58 results about "Retinal blood vessels" patented technology

The invention discloses a retinal blood vessel segmentation method based on the deep-learning adaptive weight, comprising: performing sample expansion on retinal blood vessel images and grouping the samples; constructing a full-convolution neural network for blood vessel segmentation; pre-training the network through the training samples; performing the global adaptive weight segmentation on the retinal blood vessel images; obtaining the initial model parameters for retinal blood vessel segmentation; adding a conditional random field layer to the last network layer and optimizing the network; using the rotation test method to input the testing samples to the network; and obtaining the retinal blood vessel segmentation result. The blood vessel segmentation full-convolution neural network structure and the adaptive weight method proposed by the invention can realize the human-eye level image segmentation and can test on two internationally published retinal image databases DRIVE and CHASE_DB1 with the average accuracy of 96.00 % and 95.17% respectively, both higher than the latest algorithm.

The invention provides a fundus retinal blood vessel segmentation method and system based on K-Means clustering annotation and naive Bayesian model. The method of the invention comprises the followingsteps: randomly extracting color fundus images in a data set to construct a training set and a test set; extracting the gray image of G channel of the color fundus image as the object of feature extraction; carrying out feature extraction on the gray image, and representing each pixel in the image by a multi-dimensional feature vector; for each image in the training set after feature extraction,using K-Means clustering algorithm to label eigenvectors by clustering; training a naive Bayesian model based on training set data labelled based on K-Means clustering; segmenting the blood vessels ofeach image in the test set using the trained naive Bayesian model. The invention regards the result of clustering as the label with supervised training, and trains the naive Bayesian classification model for retinal blood vessel segmentation by using the label. The whole process does not require human to participate in the label, saves time and labor, and greatly improves the processing efficiency of the machine learning model.

The invention provides an automatic retinal blood vessel segmentation method for clinical diagnosis of glaucoma. Through the method, the influence of bright regions such as optic disc and exudate is eliminated by fusing five models depending on different image processing technologies, namely a matched filter, a neural network, multi-scale line detection, scale space analysis and morphology. At thesame time, massive data is not needed to establish a retinal blood vessel segmentation model, so the method greatly reduces the amount and complexity of data to be processed, is easy to realize, andcan effectively improve the efficiency of retinal blood vessel segmentation. The method also utilizes the region growth method and the gradient information to carry out iterative growth on the background and the blood vessel region on the basis of a multimodal fusion result, the segmentation results exhibit better continuity and smoothness, more retinal vascular details and more complete retinal vascular network can be kept, and thus the method effectively assists ophthalmologists in diagnosing diseases and lightens the burden of ophthalmologists.

An image processing apparatus includes an identification unit configured to identify a retinal blood vessel based on a retinal image, a measurement unit configured to measure blood flow information for the blood vessel based on the retinal image, and a display control unit configured to display the measured blood flow information by at least one selected from a depth of the identified blood vessel, a size of the identified blood vessel, and a combination of both.

The invention discloses a registration method of a three-dimensional non-rigid optical coherence tomographic image. The method comprises: firstly, carrying out pretreatment on an OCT scanning image; to be specific, carrying out OCT denoising, image layering and image projection on an the OCT scanning image successively to obtain a two-dimensional projection image of a retian blood vessel; extracting a blood vessel from the two-dimensional projection image of the retinal blood vessel to obtain two-dimensional feature point of the blood vessel, and returning to three-dimensional space based on the obtained two-dimensional feature point to obtain a three-dimensional feature point; and then on the basis of the obtained three-dimensional feature point, carrying out rough image registration by using affine transformation, and then carrying out precise image registration by using a non-rigid method to obtain a precisely registered image. According to the method disclosed by the invention, on the basis of combination of the grayscale-based non-rigid registration method and feature-affine-transformation-based registration method, a three-dimensional OCT scanning image is registered, so that the precision and reliability of the registration result can be improved.

The invention discloses an eye fundus image retinal vessel segmentation method based on a mixed attention mechanism, and the method comprises the following steps: S1, obtaining a retinal eye fundus image, and dividing the retinal image into a training set and a test set; S2, constructing a hybrid attention convolutional neural network, wherein the hybrid attention convolutional neural network is used for segmenting retinal vessels in the retinal fundus image; S3, training the hybrid attention convolutional neural network by using a training set, and testing the hybrid attention convolutional neural network by using a test set to obtain a trained hybrid attention convolutional neural network; and S4, inputting a to-be-segmented retinal image into the trained mixed attention convolutional neural network, wherein the mixed attention convolutional neural network outputs a retinal image blood vessel segmentation result. According to the method, a low-contrast vascular structure is effectively and accurately segmented, and the method has high robustness for interference of complex eye fundus image focuses, blood vessel center reflection phenomena and illumination imbalance phenomena.

