in the memory to realize corresponding functions.
The processor includes a kernel, and the kernel fetches the corresponding program unit from the memory. The kernel can be provided with one or more than one, and the registration of the images is realized by adjusting kernel parameters.
The memory may include volatile memory, random Access Memory (RAM), and/or nonvolatile memory, such as Read Only Memory (ROM) or flash memory (flash RAM), among other forms in computer readable media, the memory including at least one memory chip.
An embodiment of the present invention provides a computer-readable storage medium having stored thereon a program which, when executed by a processor, implements a method of processing medical images.
As shown in fig. 5, an embodiment of the present invention provides an electronic device, where the device includes a processor, a memory, and a program stored in the memory and executable on the processor, and when the processor executes the program, the following steps are implemented: acquiring a target medical image of a first modality; registering the target medical image of the first modality through a target registration network to obtain a registered target medical image of the second modality, wherein the target registration network is trained through a target evaluator, the target evaluator is trained by a first medical sample image pair, the first medical sample image pair consists of a first medical sample image and a second medical sample image, and the first medical sample image and the second medical sample image are images of the same modality.
Acquiring a first medical sample image and a second medical sample image, and calculating to obtain a real error map between the first medical sample image and the second medical sample image; performing image enhancement processing on the first medical sample image to obtain a processed first medical sample image; determining the processed first medical sample image, the processed second medical sample image and the processed true error map as a first medical sample image pair; training the initial evaluator according to the first medical sample image to obtain a target evaluator.
Optionally, acquiring the first medical sample image and the second medical sample image comprises: acquiring an original medical sample image; performing first random spatial transformation on the original medical sample image to obtain a first medical sample image; and performing second random spatial transformation on the original medical sample image to obtain a second medical sample image.
Optionally, performing image enhancement processing on the first medical sample image to obtain a processed first medical sample image includes: normalizing pixel points on the first medical sample image to a preset numerical range to obtain a normalized first medical sample image; randomly generating a plurality of control points in a preset numerical range, and dividing the normalized first medical sample image according to the plurality of control points to obtain a plurality of segmented first medical sample images, wherein the first control point is the starting point of the preset numerical range, and the second control point is the end point of the preset numerical range; numbering the first medical sample images of the plurality of segments, and randomly disturbing the sequence of the numbering; and mapping the plurality of segmented first medical sample images according to the randomly disturbed numbering sequence to obtain a processed first medical sample image.
Optionally, training the initial evaluator based on the first medical sample image, the obtaining the target evaluator includes: performing error evaluation on the processed first medical sample image and the processed second medical sample image through an initial evaluator to obtain a prediction error map; calculating according to the prediction error map and the real error map to obtain a loss value; training the initial estimator according to the loss value to obtain a target estimator.
Optionally, after training the initial evaluator from the first medical sample image to obtain the target evaluator, the method further comprises: acquiring a second medical sample image pair, wherein the second medical sample image pair consists of a third medical sample image of the first modality and a fourth medical sample image of the second modality; training the initial registration network according to the second medical sample image pair and the target evaluator to obtain a target registration network.
Optionally, training the initial registration network in accordance with the second medical sample image pair and the target evaluator, the obtaining the target registration network comprising: inputting the second medical sample image pair into an initial registration network to obtain a deformation field from a third medical sample image to a fourth medical sample image, and registering the third medical sample image according to the deformation field to obtain a registered third medical sample image of a second mode; performing similarity evaluation on a third medical sample image of the second mode and a fourth medical sample image of the second mode according to the target evaluator to obtain an evaluation result; acquiring a smooth loss function of the deformation field; training the initial registration network according to the evaluation result and the smooth loss function to obtain a target registration network.
Optionally, registering the target medical image of the first modality through a target registration network, and obtaining the registered target medical image of the second modality includes: inputting the target medical image of the first modality into a target registration network, and obtaining a target deformation field for registering the target medical image of the first modality to the target medical image of the second modality through the target registration network; registering the target medical image of the first modality according to the target deformation field to obtain a registered target medical image of the second modality.
The device herein may be a server, PC, PAD, cell phone, etc.
The present application also provides a computer program product adapted to perform, when executed on a data processing device, a program initialized with the method steps of: acquiring a target medical image of a first modality; registering the target medical image of the first modality through a target registration network to obtain a registered target medical image of the second modality, wherein the target registration network is trained through a target evaluator, the target evaluator is trained by a first medical sample image pair, the first medical sample image pair consists of a first medical sample image and a second medical sample image, and the first medical sample image and the second medical sample image are images of the same modality.
