33 results about "Retina vessels" patented technology

A fundus camera using infrared light sources, which included an imaging optical path and an optical path for focusing and positioning and two optical paths share a common set of retina objective lens, and a computer-assisted method for retinal vessel segmentation in general and diabetic retinopathy (DR) image segmentation in particular. The method is primarily based on Multiscale Production of the Matched Filter (MPMF) scheme, which together the fundus camera, is useful for non-invasive diabetic retinopathy detection and monitoring.

The invention discloses a retinal vessel image segmentation method based on deep learning, and the method comprises the steps: carrying out the enhancement of a fundus image, amplifying the data of atraining set, constructing a dense connection convolution block, replacing a conventional convolution block with the dense connection convolution block, achieving the feature reuse, and improving thefeature extraction capability; constructing an attention mechanism module, and performing adaptive adjustment on the feature map to highlight important features so as to suppress invalid features; building a model, building a DA-Unet network, using the processed data set to perform training and parameter adjustment, and obtaining and storing an optimal segmentation model; and carrying out actual segmentation, segmenting the eye fundus image needing retinal vessel segmentation into 48 * 48 sub-block images by using a sliding window, inputting the 48 * 48 sub-block images into a DA-Uet network for segmentation, outputting segmented sub-block image results, and splicing the segmented small block images into a complete retinal vessel segmentation image. The blood vessel segmentation method canautomatically segment blood vessels and has a good segmentation effect on tiny blood vessels.

The invention discloses a blood vessel segmentation network and method based on a generative adversarial network. The segmentation network comprises a generation model and a discrimination model. Residual connection is added to the generation model on the basis of adopting a U-shaped encoding-decoding structure; the discrimination model adopts a full convolution form of a VGG network, and replacesthe convolution layer of the middle part with a dense connection module; according to the segmentation method, a generation sample of a color fundus image is generated through a generation model, thegeneration sample and a corresponding real sample are input into a discrimination model, alternate training optimization is performed on the generation model and the discrimination model, and finally, a to-be-segmented retinal vessel color image is input into the trained and optimized model, so that a vessel segmentation result can be output. According to the method, more tiny capillaries in theretinal vessel image can be detected, the vessel edge can be positioned more accurately, the vessel segmentation precision is greatly improved, and the vessel image segmentation sensitivity, effectiveness and stability are greatly improved.

The invention relates to a method and device for measuring blood oxygen saturation of a fundus retina. The method is characterized by comprising the following steps of: taking a platform based on an ASOLO (adaptive optics confocal scanning laser ophthalmoscope), selecting a light with at least two different wavelengths as the light source of the ASOLO, enabling the retina to be sequentially imaged after correcting a fundus aberration by adopting the adaptive optics; generating defocus via a distorting lens for realizing the longitudinal chromatography of the retina so as to make the same position of a retina vessel imaged; registering the obtained high-resolution retinal image with multiple wavelengths, extracting a plurality of darkest spots and the spot arranged at a fixed distance from the darkest spots and in a tissue along with the vessel; and processing data to obtain the blood oxygen saturation of the vessel. In the invention, the high-resolution retinal image can be obtained by adopting the adaptive optics to correct the fundus aberration, and the blood oxygen saturation of the arteries, veins and capillaries of the fundus retina can be measured by processing a multi-wavelength image.

The invention relates to application of a deep learning algorithm in the field of medical image analysis, in particular to a retinal vessel segmentation algorithm fusing W-net and a conditional generative adversarial network. According to the method, the problems of low segmentation sensitivity and insufficient micro blood vessel segmentation are well solved, great progress is made in the aspectsof the parameter utilization rate, the information circulation and the feature analysis ability of the network, complete segmentation of the main blood vessel and fine segmentation of the micro bloodvessel are facilitated, the blood vessel intersection is not prone to breakage, and a focus and an optic disc are not prone to being segmented into the blood vessel by mistake. According to the invention, a plurality of network models are fused under the condition of low complexity; the whole segmentation performance on a DRIVE data set is excellent, the sensitivity and accuracy of the method are87.18% and 96.95% respectively, the ROC curve value reaches 98.42%, the method can be used for computer-aided diagnosis in the medical field, and rapid and automatic retinal vessel segmentation is achieved.

