109 results about "Lesion segmentation" patented technology

Computerized interpretation of medical images for quantitative analysis of multi-modality breast images including analysis of FFDM, 2D / 3D ultrasound, MRI, or other breast imaging methods. Real-time characterization of tumors and background tissue, and calculation of image-based biomarkers is provided for breast cancer detection, diagnosis, prognosis, risk assessment, and therapy response. Analysis includes lesion segmentation, and extraction of relevant characteristics (textural / morphological / kinetic features) from lesion-based or voxel-based analyses. Combinations of characteristics in several classification tasks using artificial intelligence is provided. Output in terms of 1D, 2D or 3D distributions in which an unknown case is identified relative to calculations on known or unlabeled cases, which can go through a dimension-reduction technique. Output to 3D shows relationships of the unknown case to a cloud of known or unlabeled cases, in which the cloud demonstrates the structure of the population of patients with and without the disease.

A method for lesion segmentation in 3-dimensional (3D) digital images, includes selecting a 2D region of interest (ROI) from a 3D image, the ROI containing a suspected lesion, extending borders of the ROI to 3D forming a volume of interest (VOI), where voxels on the borders of the VOI are initialized as background voxels and voxels in an interior of the VOI are initialized as foreground voxels, propagating a foreground and background voxel competition where for each voxel in the VOI, having each neighbor voxel in a neighborhood of the voxel attack the voxel, and, if the attack is successful, updating a label and strength of the voxel with that of the successful attacking voxel, and evolving a surface between the foreground and background voxels in 3D until an energy functional associated with the surface converges in value, where the surface segments the suspected lesion from the image.

The invention discloses a nasopharyngeal-carcinoma (NPC) lesion automatic-segmentation method and nasopharyngeal-carcinoma lesion automatic-segmentation systems based on deep learning. The method comprises: carrying out registration on a PET (Positron Emission Tomography) image and a CT (Computed Tomography) image of nasopharyngeal carcinoma to obtain a PET image and a CT image after registration;and inputting the PET image and the CT image after registration into a convolutional neural network to carry out feature representation and scores map reconstruction to obtain a nasopharyngeal-carcinoma lesion segmentation result graph. The method carries out registration on the PET image and the CT image of the nasopharyngeal carcinoma, obtains a nasopharyngeal-carcinoma lesion by automatic segmentation through the convolutional neural network, and is more objective and accurate as compared with manual segmentation manners of doctors; and the convolutional neural network in deep learning isadopted, consistency is better, feature learning ability is higher, the problems of dimension disasters, easy falling into a local optimum and the like are solved, lesion segmentation can be carried out on multi-modal images of the PET-CT images, and an application range is wider. The method can be widely applied to the field of medical image processing.

A method and system for acquiring information on lesions in dynamic 3D medical images of a body region and / or organ, having the steps: applying a registration technique to align a plurality of volumetric image data of the body region and / or organ, yielding multi-phase registered volumetric image data of the body region and / or organ; applying a hierarchical segmentation on the multi-phase registered volumetric image data of the body region and / or organ, the segmentation yielding a plurality of clusters of n-dimensional voxel vectors of the multi-phase aligned volumetric image; determining from the plurality of clusters a cluster or set of clusters delineating the body and / or organ; identifying the connected region(s) of voxel vectors belonging to the body region and / or organ; refining / filling the connected region(s) corresponding to the body region and / or organ; reapplying the segmentation step to the refined / filled connected region(s) corresponding to the body region and / or organ, to obtain a more accurate segmentation; and acquiring information on the presence of lesions in cluster of set of clusters of said plurality of clusters.

