30 results about "Vascular segmentation" patented technology

A technique is provided for automatically generating a bone mask in CTA angiography. In accordance with the technique, an image data set may be pre-processed to accomplish a variety of function, such as removal of image data associated with the table, partitioning the volume into regionally consistent sub-volumes, computing structures edges based on gradients, and / or calculating seed points for subsequent region growing. The pre-processed data may then be automatically segmented for bone and vascular structure. The automatic vascular segmentation may be accomplished using constrained region growing in which the constraints are dynamically updated based upon local statistics of the image data. The vascular structure may be subtracted from the bone structure to generate a bone mask. The bone mask may in turn be subtracted from the image data set to generate a bone-free CTA volume for reconstruction of volume renderings.

A method of segmenting tubular structures in digital images includes selecting a point in an image of a tubular object to be segmented, defining an initial state of the selected point, initializing measurement weights, a conditional probability distribution and a prior probability distribution of a feature space of the initial state, sampling the feature space from the prior probability distribution, estimating a posterior probability distribution by summing sample measurements weighted by the measurement weights, and segmenting a cross section of the tubular object from the posterior probability distribution.

A technique is provided for automatically generating a bone mask in CTA angiography. In accordance with the technique, an image data set may be pre-processed to accomplish a variety of function, such as removal of image data associated with the table, partitioning the volume into regionally consistent sub-volumes, computing structures edges based on gradients, and / or calculating seed points for subsequent region growing. The pre-processed data may then be automatically segmented for bone and vascular structure. The automatic vascular segmentation may be accomplished using constrained region growing in which the constraints are dynamically updated based upon local statistics of the image data. The vascular structure may be subtracted from the bone structure to generate a bone mask. The bone mask may in turn be subtracted from the image data set to generate a bone-free CTA volume for reconstruction of volume renderings.

The invention relates to a medical image processing technology, and provides a fundus image-oriented microaneurysm detection method for overcoming the defects of detection of a microaneurysm which isa micro target in the prior art, so as to effectively detect the microaneurysm in a fundus image, and better assist a doctor in diagnosis while realizing automatic detection. The method comprises: preprocessing the fundus image, so that the features of a small target are enhanced, and making and training a data set on a built basic feature extraction network; during detection, extracting image basic features from the input image, and performing blood vessel segmentation on the input image by utilizing a segmentation model to obtain a feature map and a segmentation map for subsequent processing; then integrating an attention mechanism into a feature fusion process to obtain a fused convolution feature layer; inputting the fused convolution feature layer into a candidate region generation network, and obtaining a candidate region by considering the position relationship between the target and the blood vessel; and finally, further classifying and regressing the candidate regions to obtain a detection result.

A method of segmenting tubular structures in digital images includes selecting a point in an image of a tubular object to be segmented, defining an initial state of the selected point, initializing measurement weights, a conditional probability distribution and a prior probability distribution of a feature space of the initial state, sampling the feature space from the prior probability distribution, estimating a posterior probability distribution by summing sample measurements weighted by the measurement weights, and segmenting a cross section of the tubular object from the posterior probability distribution.

The invention discloses a method for fine segmentation of the eye ground optic disc based on SLIC super-pixel segmentation. The steps are as follows: super-pixel segmentation, vascular segmentation based on morphological processing and R-G double-channel color threshold segmentation are carried out on an input eye ground image, and after expansion of a connected region obtained from color threshold segmentation, a corresponding Toeplitz matrix template is selected according to the pixel coordinates of the connected region to filter an eye ground vascular image to obtain the center position ofthe optic disc; and then, a candidate area of the optic disc is extracted and the internal blood vessels are removed, threshold segmentation is carried out on the candidate area of the optic disc through a binarization method, the ROI (Region of Interest) of the optic disc ellipse is determined by using an ellipse fitting method based on least squares, super pixels with a certain overlapping areaare retained according to the SLIC super-pixel segmentation result, and thus, fine segmentation of the optic disc is completed. Through the method, automatic positioning and fine segmentation of the optic disc are realized, better contour information of the optic disc can be retained, less time is consumed, other follow-up processing of the eye ground image is facilitated, and assistant diagnosiscan be provided for ophthalmologists.