The invention discloses a retinal vessel segmentation method combining U-Net and adaptive PCNN. The method comprises the steps of performing data augmentation on a fundus image database selected in anexperiment; graying processing is carried out on the data set pictures; carrying out CLAHE processing on the data set pictures to increase the contrast between retinal vessels and the background; partitioning the image; constructing and training a U-Net neural network model, and enhancing a picture; building a self-adaptive PCNN neural network model; and carrying out blood vessel segmentation byusing the adaptive PCNN. On one hand, the invention provides the fundus blood vessel image enhancement method based on the improved U-Net quadratic iteration, the background can be significantly inhibited, the blood vessel region is highlighted, the noise interference is weakened, and the picture contrast is increased, so that the picture quality of a data set is improved. The invention further provides a fundus blood vessel image segmentation method based on the self-adaptive PCNN. Accurate parameters are estimated by using an Otsu algorithm, then a U-Net secondary iteration enhancement output result is sent to an adaptive PCNN, and effective segmentation of a complete fundus vessel is realized.

The invention discloses a retina vessel segmentation system based on combination of a hessian matrix and region growing. The retina vessel segmentation system comprises a retina image preprocessing unit which is used for extracting a retina image green channel and enhancing the extracted image to improve the contrast; a hessian matrix enhancement unit which re-enhances the image through the hessian matrix and extracts the vessel direction in the retina image; a connected domain classification unit which morphologically classifies the image enhanced by the hessian matrix so as to extract smallvessels; and a region growing unit which performs region growing on the classified image to connect the broken vessel structures in the image so as to enhance the segmentation image and improve the segmentation result. Vessel segmentation of the retina fundus image can be realized by the system, and the algorithm of combination of the hessian matrix and region growing is put forward by using the mode of combining the hessian matrix and region growing for aiming at the problem of appearance of breaking points in the segmentation result of the segmentation algorithm so that the situation of vessel breaking in the extracted image can be further improved and the accuracy of vessel segmentation can be enhanced.

An existing fundus retina image segmentation technology has defects in retina vessel segmentation precision, and a good method for segmenting small capillaries does not exist. In order to realize accurate segmentation of a fundus retina image, the invention provides a fundus retina image segmentation method based on a generative adversarial network (C-GAN). The method comprises: a generation modelextracting abstract features through a coding-decoding network structure and carrying out detail recovery to obtain a blood vessel segmentation probability graph close to a sample true value; the discrimination network inputting the sample probability graph obtained by the generation model and a sample true value label into a deep convolutional network, giving a higher label to a real sample, andgiving a lower label to a generated sample. The network parameter weight is adjusted through mutual counterbalance of the generation model and the discrimination model, and the network converges thewhole network when the generation model and the discrimination model reach balance. According to the invention, tiny capillaries of the fundus image can be segmented, and the segmentation precision ofthe whole retinal vessel is improved.

The invention relates to the field of computer vision image processing, and discloses a retinal blood vessel image segmentation method and device. The method comprises the following steps: in a data processing stage, firstly, acquiring a to-be-detected retinal blood vessel image; then, performing data enhancement and preprocessing on the retina fundus image; in the training stage, firstly, a network model based on Transform optimization is constructed, and then network model training is carried out by using a processed training image; in a test stage, inputting the retinal blood vessel image into a trained network model for image segmentation; and finally, weighted average is carried out on prediction results of a plurality of retinal blood vessel images output by the network model to obtain the classification probability of each pixel, a final segmentation result graph is obtained, and the segmentation precision of the retinal blood vessel images is improved by adopting the retinal blood vessel image segmentation method.