The present invention provides a method, device, equipment for the segmentation of a lesion in a biological image and a storage medium. The method comprises a step of acquiring a target biological image, a step of performing coarse segmentation processing on the target biological image and obtaining a coarse segmentation mask after the rough segmentation processing, wherein the coarse segmentationmask includes information of candidate lesions in the target biological image, a step of identifying a non-real lesion from the candidate lesions, correcting the rough segmentation mask based on a recognition result such that the information of an identified non-real lesion is not included in the coarse segmentation mask and a target segmentation mask obtained after the correction is used as a lesion segmentation mask corresponding to the target biological image. According to the method, the device, the equipment and the storage medium, the lesion can be automatically positioned from the target biological image, the mode is labor-saving, the time consumption of the positioning of the lesion is reduced, misdiagnosis and missed diagnosis caused by the manual positioning of the lesion are avoided, the positioned lesion can also assist the doctor to carry out fast and accurate analysis, and the diagnostic efficiency and diagnostic accuracy of doctors are improved.

The invention provides an improved image segmentation training method based on a full convolutional neural network in allusion to the current situation that the difficulty in melanoma skin lesion segmentation is high and a simple effective rapid image segmentation method is absent. The method comprises the following steps: firstly executing data enhancement to a training sample, executing normalization processing, executing sampling segmentation to the processed sample, and classifying images after the sampling segmentation so as to realize classification and identification training based on the traditional convolutional neural network; and assigning parameters of the classification network to an improved full convolutional network, placing the training sample with the original size in thenetwork, and training to obtain a prediction probability graph to provide segmentation for a skin melanoma lesion picture. The method is capable of effectively improving the supervision of the full convolutional network upon image segmentation training, improving the training efficiency, and increasing segmentation accuracy.

The invention aims to solve problems of poor effect, small quantity of training samples, large difference among samples of a traditional extracted feature resorting method due to large difficulty in melanoma skin lesion segmentation and provides a multi-scale deep learning recognition system based on expansion and convolution. The system includes performing data enhancement and normalization processing on training samples and training an extracted expansion and convolution based multi-scale feature learning neural network, performing multi-threshold segmentation based on an obtained predictionprobability map, and thus implementing segmentation of a melanoma skin lesion image. Finally, the segmentation accuracy is improved.

The present invention, referred to as Oasis is Automated Statistical Inference for Segmentation (OASIS), is a fully automated and robust statistical method for cross-sectional MS lesion segmentation. Using intensity information from multiple modalities of MRI, a logistic regression model assigns voxel-level probabilities of lesion presence. The OASIS model produces interpretable results in the form of regression coefficients that can be applied to imaging studies quickly and easily. OASIS uses intensity-normalized brain MRI volumes, enabling the model to be robust to changes in scanner and acquisition sequence. OASIS also adjusts for intensity inhomogeneities that preprocessing bias field correction procedures do not remove, using BLUR volumes. This allows for more accurate segmentation of brain areas that are highly distorted by inhomogeneities, such as the cerebellum. One of the most practical properties of OASIS is that the method is fully transparent, easy to implement, and simple to modify for new data sets.

An apparatus and method are provided including a first segmenter and a second segmenter. The first segmenter is configured to generate a first segmentation result from a medical image using a first segmentation parameter for a candidate lesion. The second segmenter is configured to determine a target lesion to segment from among the candidate lesion based on the first segmentation result, and generate a second segmentation result using a second segmentation parameter to segment the target lesion.

The embodiment of the invention provides a pneumonia lesion segmentation method and device, and solves the problems of low accuracy and low efficiency of an existing pneumonia lesion segmentation mode. The pneumonia lesion segmentation method comprises the following steps: predicting a lesion area on medical image data of a positive level based on an image semantic segmentation model; counting thelesion area of each parallel layer, and calculating the lesion volume by combining the lesion area of each parallel layer, wherein the image semantic segmentation model is established through the following training steps: inputting all or part of marked sample data into a focus segmentation model to obtain a prediction result output by the focus segmentation model; based on a focus detection frame predicted by a focus detection model and a lung region predicted by a lung lobe lung segment segmentation model, screening out a low-level false positive region from a prediction result to obtain afalse label of sample data, and adding unmarked sample data; and rechecking the pseudo label, and marking the marked sample data to update the marked sample data.