Optionally, acquiring a first medical sample image and a second medical sample image, and calculating to obtain a real error map between the first medical sample image and the second medical sample image; performing image enhancement processing on the first medical sample image to obtain a processed first medical sample image; determining the processed first medical sample image, the processed second medical sample image and the processed true error map as a first medical sample image pair; training the initial evaluator according to the first medical sample image to obtain a target evaluator.
Optionally, acquiring the first medical sample image and the second medical sample image comprises: acquiring an original medical sample image; performing first random spatial transformation on the original medical sample image to obtain a first medical sample image; and performing second random spatial transformation on the original medical sample image to obtain a second medical sample image.
Optionally, performing image enhancement processing on the first medical sample image to obtain a processed first medical sample image includes: normalizing pixel points on the first medical sample image to a preset numerical range to obtain a normalized first medical sample image; randomly generating a plurality of control points in a preset numerical range, and dividing the normalized first medical sample image according to the plurality of control points to obtain a plurality of segmented first medical sample images, wherein the first control point is the starting point of the preset numerical range, and the second control point is the end point of the preset numerical range; numbering the first medical sample images of the plurality of segments, and randomly disturbing the sequence of the numbering; and mapping the plurality of segmented first medical sample images according to the randomly disturbed numbering sequence to obtain a processed first medical sample image.
Optionally, training the initial evaluator based on the first medical sample image, the obtaining the target evaluator includes: performing error evaluation on the processed first medical sample image and the processed second medical sample image through an initial evaluator to obtain a prediction error map; calculating according to the prediction error map and the real error map to obtain a loss value; training the initial estimator according to the loss value to obtain a target estimator.
Optionally, after training the initial evaluator from the first medical sample image to obtain the target evaluator, the method further comprises: acquiring a second medical sample image pair, wherein the second medical sample image pair consists of a third medical sample image of the first modality and a fourth medical sample image of the second modality; training the initial registration network according to the second medical sample image pair and the target evaluator to obtain a target registration network.
Optionally, training the initial registration network in accordance with the second medical sample image pair and the target evaluator, the obtaining the target registration network comprising: inputting the second medical sample image pair into an initial registration network to obtain a deformation field from a third medical sample image to a fourth medical sample image, and registering the third medical sample image according to the deformation field to obtain a registered third medical sample image of a second mode; performing similarity evaluation on a third medical sample image of the second mode and a fourth medical sample image of the second mode according to the target evaluator to obtain an evaluation result; acquiring a smooth loss function of the deformation field; training the initial registration network according to the evaluation result and the smooth loss function to obtain a target registration network.
Optionally, registering the target medical image of the first modality through a target registration network, and obtaining the registered target medical image of the second modality includes: inputting the target medical image of the first modality into a target registration network, and obtaining a target deformation field for registering the target medical image of the first modality to the target medical image of the second modality through the target registration network; registering the target medical image of the first modality according to the target deformation field to obtain a registered target medical image of the second modality.
It will be appreciated by those skilled in the art that embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
In one typical configuration, a computing device includes one or more processors (CPUs), input/output interfaces, network interfaces, and memory.
The memory may include volatile memory in a computer-readable medium, random Access Memory (RAM) and/or nonvolatile memory, etc., such as Read Only Memory (ROM) or flash RAM. Memory is an example of a computer-readable medium.
Computer readable media, including both non-transitory and non-transitory, removable and non-removable media, may implement information storage by any method or technology. The information may be computer readable instructions, data structures, modules of a program, or other data. Examples of storage media for a computer include, but are not limited to, phase change memory (PRAM), static Random Access Memory (SRAM), dynamic Random Access Memory (DRAM), other types of Random Access Memory (RAM), read Only Memory (ROM), electrically Erasable Programmable Read Only Memory (EEPROM), flash memory or other memory technology, compact disc read only memory (CD-ROM), digital Versatile Discs (DVD) or other optical storage, magnetic cassettes, magnetic tape disk storage or other magnetic storage devices, or any other non-transmission medium, which can be used to store information that can be accessed by a computing device. Computer-readable media, as defined herein, does not include transitory computer-readable media (transmission media), such as modulated data signals and carrier waves.
It should also be noted that the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article or apparatus that comprises an element.
It will be appreciated by those skilled in the art that embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The foregoing is merely exemplary of the present application and is not intended to limit the present application. Various modifications and changes may be made to the present application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc. which are within the spirit and principles of the present application are intended to be included within the scope of the claims of the present application.
