77 results about "Cardiac magnetic resonance" patented technology

The invention discloses a heart left ventricle segmentation method based on a deep full convolutional neural network. According to the method, a deep learning idea is introduced into heart magnetic resonance short-axis image left ventricle segmentation; The process is mainly divided into a training stage and a prediction stage, in the training stage, a preprocessed 128 * 128 heart magnetic resonance image serves as input, a manually processed label serves as a label of a network to be used for calculating errors, and along with increase of training iteration times, the error of a training setand the error of a verification set are gradually reduced; And in the test stage, inputting data in the test set into the trained model, and finally outputting prediction of each pixel by the networkto generate a segmentation result. According to the method, segmentation of the heart magnetic resonance short-axis image is achieved from the perspective of data driving, the problem that manual outline drawing is time-consuming and labor-consuming is effectively solved, the defects of a traditional image segmentation algorithm can be overcome, and high-precision and high-robustness left ventricle segmentation is achieved.

The invention discloses a method for fully-automatically segmenting and quantifying the left ventricle of a cardiac magnetic resonance image, comprising the following specific steps of: carrying out automatic denoising and edge enhancement processing on a 4D (four-dimensional) cardiac magnetic resonance image first; automatically and preprimarily determining the center of the left ventricle by utilizing Hough transform, and implementing a region growing technology by taking the center of the left ventricle as a starting seed point to provide a left ventricle full-voxel blood region, and taking the mass center of the left ventricle full-voxel blood region as a center of the left ventricle in the current layer; finding out the center of each layer by using a seed propagation technology; implementing a region growing technology based on an iterative falling threshold by taking the center of the left ventricle of each layer as a starting seed point to automatically provide a left ventricle blood region of each layer, and calculating the area of the left ventricle blood region of each layer; automatically segmenting the top of the left ventricle according to the time-space continuity of the area of the left ventricle and calculating the area of the top of the left ventricle; positioning the bottom of the left ventricle according to the area and shape time-space continuity of the left ventricle, and automatically segmenting the bottom of the left ventricle of the heart by adopting a region growing technology which is constricted by the left ventricle shape with time-space continuity, and calculating the area of the bottom of the left ventricle; and finally, realizing whole segmentation of the left ventricle image. The method disclosed by the invention is of a fully-automaticprocess without any manual intervene.

A system and method for providing a dark-blood technique for contrast-enhanced cardiac magnetic resonance, improving visualization of subendocardial infarcts or perfusion abnormalities that may otherwise be difficult to distinguish from the bright blood pool. In one technique the dark-blood preparation is performed using a driven-equilibrium fourier transform (DEFT) preparation with motion sensitizing gradients which attenuate the signal in the ventricular cavities related to incoherent phase losses resulting from non-steady flow within the heart. This dark-blood preparation preserves the underlying contrast characteristics of the pulse sequence causing a myocardial infarction to be bright while rendering the blood pool dark. When applied to perfusion imaging, this dark-blood preparation will help eliminate artifacts resulting from the juxtaposition of a bright ventricular cavity and relatively dark myocardium.

The disclosure relates to a method of automatically producing a three-dimensional (3D) segmentation of a heart chamber, the method comprising: obtaining data sets from cardiac magnetic resonance imaging (MRI) or ultrasound, generating a 3D segmentation of the heart chamber from the data sets using an active contour method, modifying the 3D segmentation by adding a plurality of intra-chamber structures; and identifying an enclosing myocardium using the 3D segmentation generated by the method.

The invention provides an automatic right ventricle segmentation method based on deep learning, and the method is characterized in that the method comprises the following steps: carrying out the preprocessing of all collected heart magnetic resonance film short-axis images; extracting the ROI of the region of interest; expanding the data set; constructing a U-shaped network model in a keras library of deep learning, and training the U-shaped network model; and performing result prediction by using the trained U-shaped network model. The invention provides a full-automatic right ventricle segmentation method based on deep learning. Firstly, the influence of surrounding tissues on a segmentation result is reduced by automatically identifying a ventricle region, and an original U-shaped network is improved, so that accurate and full-automatic segmentation of a right ventricle is realized, and a basis is provided for further functional analysis and disease diagnosis of the heart.