The invention belongs to the field of computer aided diagnosis and medical image processing, and relates to a skin disease image focus segmentation method based on a deep convolutional neural network,which is used for improving the quality of a skin disease image and further improving the focus segmentation accuracy so as to obtain more accurate focus information. The method comprises the specific steps that data preprocessing is responsible for carrying out noise reduction processing on a skin disease image, and removing artificial and natural noise which hinders focus position determinationin the image; the data expansion is responsible for expanding a data set by deforming and rotating the image subjected to the noise reduction processing; a segmentation model is constructed to perform first feature extraction on the image, encoding is performed to obtain more detail features, and the features obtained at the first time are fused to obtain a prediction graph;.

The invention discloses a pancreatic neuroendocrine tumor automatic segmentation method and system based on deep learning. The method comprises the steps of obtaining a computed tomography enhanced image of a pancreatic neuroendocrine tumor patient; adopting a deep learning method to automatically segment the focus of the obtained computed tomography enhanced image, wherein the deep learning method adopts a deep convolutional neural network. According to the invention, automatic focus segmentation is carried out on the obtained computed tomography enhanced image by using a deep learning method; deep learning and computed tomography enhanced images are combined and applied to lesion segmentation of pancreatic neuroendocrine tumors. The method can automatically segment the focus area of thetumor through the feature learning of the deep convolutional neural network, is less affected by the experience of a doctor, is more accurate, saves the time and energy of the doctor for manually sketching the focus area, and is higher in efficiency. The method can be widely applied to the field of medical image processing.

The invention belongs to the technical field of artificial intelligence image processing, and particularly relates to a medical image lesion segmentation method, which comprises the following steps ofobtaining medical CT image data, and performing preprocessing work such as binarization and feature labeling to obtain an original data set; carrying out denoising processing on the medical image byusing a denoising method, and carrying out image fusion on various images with different modes; carrying out data enhancement methods such as translation, rotation, overturning and shifting by utilizing the obtained data set to realize expansion of the data set; constructing a segmentation network model through a fusion module; fusing and splicing the feature maps obtained in different stages in the segmentation network model, and performing convolution operation to output a predicted segmentation map; in the constructed segmentation network model, carrying out training by using the data set to obtain information of a loss function and a segmentation result; and according to the medical image segmentation method and system, generating and storing a trained medical image segmentation network model. The segmentation accuracy of the model is improved.

A method of analyzing image data comprises: obtaining a first image of a first part of an object; obtaining a second image of a second part of the object having overlap with the first part; obtaining a mapping between the first and second images; segmenting the second image to obtain a segmentation; detecting outliers in the first image by identifying extreme intensity values of elements within one or more classes of elements on the basis of the segmentation; replacing elements of the second image that correspond to at least some outliers of the first image, with replacement values, to obtain a corrected second image; and updating the segmentation by performing the segmenting on the corrected second image. The detecting outliers, the replacing, and the updating are performed iteratively until a predetermined convergence criterion is met, which represents a point where there is no significant change in the tissue and lesion segmentations.

The invention relates to an automatic segmentation method for an abdominal CT liver lesion image based on a three-stage cascade network. The method comprises the following steps: S1, acquiring three-dimensional abdominal liver CT image data; S2, preprocessing and standardizing the obtained three-dimensional abdominal liver CT image data; S3, inputting the three-dimensional abdominal liver CT imagedata after preprocessing and data standardization into AuxResUnet liver image segmentation model, and then taking 3D maximum connected region from the obtained three-dimensional abdominal liver CT image data segmentation result to exclude false positive region, so as to obtain liver VOI; S4, adopting the segmentation result of the three-dimensional abdominal liver CT image data obtained by S3 asa mask of the CT liver image data to cover the liver VOI obtain by S3; S5, inputting the covered liver VOI into AuxResUnet liver image lesion segmentation model for lesion segmentation, and obtainingthe liver image lesion segmentation result. The image segmentation method provided by the invention can realize fast and accurate segmentation of liver and liver pathological changes.

A method of collecting information regarding an anatomical object of interest includes displaying an image characterized by a first region and a second region, wherein the first and second regions are mutually exclusive and the object is displayed within the second region, selecting first and second points spanning the object in the displayed image, at least one of the points being within the first region, and extracting a plurality of statistical values from image voxels, lying on a line segment between the first and second points, that correspond to the object.