Claims (11)
1. A method of processing a medical image, comprising:
acquiring a target medical image of a first modality;
registering the target medical image of the first modality through a target registration network to obtain a registered target medical image of a second modality, wherein the target registration network is trained through a target evaluator, the target evaluator is trained by a first medical sample image pair, the first medical sample image pair consists of a first medical sample image and a second medical sample image, and the first medical sample image and the second medical sample image are images of the same modality;
Wherein if x j And x k Can be obtained by only spatial transformation, x is j And x k Images belonging to the absolute same modality;
before registering the target medical image of the first modality through a target registration network, the method further comprises:
acquiring the first medical sample image and the second medical sample image, and calculating to obtain a real error map between the first medical sample image and the second medical sample image;
performing image enhancement processing on the first medical sample image to obtain a processed first medical sample image;
and determining the processed first medical sample image, the second medical sample image and the true error map as a first medical sample image pair.
2. The method of processing medical images according to claim 1, wherein prior to registering the target medical image of the first modality through a target registration network, the method further comprises:
training an initial evaluator according to the first medical sample image to obtain the target evaluator.
3. The method of processing medical images according to claim 2, wherein acquiring the first medical sample image and the second medical sample image comprises:
Acquiring an original medical sample image;
performing first random spatial transformation on the original medical sample image to obtain the first medical sample image;
and performing second random spatial transformation on the original medical sample image to obtain the second medical sample image.
4. The method of processing a medical image according to claim 2, wherein performing image enhancement processing on the first medical sample image to obtain a processed first medical sample image includes:
normalizing the pixel points on the first medical sample image to a preset numerical range to obtain a normalized first medical sample image;
randomly generating a plurality of control points in the preset numerical range, and dividing the normalized first medical sample image according to the plurality of control points to obtain a plurality of segmented first medical sample images, wherein the first control point is the starting point of the preset numerical range, and the second control point is the end point of the preset numerical range;
numbering the plurality of segmented first medical sample images and randomly scrambling the order of the numbering;
and mapping the plurality of segmented first medical sample images according to the randomly disturbed numbering sequence to obtain the processed first medical sample image.
5. The method of processing medical images according to claim 2, wherein training an initial evaluator from the first medical sample image to obtain the target evaluator comprises:
performing error evaluation on the processed first medical sample image and the processed second medical sample image through the initial evaluator to obtain a prediction error map;
calculating a loss value according to the prediction error map and the real error map;
training the initial evaluator according to the loss value to obtain the target evaluator.
6. The method of processing medical images according to claim 1, wherein after training an initial evaluator from the first medical sample image to obtain the target evaluator, the method further comprises:
acquiring a second medical sample image pair, wherein the second medical sample image pair is composed of a third medical sample image of the first modality and a fourth medical sample image of the second modality;
training an initial registration network according to the second medical sample image pair and the target evaluator to obtain the target registration network.
7. The method of claim 6, wherein training an initial registration network based on the second medical sample image pair and the target evaluator to obtain a target registration network comprises:
inputting the second medical sample image pair into the initial registration network to obtain a deformation field from a third medical sample image to a fourth medical sample image, and registering the third medical sample image according to the deformation field to obtain a registered third medical sample image of a second mode;
performing similarity evaluation on a third medical sample image of the second modality and a fourth medical sample image of the second modality according to the target evaluator to obtain an evaluation result;
acquiring a smooth loss function of the deformation field;
and training the initial registration network according to the evaluation result and the smooth loss function to obtain the target registration network.
8. The method of processing medical images according to claim 1, wherein registering the target medical image of the first modality through a target registration network, obtaining a registered target medical image of a second modality comprises:
Inputting the target medical image of the first modality into the target registration network, and obtaining a target deformation field for registering the target medical image of the first modality to the target medical image of the second modality through the target registration network;
registering the target medical image of the first modality according to the target deformation field to obtain a registered target medical image of the second modality.
9. A medical image processing apparatus, comprising:
a first acquisition unit configured to acquire a target medical image of a first modality;
the registration unit is used for registering the target medical image of the first modality through a target registration network to obtain a target medical image of a registered second modality, wherein the target registration network is trained through a target evaluator, the target evaluator is trained by a first medical sample image pair, the first medical sample image pair consists of a first medical sample image and a second medical sample image, and the first medical sample image and the second medical sample image are images of the same absolute modality;
wherein if x j And x k Can be obtained by only spatial transformation, x is j And x k Images belonging to the absolute same modality;
the apparatus further comprises: the second acquisition unit is used for acquiring the first medical sample image and the second medical sample image before registering the target medical image of the first modality through a target registration network, and calculating to obtain a real error map between the first medical sample image and the second medical sample image; the processing unit is used for carrying out image enhancement processing on the first medical sample image to obtain a processed first medical sample image; and the confirming unit is used for confirming the processed first medical sample image, the second medical sample image and the real error mapping image as a first medical sample image pair.