The invention discloses a vascular segmentation method and device and a computer device. The vascular segmentation method includes: acquiring perfusion image data of a selected object; performing vascular reinforcement on the perfusion image data on the basis of change information of concentration of a contrast in a vessel and / or vascular space tubular structural information to obtain vascular reinforcement image data; using a preset threshold to perform image segmentation on the vascular reinforcement image data so as to extract macro-vascular image data and micro-vascular image data; fusingthe macro-vascular image data and the micro-vascular image data to obtain a vascular image mask. Vascular reinforcement is performed through time and space information, such as changes in the contrastin the vessels and tubular structure of the vessels so as to extract macro vessels and micro vessels, the influence of non-rigid deformation portions, such as nasal cavity, upon vascular segmentationis effectively avoided, richer vascular structure can be extracted for clinical diagnosis and reference, and vascular segmentation accuracy is greatly improved.

The embodiment of the invention discloses a premature infant retinopathy plus lesion classification method. The method comprises: constructing a blood vessel segmentation model capable of segmenting ablood vessel image from the fundus image; and constructing a classification model capable of carrying out plus lesion classification on the vascular image, segmenting the vascular image in the targetfundus image from the target fundus image by applying the vascular segmentation model, and classifying the vascular image in the target fundus image by applying the classification model to obtain a plus lesion category to which the vascular image in the target fundus image belongs. Therefore, blood vessel segmentation and blood vessel image classification are carried out on the fundus image basedon the blood vessel segmentation model and the classification model, and compared with an existing mode of manually classifying the plus lesions, the classification efficiency of the premature infantretinopathy plus lesions can be improved.

A method of segmenting tubular structures in digital images comprises providing a digitized image, selecting a point within an object for segmenting in the image, defining an initial state of the selected point in the object, performing an initial segmentation of a 2D cross section of the object based on the initial state, predicting a new state of said object about a new point that is a translation of said selected point along the object tangent, correcting said new state prediction based on a measurement of said new point in said image, and segmenting a 2D cross section of said object based on said new state.

The present application relates to a cerebral vascular segmentation method, a system and an electronic device. The method comprises the following steps: the original blood vessel image data is processed by multi-scale filtering and enhancement to obtain the enhanced blood vessel image data and the corresponding direction vector field; the finite mixing model is established, and the parameters of the finite mixing model are estimated, and the class conditional probability P (y | x) is obtained; an initial marker field of the blood vessel image data is calculated and the initial marker field iscombined with a corresponding direction vector field to form a new Markov random field; according to the equivalence of Markov random field and Gibbs distribution, a priori-like probability P (x) is obtained; based on the conditional probability P (y | x) and the priori-like probability P (x), the blood vessel segmentation results of the blood vessel image data are obtained through the maximum posterior probability and the conditional iterative mode. The present application can extract more effective blood vessel candidate space, extract blood vessel structure under low contrast, and make theblood vessel network more complete.

Methods, systems, computer programs, circuits and workstations are configured to generate at least one two-dimensional weighted CBF territory map of color-coded source artery locations using an automated vascular segmentation process to identify source locations using mutual connectivity in both image and label space.

Methods, systems, computer programs, circuits and workstations are configured to generate at least one two-dimensional weighted CBF territory map of color-coded source artery locations using an automated vascular segmentation process to identify source locations using mutual connectivity in both image and label space.

An X-ray image processing device for providing segmentation information with reduced X-ray dose that includes an interface unit, and a data processing unit. The interface unit is configured to provide a sequence of time series angiographic 2D images of a vascular structure obtained after a contrast agent injection. The data processing unit is configured to determine an arrival time index of a predetermined characteristic related to the contrast agent injection for each of a plurality of determined pixels along the time series, and to compute a connectivity index for each of the plurality of the determined pixels based on the arrival time index. The data processing unit is configured to generate and provide segmentation data of the vascular structure from the plurality of the determined pixels, wherein the segmentation data is based on the connectivity index of the pixels.