Provided is a treatment method and a medical device that can intentionally and locally administer a drug to a lesion area appearing in a narrow blood vessel such as a retinal blood vessel and a spinal blood vessel. The treatment method has an introduction step of introducing a medical elongated body having a drug holder for holding a drug into a living body, an arrangement step of arranging the drug holder at a treatment target inside the living body, and a discharge step of discharging the drug to the treatment target by releasing the drug held by the drug holder.

The invention discloses a GAN-based retinal vessel image intelligent segmentation method, and the method comprises the following steps: 1, giving a retinal image set, and dividing the set into a training set and a test set; 2, designing a generator network G and a discriminator network D, and constructing an Adam optimizer; 3, inputting the training set into G; 4, generating a blood vessel segmentation image through the G; 5, discriminating and calculating the segmented image generated by the G through the D; 6, updating G and D parameters; 7, evaluating the G and obtaining an optimal model G', and repeating the steps 3-7 until iteration is finished; and 8, inputting the retina image into the G' to generate a blood vessel segmentation image. According to the invention, a large receptive field network model is used to carry out intelligent segmentation on a retinal image to obtain a final retinal vessel segmentation image. The network model has good robustness, and the obtained blood vessel segmentation image contains less noise and is superior to an existing retinal blood vessel image segmentation method on the whole.

The invention discloses a retinal vessel segmentation method based on symmetric bidirectional cascade network deep learning, and belongs to the field of medical image processing. According to the method, data enhancement is carried out through a series of modes such as contrast change, rotation, zooming and translation, data set amplification is realized, and then a preprocessed image is input into a bidirectional cascade network for training and learning to obtain a predicted retinal vessel segmentation result. The network is composed of five scale detection modules, retinal vessel characteristics of different diameter scales are extracted by changing the expansion rate, two vessel contour prediction images are generated in the two directions from the lower layer to the upper layer and from the upper layer to the lower layer of the network respectively, and the two vessel contour prediction images are structurally distributed in an up-and-down symmetrical mode; the outputs of the twopaths of the dense hole convolution module are combined; and finally, the blood vessels and the background pixels are classified by adopting a quasi-balanced cross entropy loss function so as to realize accurate segmentation of the retinal blood vessels.

The invention belongs to the technical field of medical image processing, and discloses a fundus image retina blood vessel diameter automatic quantification method. The method comprises the followingsteps of using a two-step positioning method combining coarse positioning and fine positioning to automatically detect the visual disc; preprocessing the fundus image, including denoising and enhancement processing of the fundus image; separating the retinal blood vessel from the background of the fundus image by adopting a multi-scale analysis method, and reconstructing and eliminating pseudo blood vessels in segmentation by utilizing mathematical morphology; and measuring the diameter of the blood vessel by using a boundary method, and displaying a result. On the basis of fundus image features, the fundus image feature extraction method firstly detects the optic disc, then enhances the image contrast ratio, segments the retinal blood vessel, finally measures the size of the retinal bloodvessel diameter, displays the result on the system interface, provides reference for early prediction and auxiliary diagnosis of clinicians, saves manpower and improves the efficiency.

The invention discloses a retinal vessel segmentation method based on regional growth PCNN. The retinal vessel segmentation method comprises the following steps: selecting seed points from unmarked pixels of a target retinal vessel image; Increasing the connection strength of the PCNN model, and extracting blood vessel characteristics in the target retina blood vessel image by using the PCNN modelwith the increased connection strength and taking the seed point as a starting point until the increased connection strength is greater than a first preset threshold value; If the blood vessel characteristics extracted through iteration at this time do not meet the first preset condition and the second preset condition at the same time, marking pixels corresponding to the blood vessel characteristics extracted through iteration at this time with the same label until all pixels are marked with labels, Wherein the first preset condition is that the proportion of the number of the blood vessel edge pixels to the total number of the blood vessel pixels is smaller than or equal to a second preset threshold value, and the second preset condition is that the ratio of the blood vessel area to thearea of the whole image is smaller than or equal to a third preset threshold. Automatic growth of the blood vessel area is achieved, and the retinal blood vessel segmentation precision is improved.

The invention discloses a fundus image blood vessel segmentation method based on skeleton prior and contrast loss. The fundus image blood vessel segmentation method comprises the following steps: S1, performing data augmentation on a color fundus image; s2, performing expert annotation on the eye fundus image to extract a skeleton; s3, inputting the eye fundus image into a segmentation network, and calculating segmentation loss; s4, the foreground and background features of the middle features are compared to learn loss; s5, outputting a skeleton continuity constraint for the segmentation model, and solving a loss function; s6, superposing the three loss functions to obtain total loss, carrying out gradient back propagation, and stopping training when the total loss is not reduced any more for four consecutive rounds; and S7, obtaining a binary vascular tree segmentation result. Compared with the prior art, the contrast loss function adopted in the two types of pixel feature sample sets can further improve the discrimination capability of the model for hidden layer features in a high-dimensional space, can extract small blood vessels and prevent the blood vessels from being broken, can inhibit the interference of biomarkers, and is very suitable for fine retinal vessel tree segmentation.