10. A computer-readable storage medium, characterized in that the storage medium stores a program, wherein the program performs the medical image processing method according to any one of claims 1 to 8.
11. An electronic device comprising one or more processors and a memory for storing one or more programs, wherein the one or more programs, when executed by the one or more processors, cause the one or more processors to implement the method of processing medical images of any of claims 1-8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310175697.5A CN115861394B (en) | 2023-02-28 | 2023-02-28 | Medical image processing method and device, storage medium and electronic equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310175697.5A CN115861394B (en) | 2023-02-28 | 2023-02-28 | Medical image processing method and device, storage medium and electronic equipment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115861394A CN115861394A (en) | 2023-03-28 |
| CN115861394B true CN115861394B (en) | 2023-05-05 |
Family
ID=85659301
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310175697.5A Active CN115861394B (en) | 2023-02-28 | 2023-02-28 | Medical image processing method and device, storage medium and electronic equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115861394B (en) |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103948432A (en) * | 2014-04-30 | 2014-07-30 | 深圳先进技术研究院 | Algorithm for augmented reality of three-dimensional endoscopic video and ultrasound image during operation |
| CN109378043A (en) * | 2017-10-13 | 2019-02-22 | 北京昆仑医云科技有限公司 | Medical image based on patient generates the system and method and medium of diagnosis report |
| CN110298871A (en) * | 2019-06-10 | 2019-10-01 | 东软医疗系统股份有限公司 | Method for registering images and device |
| CN111862175B (en) * | 2020-07-13 | 2022-09-13 | 清华大学深圳国际研究生院 | Cross-modal medical image registration method and device based on cyclic canonical training |
| CN113822792A (en) * | 2021-06-15 | 2021-12-21 | 腾讯科技（深圳）有限公司 | Image registration method, device, equipment and storage medium |
-
2023
- 2023-02-28 CN CN202310175697.5A patent/CN115861394B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN115861394A (en) | 2023-03-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Yang et al. | Quicksilver: Fast predictive image registration–a deep learning approach | |
| CN108399452B (en) | Hierarchical learning of weights for neural networks performing multiple analyses | |
| CN110599526B (en) | Image registration method, computer device, and storage medium | |
| Gilbert et al. | Automated left ventricle dimension measurement in 2d cardiac ultrasound via an anatomically meaningful cnn approach | |
| Iglesias | A ready-to-use machine learning tool for symmetric multi-modality registration of brain MRI | |
| CN116563096B (en) | Method and device for determining deformation field for image registration and electronic equipment | |
| Solari et al. | Design strategies for direct multi-scale and multi-orientation feature extraction in the log-polar domain | |
| Stolt-Ansó et al. | Nisf: Neural implicit segmentation functions | |
| CN115861394B (en) | Medical image processing method and device, storage medium and electronic equipment | |
| Nika et al. | Change detection of medical images using dictionary learning techniques and principal component analysis | |
| Svärm et al. | Improving robustness for inter-subject medical image registration using a feature-based approach | |
| Pyatov et al. | Affine registration of histological images using transformer-based feature matching | |
| Menchón-Lara et al. | Fast 4D elastic group-wise image registration. Convolutional interpolation revisited | |
| Öfverstedt et al. | INSPIRE: Intensity and spatial information-based deformable image registration | |
| Zeineldin et al. | Self-supervised iRegNet for the Registration of Longitudinal Brain MRI of Diffuse Glioma Patients | |
| Zhang et al. | A comparative study for non-rigid image registration and rigid image registration | |
| Ireta Muñoz et al. | Point-to-hyperplane ICP: fusing different metric measurements for pose estimation | |
| Almogadwy et al. | A deep learning approach for slice to volume biomedical image integration | |
| Bouza et al. | Geometric deep learning for unsupervised registration of diffusion magnetic resonance images | |
| Tramnitzke et al. | GPU based affine linear image registration using normalized gradient fields | |
| Chessa et al. | A quantitative comparison of speed and reliability for log-polar mapping techniques | |
| Ge et al. | Image registration based on subpixel localization and Cauchy–Schwarz divergence | |
| US11605206B2 (en) | Method and apparatus with human body estimation | |
| Hu et al. | Manifold‐based feature point matching for multi‐modal image registration | |
| Domingos et al. | Local phase-based fast ray features for automatic left ventricle apical view detection in 3D echocardiography |
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 |