The invention belongs to the technical field of medical image processing, and discloses a blood vessel image segmentation method based on centerline extraction, a nuclear magnetic resonance imaging system, and preprocessing of brain blood vessel data by vesselness filtering based on Hessian matrix; topology refinement method for blood vessel centerline Extraction; take the centerline point as a positive sample and the non-vascular point as a negative sample to extract the features of the training sample and the test sample; use the features of the training sample and the corresponding label to train the SVM model, and use the feature of the test sample as the input of the trained SVM model , the output label is the segmentation result of blood vessels. The invention reduces the workload and improves the calculation efficiency; it does not need to manually calibrate the target and the background, completes automatic blood vessel segmentation, and greatly improves the segmentation efficiency. The invention realizes the segmentation of cerebral blood vessels, which is accurate, fast and does not require human intervention; the true positive rate and true negative rate can reach 0.85.

The invention discloses a method and a system for establishing a loss function for liver blood vessel segmentation, and belongs to the technical field of medical image processing and artificial intelligence. The method comprises the following steps: performing morphological corrosion on an original liver blood vessel mask, and removing free miscellaneous points to obtain a second mask; performing morphological expansion on the second mask to obtain a third mask; calculating an intersection of pixel positions, which are equal to 1, in the original mask and the third mask to obtain a fourth mask; calculating a difference set of pixel positions which are equal to 1 in the original mask and the fourth mask to obtain a fifth mask; dividing an original liver blood vessel mask pixel position set into three parts, and establishing a loss function of liver blood vessel segmentation; the system comprises a corrosion module, a removal module, an expansion module, a first calculation module, a second calculation module and an establishment module. The loss function established by the invention can perform different weighting on branch blood vessels, main blood vessels and background pixels, and the optimization direction of the network is more flexibly and accurately controlled.

The present application provides a method and device for training a blood vessel segmentation model, and an application method and device. The training method of the blood vessel segmentation model includes: extracting the blood vessel midline in the blood vessel sample image; determining the loss weight of each pixel point according to the distance between each pixel point in the blood vessel sample image and the blood vessel midline; The weight, the blood vessel segmentation label corresponding to the blood vessel sample image, and the blood vessel segmentation result output by the blood vessel segmentation model, determine the first loss function value; the blood vessel segmentation model is trained according to the first loss function value, which can solve the category imbalance of different parts of the blood vessels. problem, improve the training effect and segmentation effect.

The invention discloses a blood vessel segmentation method, equipment and computer storage medium. The method includes: acquiring and segmenting blood vessel image data, obtaining the prediction result of the aorta model and the prediction result of the small blood vessel model; judging the prediction result of the aorta model and Whether the prediction results of the small blood vessel model can be connected as a whole; when the prediction results of the aorta model and the prediction results of the small blood vessel model cannot be connected as a whole, perform coronal segmentation prediction on the blood vessel image data to obtain A prediction result of the coronary ostium model; integrating the prediction result of the aorta model and the prediction result of the small vessel model according to the prediction result of the coronary ostium model to obtain the prediction result of the normal coronary artery model. The invention makes the segmentation of the coronary ostium area more accurate, thereby overcoming the inaccurate or even abnormal segmentation of the joint between the coronary artery and the aorta in the traditional method, thereby obtaining a more complete and accurate coronary artery segmentation result.

The embodiment of the present invention relates to a method for segmenting blood vessels in a fundus image. Each blood vessel feature processing layer sends an up-sampled image obtained by this layer to a connected blood vessel feature optimization layer; the lowest blood vessel feature optimization layer also acquires the lowest blood vessel feature. The feature image of the extraction layer; the blood vessel feature processing layer of the highest layer sends the up-sampled image obtained by this layer to the blood vessel feature optimization layer of the lowest layer through the backward short connection; each blood vessel feature optimization layer sequentially acquires each image. Performing blood vessel feature extraction and non-linear processing to obtain a non-linear image corresponding to each image; each blood vessel feature optimization layer sends each acquired image to a higher blood vessel feature optimization layer through forward short connections. In the embodiment of the present invention, the high-level information is sent to the low-level through the backward short connection, and the low-level information is sent to the high-level through the forward short connection, so that the features of all levels are fully integrated, and the blood vessel segmentation is more accurate.