Isolated, mammalian, mature bone marrow-derived, lineage-negative hematopoietic stem cell populations (Lin-HSCs) containing endothelial progenitor cells (EPCs) are capable of repairing retinal vascular and neural networks in the eye. Preferably, at least about 20% of the cells in the isolated Lin-HSCs express the cell surface antigen CD31. The isolated Lin-HSC population has utility for the treatment of ocular vascular diseases. In a preferred embodiment, the Lin-HSC population is obtained by: extracting the bone marrow of an adult mammal, isolating multiple monocytes from the bone marrow, using biotin-conjugated lineage populations directed against one or more lineage surface antigens These monocytes were labeled with antibodies; monocytes positive for lineage surface antigens were removed from various monocytes, and then the Lin-HSCs population containing EPCs were recovered for isolation. Also provided are isolated Lin-HSCs transfected with a therapeutically useful gene for gene delivery into the eye in cell-based gene therapy. Also described are methods of making the isolated stem cell populations of the invention, and methods of treating eye diseases and eye injuries.

The invention relates to the field of medical optical detection and discloses a human eye retina blood flow measuring device based on a light beam parallel scanning mode. Blood vessel branches of different parts of the retina of the human eye and the total blood flow of the retina can be quantitatively detected, so that a detection light beam passes through the front focus of the human eye, and then the retina of the human eye is scanned in a manner of being parallel to the main optical axis or the radial axis of the human eye; the light beam parallel incidence mode can effectively increase the included angle between the detection light beam and the retinal vessel, so that the quantitative detection of the blood flow of vessel branches at different parts of the retina of the human eye is realized, and the accuracy of blood flow measurement results is improved. Meanwhile, a single-beam OCT system is used for measuring the blood flow of the retina of the human eye, so that the complexityof the OCT retina blood flow measuring device is reduced, the cost is greatly reduced, popularization and application of the retina blood flow measuring technology in medical clinic are promoted, andassistance is offered for diagnosis and research of ophthalmic diseases.

The invention discloses a multispectral fundus image analysis method and system based on adversarial learning, and the method comprises the steps: obtaining a multispectral fundus image which comprises a blood vessel label image and a blood vessel label-free image; inputting the multispectral fundus image into a trained fundus image analysis model to obtain a registration result of the multispectral fundus image; wherein the eye fundus image analysis model comprises an eye fundus image registration model and a retinal blood vessel segmentation model, and during training, the eye fundus image registration model and the retinal blood vessel segmentation model are independently trained by adopting blood vessel label images respectively; and the fundus image registration model and the retinalvessel segmentation model perform adversarial learning according to the vessel-free label image. The fundus image registration model and the retinal blood vessel segmentation model can perform adversarial learning according to the image without the blood vessel label besides performing independent training through the image with the blood vessel label, so that the precision of image registration and blood vessel segmentation is improved.

The invention discloses an OCT blood vessel image dislocation correction method. The method comprises the steps of preprocessing blood vessels; performing shift correction on an image based on a Fourier formula and a Euclidean distance transformation formula; and removing noise through a coherent denoising method. Dislocation correction is carried out on a blood vessel image, related noise is suppressed, the quality of retinal blood vessel imaging data is improved, and the correction algorithm is high in efficiency, excellent in image quality, high in robustness and high in efficiency; the method can be used for blood vessel image dislocation correction.

The invention provides a retinal blood vessel segmentation method and device. The retinal blood vessel segmentation method is a color retinal image automatic segmentation method based on a deep convolutional neural network. The method comprises the following steps of: firstly, preprocessing a retina image, and processing a color image into a grey-scale map with higher contrast after a G channel, histogram equalization and gamma transformation; secondly, randomly segmenting the processed image into uniform small blocks to form a training data set; and then sending the data set into a deep convolutional neural network combining void space pyramid pooling and an efficient fusion attention mechanism, and carrying out training of a retinal vessel segmentation model; and finally, adjusting model parameters through a cross entropy loss function fused with the cost-sensitive matrix, and segmenting the blood vessel in the color retina image by using the optimized model. According to the invention, high segmentation accuracy and speed can be realized, and the manpower cost of doctors is reduced.