The invention discloses a method for segmenting capillaries in a microscopic image based on a generative adversarial network. The objective of the invention is to solve the problems that a vascular fracture phenomenon often occurs in an algorithm segmentation result and vascular branch details are redundant due to the limitation of an algorithm or low actual imaging contrast of a microvascular image. The method comprises the steps of establishing a training model and a sample set based on a generative adversarial network; inputting the color fundus image in the sample set into a generation model, extracting image feature information, and outputting a microvascular probability image under a microscopic image as a generation sample; carrying out enhancement processing on the microscopic image comparison adaptive histogram equalization; increasing the data volume of training, and carrying out enhancement processing on the microscopic image processed by the preprocessing unit again; distinguishing the real sample from the generated sample; inputting a to-be-segmented retinal vessel color image into the segmentation model, and outputting a vessel segmentation result. The method is applied to capillary segmentation in microscopic images.

The invention discloses a retinal vessel image segmentation system based on a global information convolutional neural network, and relates to a retinal vessel image segmentation system, and aims to solve the problems of limited global information utilization and easy loss of important features in existing convolutional neural network retinal vessel image segmentation. The system comprises an imageprocessing main module, a neural network main module, a training main module and a detection main module. The image processing main module is used for acquiring an original retina image, preprocessing the acquired original retina image and inputting the processed image into the training main module and the detection main module; the neural network main module is used for establishing a convolutional neural network capable of extracting global information and strengthening features; the training main module is used for initializing network parameters to obtain a trained convolutional neural network model; and the detection main module is used for testing by using the trained model and calculating a model performance index. The invention belongs to the field of retinal vessel image segmentation systems.

The invention provides a retina vessel segmentation method based on residual network feature extraction, which is applied to a neural network model, and comprises the following steps: enabling an original retina vessel image to pass through a pre-trained VGG coding layer to obtain a plurality of images, wherein the number of the images is five, the sizes of the images have a preset proportional relation with the sizes of the original retinal blood vessel images. Therefore, the segmentation precision of the network is obviously improved, and the fitting ability and generalization ability of the model are better optimized. Compared with a Unet, the network also generally has better performance in other data sets.

The invention discloses a retinal blood vessel segmentation method based on a convolutional neural network, which includes: preprocessing the retinal fundus image; performing block extraction on training set images; constructing a convolutional neural network for blood vessel segmentation, and using the extracted image blocks to perform Training; in the prediction stage, multiple consecutive overlapping segments are extracted for each image, and the classification probability of each pixel is obtained by averaging multiple prediction results, and the final segmentation result map is obtained. The new convolutional neural network structure proposed by the present invention for retinal vessel segmentation is a symmetrical network based on the Encoder-Decoder structure, and two skip connections are added between the Encoder part and the Decoder part. The network can not only realize the end-to-end segmentation of retinal images, but also can obtain accurate segmentation results on a limited data set, and can effectively avoid the problem of gradient disappearance, which has certain advantages compared with existing algorithms.

The invention relates to a retinal vessel reconstruction method based on a directed vessel tree. The method comprises the following steps: acquiring retinal image information and segmenting a retinal vessel by adopting a convolutional neural network; extracting a skeleton line of the blood vessel by using a convolution algorithm; intercepting a visual disc area and extracting an end node and a branch node from the skeleton line; segmenting a blood vessel section through branch nodes, and extracting skeleton lines and outer contour information of the blood vessel section to construct a directed blood vessel tree; and reconstructing the skeleton line of the retinal blood vessel and the image information of the blood vessel by using the directed blood vessel tree. According to the invention, blood vessel segments serve as main research objects, recognition and relation construction of the blood vessel segments serve as ultimate targets, blood vessels are segmented through a convolutional neural network, a skeleton of the blood vessels is extracted through a convolution algorithm, the blood vessel skeleton is segmented according to the geometrical relation, and finally a blood vessel structure diagram is reconstructed through a directed blood vessel tree. And different blood vessels are distinguished through the topological structure diagram of the reconstructed blood vessels of the directed blood vessel tree and the image information of the blood vessels.