The invention discloses a retinal artery vein diameter ratio automatic measuring method and a device. The retinal artery vein diameter ratio automatic measuring device can automatically calculate thecorresponding AVR indexes, increase the accuracy and automation degree of AVR calculation, realize accurate measurement of the AVR indexes, and accelerate clinical disease diagnosis and treatment speed. The method comprises the following steps: (1) acquiring a retina image by using fundus imaging equipment; (2) carrying out optic disc positioning and blood vessel segmentation; (3) based on the center line, carrying out blood vessel topology analysis; and (4) based on a deep convolutional network, classifying the blood vessel tree into arteries or veins, calculating the diameter and category ofsampling point positions, and completing the measurement of the retinal artery-vein diameter ratio AVR.

The present application relates to a retinal vessel segmentation method, device and storage medium based on vessel features. Including obtaining the dataset of retinal vessel segmentation, and performing data enhancement operations on the dataset; establishing a deep convolutional neural network intelligent model based on multi-scale semantic information fusion, inputting the enhanced data into the intelligent model, and obtaining the output of the model; Use the image morphology method to detect blood vessel breakpoints and areas with inconsistent thickness in the output results; increase the weight of the loss function on the breakpoints and areas with inconsistent thickness, and iterate and optimize the intelligent model through the cross-entropy loss function of dynamic weights to achieve Accurate segmentation of retinal vessels. This method combines multi-scale semantic information and breakpoint information to solve vascular connectivity, and improves the loss function by extracting vascular thickness information to solve the problem of inconsistency in vascular thickness, effectively improving the accuracy of retinal vascular segmentation, which is of great significance for computer medical intelligent diagnosis .

The invention discloses a fundus image blood vessel segmentation method based on phase congruency and mainly overcomes the defect that a traditional method can not be used to accurately segment blood vessels in fundus images. The fundus image vascular segmentation method base on the phase congruency can be simultaneously used to segment small blood vessels of most tips. The method comprises the steps: (1) extracting green channels of the fundus images, (2) enhancing the contrast ratio of the images through contrast limited adaptive histogram equalization (CLAHE), (3) filtering the fundus images through the anisotropic coupled diffusion equation, (4) segmenting the blood vessels of the fundus images filtered or not filtered through the anisotropic coupled diffusion equation in a phase congruency algorithm, (5) multiplying pixel-levels of results of vessels, of two fundus images, segmented based on the phase congruency algorithm, (6) processing the images in a binaryzation mode through the iterative threshold segmentation method, (7) optimizing the images in the mathematical morphology method. The fundus image vascular segmentation method has significant application values in fields of three-dimensional splicing of the fundus images and judging existence of diabetes mellitus and severity of diabetes mellitus.

The embodiment of the present invention discloses a method for segmenting blood vessels in a fundus image based on a layered matting algorithm. The method includes: preprocessing the fundus image to generate a tripartite image of the fundus image; segmenting the fundus image using a layered matting algorithm Blood vessels in unknown regions in the three-part image; post-processing the segmented blood vessel images; testing the post-processed images on two public databases, DRIVE and STARE, to obtain the result map of blood vessel segmentation in fundus images. The present invention uses a layered matting algorithm to process the fundus image, which can efficiently and accurately segment the blood vessels of the fundus image, thereby helping doctors improve the efficiency of diagnosing eye diseases, and helping to reduce possible errors caused by doctors. Misdiagnosis due to fatigue.

The invention relates to an image display method, computer equipment and a storage medium. The method comprises the following steps: acquiring medical images of a 2D sequence and a 3D sequence of the same vascular structure of the same patient, performing vascular segmentation on the medical images of the 2D sequence to obtain a first vascular cross section, and registering the first vascular cross section and the vascular structure in the medical images of the 3D sequence to obtain a registered medical image. According to the method, the first blood vessel cross section of the 2D sequence and the blood vessel structure of the 3D sequence are displayed on the same display interface at the same time, the whole blood vessel structure is displayed from the 3D dimension, the blood vessel cross section is also displayed from the 2D level, the overall form and local information of the blood vessel structure are included, and the display effect is good. And sufficient and reliable reference data is provided for a doctor to carry out plaque analysis based on the images of the 3D and 2D sequences displayed on the display interface, so that the accuracy of plaque analysis is improved.