A system and method for providing a dark-blood technique for contrast-enhanced cardiac magnetic resonance, improving visualization of subendocardial infarcts or perfusion abnormalities that may otherwise be difficult to distinguish from the bright blood pool. In one technique the dark-blood preparation is performed using a driven-equilibrium fourier transform (DEFT) preparation with motion sensitizing gradients which attenuate the signal in the ventricular cavities related to incoherent phase losses resulting from non-steady flow within the heart. This dark-blood preparation preserves the underlying contrast characteristics of the pulse sequence causing a myocardial infarction to be bright while rendering the blood pool dark. When applied to perfusion imaging, this dark-blood preparation will help eliminate artifacts resulting from the juxtaposition of a bright ventricular cavity and relatively dark myocardium.

A system automatically calculates optimal protocol parameters for dark-blood (DB) preparation and inversion recovery. The system automatically determines pulse sequence timing parameters for MR imaging with blood related signal suppression. The system comprises an acquisition processor for acquiring data indicating a patient heart rate. A pulse timing processor automatically determines an acquisition time of an image data set readout, relative to a blood signal suppression related magnetization preparation pulse sequence, by calculating the acquisition time in response to inputs including, (a) the acquired patient heart rate, (b) data indicating a type of image contrast of the pulse sequence employed and (c) data indicating whether an anatomical signal suppression related magnetization preparation pulse sequence used has a slice selective, or non-slice selective, data acquisition readout.

The invention relates to a method and system for reconstructing a three-dimensional left ventricular profile of a cardiac magnetic resonance image. The method includes the step of building a mixture Gauss model of the cardiac magnetic resonance image, the step of initializing a movable profile model, the step of determining left ventricular inner and outer surface profiles, and the step of reconstructing the three-dimensional left ventricular profile. According to the method and system for reconstructing the three-dimensional left ventricular profile of the cardiac magnetic resonance image, the magnetic resonance image is divided into multiple areas by means of the mixture Gauss model, then the movable profile model is initialized by means of a movable square method, the left ventricular inner and outer surface profiles are obtained by solving an energy minimization equation by means of the movable profile model, and then the three-dimensional profile is reconstructed through the left ventricular inner and outer surface profiles. Due to the fact that the left ventricular inner and outer surface profiles obtained through the mixture Gauss model and the movable profile model are relatively accurate, accuracy of the reconstructed profile is relatively high.

A method is proposed for identifying an anatomical structure within a spatial-temporal image (i.e. a series of frames captured as respective times). A current frame of spatial-temporal medical image is processed using information from one or more previous and / or subsequent temporal frames, to aid in the segmentation of an object or a region of interest (ROI) in a current frame. The invention is applicable to both two- and three-dimensional spatial-temporal images (i.e., 2D+time or 3D+time), and in particular to cardiac magnetic resonance (CMR images). An initialisation process for this method segments the left ventricle (LV) in a CMR image by a fuzzy c-means (FCM) clustering algorithm which employs a circular shape function as part of the definition of the dissimilarity measure.

A hybrid filtering method called Karhunen Loeve Transform-Wavelet (KW) filtering is presented to de-noise dynamic cardiac magnetic resonance images that simultaneously takes advantage of the intrinsic spatial and temporal redundancies of real-time cardiac cine. This filtering technique combines a temporal Karhunen-Loeve transform (KLT) and spatial adaptive wavelet filtering. KW filtering has four steps. The first is applying the KLT along the temporal direction, generating a series of “eigenimages”. The second is applying Marcenko-Pastur (MP) law to identify and discard the noise-only eigenimages. The third applying a 2-D spatial wavelet filter with adaptive threshold to each eigenimage to define the wavelet filter strength for each of the eigenimages based on the noise variance and standard deviation of the signal. Lastly, the inverse KLT is applied to the filtered eigenimages to generate a new series of cine images with reduced image noise.

A method and system for retrospective image combination for free-breathing magnetic resonance (MR) images is disclose. A free-breathing cardiac MR image acquisition including a plurality of frames is received. A key frame is selected of the plurality of frames. A deformation field for each frame to register each frame with the key frame. A weight is determined for each pixel in each frame based on the deformation field for each frame under a minimum total deformation constraint. A combination image is then generated as a weighted average of the frames using the weight determined for each pixel in each frame.