The invention provides a multi-low-level feature fusion retinal vessel segmentation method, which comprises the following steps of: S1, acquiring an original fundus image set, and respectively extracting a plurality of low-level feature images; s2, stacking a plurality of low-level feature images together to form a feature vector containing background and blood vessel features; s3, acquiring a manually labeled eye fundus image set, and taking the eye fundus image set and the feature vectors as input for training a retinal vessel segmentation model to complete model training; and S4, processing a fundus image to be segmented through the steps S1 and S2, and inputting the fundus image to be segmented into a trained retinal blood vessel segmentation model to obtain a segmented image only containing blood vessels and backgrounds. According to the method, multiple low-level features of the retinal blood vessel image features are fully considered, rich retinal blood vessel features can be effectively reserved, and the classification accuracy is high. The method does not depend on a large amount of sample data training, and solves the problems of few learning samples and poor effect of a deep learning method caused by the actual condition of small samples in the fundus image.

The invention provides a retinal blood vessel segmentation method and device, electronic equipment and a storage medium. The method comprises the following steps: acquiring eye fundus images, dividing the eye fundus images into a training set and a test set, and performing corresponding preprocessing operation on the eye fundus images in the training set and the test set; respectively constructing a segmentation network and a discrimination network; inputting the marked fundus images in the training set into a segmentation network for training, inputting the unmarked fundus images in the training set into the segmentation network after training for a preset round, and alternately training the segmentation network and the discrimination network to obtain a trained retinal vessel segmentation model; inputting a fundus image to be segmented into the retinal vessel segmentation model to obtain a segmented output image; and splicing all the output images to obtain a retinal blood vessel segmentation result image. The retinal blood vessel in the fundus image can be automatically and accurately extracted, the segmentation result comprises tiny details of the blood vessel, and detail information of the image is richer and can be used for clinical auxiliary diagnosis.

The invention provides a psychological state recognition method and equipment based on an fundus image. The method comprises the following steps: acquiring a retinal blood vessel diameter parameter and the fundus image; extracting feature information from the fundus image by using a feature extraction network; obtaining fused feature information according to the feature information and the parameters; and identifying the fused feature information by using an output network to obtain an identification result about the psychological state.

The invention discloses a fundus blood vessel three-dimensional reconstruction method and device, equipment and a storage medium, and the method comprises the steps: carrying out the depth prediction of OCTA two-dimensional blood vessel front projection images of different equipment through a cross-domain depth adaptive network, and obtaining a depth image prediction result; extracting the center line position and the radius length of the blood vessel on the OCTA two-dimensional blood vessel segmentation image; obtaining a three-dimensional blood vessel center line point cloud according to depth information carried by a depth map prediction result and extraction information of the OCTA two-dimensional blood vessel segmentation image; and reconstructing the three-dimensional surface grid of the blood vessel according to the three-dimensional blood vessel center line point cloud and the space parameter equation of the circle. According to the retina blood vessel reconstruction process from two dimensions to three dimensions, the problem of blood vessel space information loss in a two-dimensional front projection image, the common artifact problem in an OCTA image and the problem that blood vessels cannot be reconstructed by using three-dimensional volume data due to three-dimensional data loss and limitation of technical means are effectively solved.

The invention provides a retinal vessel segmentation method based on an improved U-Net network. Image enhancement is performed on a color eye fundus image, so that the contrast ratio between a blood vessel and a background in the image is improved, and a training data set is amplified. A U-Net encoder-decoder structure is used as a basic segmentation framework, a dense convolution block and a CDBR layer structure are designed to replace a traditional convolution block, learning of multi-scale feature information is achieved, and the feature extraction capacity of the model is improved. Meanwhile, an attention mechanism is introduced at a jump connection part of the model, so that the model is enabled to allocate weights again, the importance degree of a feature channel is adjusted, noise is suppressed, the problem of blood vessel information loss in an up-sampling process at a decoder end is solved, and a GAB-D2BUNet network model is constructed based on the above technologies. According to the method, an internationally disclosed retina fundus blood vessel data set DRIVE is adopted for training, and finally the optimal segmentation model is reserved to verify the segmentation performance of the model. The retina fundus blood vessel segmentation method achieves the task of accurately segmenting the retina fundus blood vessel, and has better segmentation performance.

The invention provides a U-Net fundus retinal blood vessel image segmentation method based on improvement. The method comprises the following steps: acquiring a to-be-detected retinal blood vessel image; inputting the retinal blood vessel image into a trained improved U-Net network model for image segmentation; and outputting a retinal blood vessel image segmentation result. A common convolution block of the U-Net network is changed into a random discarded convolution block, so that image related features can be better extracted, and network over-fitting is effectively relieved; then, a NoL-block attention module is added to the bottom of a coding and decoding structure, and on the basis of not increasing the parameter quantity, the receptive field is expanded, and the correlation of pixel information is enhanced; the NoL-block is used as a component and is embedded into the U-Net network structure, so that the NoL-block can be used for directly capturing the spatial dependency relationship between pixels with distances in image segmentation in a forward propagation manner, and the blood vessel can be better positioned and segmented.