The invention discloses a method, a system and a medium for generating registered multi-modal MRI with focus segmentation tags. The method comprises the following steps: acquiring a random matrix in normal distribution, inputting the random matrix into a trained structural feature decoder in a generative adversarial network, and decoding to generate a structural feature map; fusing the structuralfeature map and the randomly selected focus segmentation label map through random input to obtain a fusion result, and inputting the fusion result into a trained random encoder in the generative adversarial network to obtain a code; and inputting the code into a decoder of each mode trained in the generative adversarial network, and respectively generating registered multi-mode MRI. The generatorin the generative adversarial network is modularized into the encoders and the decoders, through combined training of the encoders, the decoders and the discriminators, random input conforming to design specifications can be received, then a set of registered multi-mode MRI images with focus segmentation tags are generated, and the method can be widely applied to the field of medical images.

The invention discloses a hepatic echinococcosis focus segmentation method and a hepatic echinococcosis focus segmentation system based on a neural network. The method comprises the following steps: S1, training and verifying a cystic echinococcosis focus segmentation model; S2, training and verifying a follicular echinococcosis focus segmentation model; S3, obtaining a segmented liver region fromthe one-pack worm CT image, and inputting the liver region into the focus recognition model to obtain a recognition result; S4, when it is determined that the recognition result is the cystic echinococcosis focus, inputting the VOI region into the cystic echinococcosis focus segmentation model to obtain a first segmentation result; And S5, when it is determined that the recognition result is thefollicular echinococcosis lesion, performing blood vessel recognition and segmentation on the VOI region, and inputting the blood vessel segmentation result and the VOI region into the follicular echinococcosis lesion segmentation model to obtain a second segmentation result. According to the method and the system provided by the invention, fusion recognition and feature extraction are carried outon the multi-modal medical image through various models, a doctor is assisted to carry out echinococcosis screening work, and the diagnosis efficiency and accuracy are improved.

The invention discloses a fundus image lesion segmentation method based on deep network aggregation, and the method comprises the following steps: 1), obtaining a plurality of fundus lesion images, carrying out the manual segmentation of the lesion contour in each fundus lesion image, obtaining a true value label, and constructing a training set and a test set; 2) adding a deep aggregation networkmodule into a backbone network of the U-Net model; 3) migrating the U-Net model obtained in the step 2) to focus segmentation of the fundus image, training the U-Net model, and taking the trained U-Net model as a fundus image focus segmentation model; and 4) segmenting the fundus image to be segmented by using the fundus image lesion segmentation model. The method can effectively solve the problem of poor fundus image lesion segmentation effect based on the deep convolutional neural network in the prior art.

The invention discloses a multi-mode nuclear magnetic image cerebral arterial thrombosis lesion segmentation method based on a convolutional neural network. The method comprises the steps of designinga deep convolutional neural network 1 based on an original framework of a 3D U-Net network; basic modules based on a 3 * 3 * 3 convolution layer and a batch standardization layer; a 3D deformable convolution module based on a 3 * 3 * 3 convolution layer and trilinear interpolation; a cascade deformation module based on the 3D deformable convolution module and the basic module; constructing a deepconvolutional neural network 2 based on the 3D U-Net network and the cascaded deformation module; inputting the processed data into a neural network 1 for pre-training; assigning a weight obtained bypre-training to a neural network 2 for training; and verifying the segmentation effect on the test set of the pixel-level label. The invention provides an automatic labeling method for ischemic stroke lesion segmentation, so that the cost of labeling data is greatly reduced, the engineering operability is enhanced to a certain extent, and doctors are assisted in clinical diagnosis of ischemic stroke patients.