The invention provides a right ventricle multi-map partitioning method based on a cardiac magnetic resonance movie minor-axis image. A magnetic resonance imaging system is used for collecting a certain number of heart original magnetic resonance images of a tested person, and a region of interest is extracted. A fixed number of map images of right ventricle are selected from the original magneticresonance images to be added into a map set, and an expert manual partitioning result is obtained by an expert manually partitioning the map image. The map images and target images obtain a right ventricle coarse partitioning result by adopting B sample conversion based on normalized mutual information, COLLATE fusion is adopted for the coarse partitioning result, firstly log likelihood estimationis carried out on complete data, and then iterative solution is carried out by using a maximum expectation algorithm until convergence, so that the right ventricle final partitioning result is obtained through amending treatment. The method has higher robustness, the accuracy and precision of fusion can be improved, and the method is used for accurately partitioning the heart right ventricle minor-axis image.

The invention discloses a three-dimensional myocardial deformation strain calculation method. The invention divides the myocardium into a plurality of local (sub) regions according to the time and space coordinates on the basis of completing the segmentation processing of the marked cardiac magnetic resonance image sequence, a feed forward neural network (BPNN) and a polynomial or support vector machine (SVM) are used for fitting a local displacement field, the calculation of the established local continuous displacement field is done by the Newton iterative method, the motion parameters of the myocardial particle are calculated and the non-linear interpolation technology is used for calculating the myocardial strain parameters. The invention has clear physical significance and simple and effective algorithm, a model is applicable to parallel calculation, which can calculate the forward and backward motion of arbitrary myocardial particle; and the strain calculation results can be used as the judgment criteria of the three-dimensional cardiac ultrasound strain measurement results.

The invention relates to a heart magnetic resonance image analysis and cardiomyopathy prediction method and device, and the method carries out the processing through a trained neural network, and theprocessing at least comprises the steps: obtaining a first image at least comprising a heart region according to a heart magnetic resonance image; obtaining a heart index according to the first image;and performing classification processing according to the heart indexes to obtain a classification result. In the embodiment of the invention, the artificial intelligence analysis technology of the heart magnetic resonance image is adopted, the full automation of image analysis can be realized through the trained neural network, the analysis of the heart magnetic resonance image can be quickly completed, and higher accuracy can be achieved.

The embodiment of the invention discloses a radial golden angle magnetic resonance heart film imaging method and device and equipment. The method comprises the steps of obtaining magnetic resonance K-space data sampled by a radial golden angle with a preset channel number of a target object; performing re-arranging on the K-space data to obtain preset frame number undersampled K-space data; inputting the preset frame number of undersampled K-space data to a cross region image rebuilding network to obtain a target image. According to the embodiment, the problem that long time is consumed in theimage rebuilding process on the basis of the radial golden angle sampled data is solved; the purpose of shortening the image rebuilding time of heart magnetic resonance data sampled by the radial golden angle can be achieved, and the quality of a rebuilt image is improved.

An exemplary system, method and computer-accessible medium can be provided for generating an image(s) of a tissue(s) that can include, for example, receiving magnetic resonance imaging information regarding the tissue(s) based on a golden-angle radial sampling procedure, sorting and synchronizing the MRI information into at least two dimensions, and generating the image(s) of the tissue(s) based on the sorted and synchronized information. The tissue(s) include cardiac tissue and respiratory-affected tissue. The MRI information can include a motion of the cardiac tissue and a motion of the respiratory tissue.

The invention belongs to the technical field of medical imaging, and particularly relates to a cardiac magnetic resonance image segmentation method. A cardiac anatomic structure-based left and right ventricle level set segmentation method specifically comprises the following steps: performing geometry decomposition to a cardiac structure; resolving a convex hull of an initialized contour; producing a signaled distance function; and generating intima and extima through a dual-level set algorithm, and the like. By using the method, the simultaneous segmentation of the ventricular intima and extima and cardiac muscle can be realized, and the segmentation results can meet the physiological structure characteristics of a heart as far as possible.

The invention provides a cardiac magnetic resonance real-time film imaging method and system, and is suitable for the technical field of medical information. The method comprises the following steps of: monitoring respiratory movement of a subject; judging whether the respiratory movement enters to the last period of expiration or not; and if so, fixing each cardiac imaging layer one by one, collecting cardiac film imaging data in one cardiac cycle until all imaging layers of the whole cardiac are collected, and if not, continuously monitoring the respiratory movement of the subject. The method is in combination with a respiratory navigation technology; the cardiac real-time film imaging data of the last period of expiration can be directly collected; the complex posttreatment processes needed by the existing real-time film imaging data are avoided; the imaging collection time is not obviously prolonged; the data processing procedure is greatly simplified; the method is closer to clinical practical application, simple and convenient; and the method has excellent practical applicability and excellent clinical application prospect.