The invention discloses a method and system for segmenting retinal blood vessels based on retinal fundus images, belonging to the technical field of image processing, including: acquiring a retinal fundus image to be detected; constructing a basic module according to the characteristics of the retinal fundus image; cascading N basic modules as In the final network model, the retinal fundus image to be detected is used as the input of the overall network model to obtain the segmentation result of retinal blood vessels. By passing the foreground features of the previous basic module and the original picture into the next basic module, the latter basic module can inherit the learning experience of the previous basic module, thus speeding up the training process and effectively solving the problem of data imbalance. Problem: Using the retinal fundus image to be detected as the input of the overall model S-UNet, the segmentation result of retinal blood vessels is more accurate.

The present application discloses a method for cerebrovascular segmentation based on a multi-view cascaded deep learning network. The method includes acquiring MRA images and MRV images corresponding to the target part; determining several slice images based on the MRA images and MRV images; The trained first segmentation network model determines the reference segmentation map corresponding to each slice image; based on the MRA image, the MRV image, the obtained reference segmentation map and the trained second segmentation network model, determine the MRA The image corresponds to the segmented image. In this application, the first network segmentation model acquires the reference segmentation images corresponding to each section image as context information together with the MRA image and the MRV image as the input information of the second segmentation network model, so that the blood vessels in the MRA image can be learned from multiple perspectives Information, so that the accuracy of the second segmentation network model to determine the segmented image can be improved, thereby improving the accuracy of cerebrovascular segmentation.

The invention relates to a tumor area capillary segmentation method and device. The method comprises the following steps: acquiring an image block containing a tumor area and microvessels; and inputting the image block into a tumor area microvessel segmentation model to obtain a prediction probability graph of the tumor area and the microvessels, wherein the prediction probability graph comprises a capillary region, a tumor region and a background region, and the tumor area capillary segmentation model comprises two convolution layers with preset scales, a plurality of dense connection blocks and a sigmoid activation module. By adopting the method, the accuracy of tumor area capillary segmentation can be improved.

The invention discloses a fundus image blood vessel segmentation method based on a shared decoder and a residual tower structure. The method comprises the following steps: obtaining an image block of a training data set and an image block of a test data set through a data input module; The tower module obtains the residual tower sequence; the multi-level semantic features are obtained through the encoding module; the multi-level probability map is obtained through the shared decoding module; the multi-scale label, the residual tower sequence, and the probability map obtained by the shared decoder are constructed The total loss of the model is formed, and PyTorch is used for gradient optimization to train the parameters in the encoding module and the shared decoding module; the image blocks of the test data set are sequentially input into the trained encoding module and the shared decoding module to obtain a probability map, and the obtained probability map And stitching, binarization processing to get the final segmentation results. The invention solves the problems of uneven distribution of blood vessel calibers and weak contrast of fundus images.

The invention discloses a wrapping-degree determination method and device for a the tumor-wrapping vessel and relates to the technical field of biomedicine with a main aim to be capable of ensuring accuracy in wrapping-degree calculation of the tumor-wrapping vessel. The method includes: acquiring vascular segmentation images which are acquired by subjecting the tumor-wrapping vessel to vascular segmentation; identifying vascular contours of the vessel according to vascular points segmented according to the vascular segmentation images; determining points that the vessel is adhered to the tumor by traversing the points in the vascular contours and two points in the vascular contours in a manner of neighboring back and forth; determining contour length of the vascular contours by traversing the points in the vascular contours and determining wrapping length of the tumor-wrapping tumor by traversing the points where the vessel is adhered to the tumor; according to the contour length and wrapping length, calculating the wrapping degree of the tumor-wrapping vessel. By the arrangement, wrapping degree of the tumor-wrapping tumor is determined.

A blood vessel segmentation method for fundus images based on multi-scale features of fully convolutional neural networks, comprising the following steps: 1) Preprocessing the retinal images of the fundus; 2) Segmenting the preprocessed images into image blocks for data expansion; 3) Construct a convolutional neural network model, and use the expanded data for network training; 4) Test the trained model to obtain segmentation results. In the present invention, by connecting one encoding and two different decoding structures, and adopting multiple skip connections, it can overcome the shortcomings of the small number of blood vessel image data sets and the low segmentation accuracy caused by low image quality, and more fully integrate images of different depths. Features, and effectively alleviate the gradient disappearance problem caused by the increase of network depth, compared with traditional segmentation methods, it has higher accuracy and higher robustness.