The invention relates to a cardiac nuclear magnetic resonance image segmentation method which comprises the following steps of: 1, carrying out Gaussian filtering pretreatment on an image; 2, calculating an external force field of an edge for keeping a normal gradient vector flow; 3, defining an initialization outline position of an inner membrane of the cardiac ventriculus sinister; 4, adding a circular energy constraint in a curve evolution process, and segmenting to obtain the inner membrane; 5, defining a final segmentation outline result of the inner membrane as an initialization outlineposition of an outer membrane; 6, setting the edge strength of a zone enclosed by the inner membrane outline in an original edge graph as 0, and re-calculating the external force field; and 7, addinga circular energy constraint in a curve evolution process, and segmenting to obtain the outer membrane. The invention has the advantages of large capturing range, strong noise proof capacity, better robustness to weak edge leakage, and capability of accurately segmenting the inner membrane and the outer membrane of the ventriculus sinister wall.

The invention discloses a method for segmenting the endocardium and the epicardium in a heart cardiac function magnetic resonance image. The method comprises positioning the diastasis in heart magnetic resonance images I<NP> including several lamellas of the left ventricular myocardium and in different cardiac cycle phases, and obtaining a coarse segmentation result of blood pools of magnetic resonance images of N lamellas in the diastasis; converting image data in the left ventricle region-of-interest in the magnetic resonance image of each lamella in the diastasis into a two-dimension polar coordinate conversion image by means of ray scanning based on a polar coordinate conversion method; detecting the endocardium and the epicardium in the two-dimension polar coordinate conversion image based on a bidynamic programming method; obtaining the endocardium and the epicardium in an original lamella image by means of polar coordinate inverse conversion, calculating the convex hull and smoothing the convex hull, and completing segmentation of the endocardium and the epicardium in the magnetic resonance images of the N lamellas in the diastasis; and deriving the segmentation result of the in the endocardium and the epicardium in the magnetic resonance images in the diastasis to the endocardium and the epicardium in the magnetic resonance images in other cardiac cycle phases.

The invention provides a heart segmentation model and pathology classification model training, heart segmentation and pathology classification method and device based on heart MRI (Magnetic Resonance Imaging), and the method comprises the steps: suppressing a residual background part with a small pixel gray level change through a standard deviation filter, highlighting a left ventricle, a right ventricle and a myocardial ,the central position of the left ventricular myocardial wall being further obtained through canny edge detection and circular Hough transform, drawing a rectangular mask, the two-dimensional image being cut based on the rectangular mask to serve as input for training a preset neural network model for training. Background interference can be greatly inhibited, and fast convergence of neural network training is promoted. The pathology classification model training method comprises the following steps: segmenting a two-dimensional image obtained by segmenting each frame of cardiac magnetic resonance imaging short axis in a cardiac cycle based on a cardiac segmentation model, calculating classification feature values, and constructing a random forest based on the classification feature values of a plurality of samples and pathology classification to obtain a cardiac pathology classification model; and realizing automatic pathological classification.

The invention discloses a non-invasive cardiac infarction classification model construction method based on a stack type self-encoder and a support vector machine. At a training stage, movement information of a heart is obtained by using a collected cardiac magnetic resonance image; a self-encoder is trained by using an image block based on fusion of image information and movement information as an input and a corresponding infarction situation as a tag; inputted data are preprocessed by using a denoising self-encoder; all variable factors of the upper layer are utilized during the high-dimensional information learning process and thus deep features of the inputted data are learned; and then with a support vector machine, sample classification is carried out by using the learned deep features as the input and a corresponding tag. According to the invention, cardiac infarction classification prediction is realized from the perspective of data driving, so that a problem of time and effort wasting caused by infarction prediction with contrast agent injection clinically is solved.

A system and method for regression-based segmentation of the mitral valve in 2D+t cardiac magnetic resonance (CMR) slices is disclosed. The 2D+t CMR slices are acquired according to a mitral valve-specific acquisition protocol introduced herein. A set of mitral valve landmarks is detected in each 2D CMR slice and mitral valve contours are estimated in each 2D CMR slice based on the detected landmarks. A full mitral valve model is reconstructed from the mitral valve contours estimated in the 2D CMR slices using a trained regression model. Each 2D CMR slice may be a cine image acquired over a full cardiac cycle. In this case, the segmentation method reconstructs a patient-specific 4D dynamic mitral valve model from the 2D+t CMR image data.

The invention provides a data processing method and a data processing system for cardiac magnetic resonance real-time film imaging. The method comprises the following steps that a selected imaging layer is obtained, and an image of the imaging layer is read; a reference line set on the image is obtained; a dynamic variation chart of a signal contour of the reference line along with acquisition time is obtained; a time interval from the beginning to the end of the end exhalation period is obtained from the dynamic variation chart; the diameter of the left ventricle within the time interval is calculated through the dynamic variation chart, and the heart phases of the end diastolic and the end systolic are determined in accordance with the diameter of the left ventricle; and images corresponding to the heart phases of the end diastolic and the end systolic are extracted and saved. By adopting the data processing method and the data processing system, the data processing speed can be increased.

The invention discloses an intelligent left ventricle magnetic resonance image classification method and device, equipment and a medium. The method comprises the steps of obtaining a first word embedding vector corresponding to first data of a detection object, a second word embedding vector corresponding to second data of the detection object, and a numerical feature vector spliced by third data; extracting a short-axis video and a long-axis video in the heart magnetic resonance image of the detected object, and performing preprocessing to obtain target video data; and splicing and inputting the three vectors into the first feature analysis model to obtain a first feature analysis result, inputting the target video data into the second feature analysis model to obtain a second feature analysis result, splicing the first feature analysis result and the second feature analysis result to obtain a third feature analysis result, and inputting the third feature analysis result into the third feature analysis model to obtain a classification probability value of the left ventricle image of the detected object. According to the method, the multi-sequence nuclear magnetic image data and the clinical data are utilized to jointly perform classification prediction, the result is more reliable and accurate, a complex post-processing flow is not needed, accumulated errors do not exist, and the robustness is improved.

The invention discloses a heart magnetic resonance imaging method based on a multi-scale low-rank model, and provides a new method for heart magnetic resonance imaging researching. A radial sampling track is adopted to realize undersampling of heart K space data, and full heart data radial undersampling is realized, and thereafter magnetic resonance data acquisition speed is accelerated, and then scanning time of a magnetic resonance device is reduced. The minimization expression of the heart magnetic resonance data based on multi-scale low-rank decomposition is used to improve the precision of the magnetic resonance imaging. A convex optimization problem of a reconstructed image is solved by an alternating direction multiplier method, and a speed of magnetic resonance imaging reconstruction is improved, and therefore the texture of the result of the reconstructed image is clear, and the edge is smooth.

The invention relates to a left ventricular segment recognition method and device, electronic equipment and a storage medium, and the method comprises the steps of obtaining a target image which comprises a cardiac magnetic resonance film short-axis image; segmenting the target image to obtain a plurality of masks of a plurality of target areas in the target image; obtaining a plurality of target contours according to the plurality of masks; and performing four-dimensional reconstruction on the target image comprising the plurality of target contours and then determining the left ventricular segment in the target image. According to the embodiment of the invention, the left ventricular segment can be quickly, automatically and accurately identified, so that a clinician can accurately judge the function and movement condition of each myocardial segment of the left ventricle, the efficiency of the clinician is greatly improved, and the invention has extremely high value in clinical application of heart magnetic resonance images.

The invention relates to a heart magnetic resonance image data enhancement method based on evolutionary GAN. According to the method, when a generator is trained, the generator is mutated to generatea plurality of sub-generation generators; the adaptability scores of the generators are judged through an adaptability score function; an optimal filial generation generator is selected as a parent generation generator of the next iteration according to the score; meanwhile, in the discriminator training stage, a new training sample is synthesized in combination with linear interpolation of the feature vector, and a related linear interpolation label is generated, so that the distribution of the whole training set is expanded, the discrete sample space is continued, the inter-domain smoothnessis improved, and the model can be better trained. According to the image enhancement method, high-quality and diverse samples can be generated to expand the training set, and finally various indexesof the classification result are improved.