The invention discloses an acute cerebral apoplexy lesion segmentation method based on small sample learning. The method comprises the steps of training a convolutional neural network by a data samplewith an image-level label, and using the classification accuracy of an image as a measurement index; constructing a new convolutional neural network by using the trained convolutional neural network,and constructing an end-to-end convolutional neural network by using a feature map obtained by the trained network from the input image; fixing trained convolutional layer parameters, training a newly constructed convolutional neural network by using a small number of data samples of pixel-level tags, and taking the segmentation precision of the image as a measurement index; and after the training is finished, verifying the segmentation effect of the network on the test set of the pixel-level label. According to the method, only a small number of pixel-level label data samples and a small number of image-level label data samples are used, so that the cost of labeling data is greatly reduced, the engineering operability is enhanced to a certain extent, and doctors are assisted in clinicaldiagnosis of acute cerebral stroke patients.

The invention discloses a GA lesion segmentation method based on depth-concatenated model for SD-OCT images. The method comprises steps of three kinds of depth network models with different layers arefirstly configured; The first layer is the input layer, the last layer is the output layer, and the middle layer is a sparse self-encoder with different number of neurons, and the encoding and decoding process are symmetrically distributed. The training is divided into two stages, self-supervised feature extraction stage and supervised-based classifier training stage. After the first stage training is completed, the first stage coding process is added with soft-max loss function training basic classifier inputs the labeled positive and negative samples with h-dimension characteristics into the depth network model, and trains the output layer of the soft-max classifier obtains the final segmentation result. Finally, based on Adaboost cascade strategy, the training process of the above model is fused to improve the final segmentation results. This method improves the segmentation accuracy of GA lesions and is of great significance for the prevention and diagnosis of macular diseases inthe elderly.

The invention discloses a mammary gland DCE-MRI image lesion segmentation model establishment based on hybrid convolution and segmentation method. The segmentation model establishing method comprisesthe following steps: firstly obtaining a three-channel image of each DCE-MRI sequence image in a mammary gland DCE-MRI image set, secondly constructing a mammary gland DCE-MRI image focus segmentationnetwork based on hybrid convolution and an ASPP network, and finally training the obtained segmentation network by using the three-channel images to obtain a trained segmentation model; and based onthe obtained segmentation model, preprocessing any DCE-MRI sequence image to be processed to obtain a three-channel image, and then inputting the three-channel image into the segmentation model to obtain a focus segmentation result. According to the method, 3D spatial features of the image are extracted through mixed 2D and 3D convolution, and a more accurate segmentation result is achieved; in addition, the ASPP is used for extracting multi-scale context features, so that the influence of lesion size difference on a segmentation result is effectively solved.

In a method and apparatus for automatic background region selection for lesion segmentation in medical images, a patient medical image dataset is loaded into a computer and an image region is delineated to obtain a segmentation containing a lesion. A background region is created from the segmentation representing the patient organ.

The invention provides a panoramic dental film focus detection and segmentation method and device for deep learning. The method comprises the steps: obtaining panoramic dental film data, carrying outthe preprocessing of the panoramic dental film data, and obtaining to-be-detected segmentation data; multi-lesion detection positioning and contour segmentation are carried out on the to-be-detected segmentation data to obtain lesion detection results, and the lesion detection results comprise lesion classification, lesion position coordinates and lesion segmentation contours; performing tooth detection positioning on the to-be-detected segmentation data by using a tooth position classification detection model to obtain a tooth position detection result, the tooth position detection result comprising an FDI standard tooth position number corresponding to each tooth; matching the tooth position detection result with the focus detection result to obtain a tooth position number correspondingto each focus; smoothing the focus segmentation contour through a contour smoothing algorithm to obtain a processed contour; and visually outputting a panoramic dental film focus detection and segmentation result.

The invention discloses a skin cancer lesion segmentation method based on deep learning. The method comprises the steps of 1, obtaining a training dermatoscope image sample; 2, normalizing data; 3, designing an edge perception neural network model; 4, training the edge perception neural network model; 5, performing segmentation. Shallow detail information and deep semantic information are fused, edge details of images can be well detected. A Multi-Block module is used for expanding the receptive field of the model so as to enhance the sensitivity to targets of different scales, and meanwhile,the interference of background information is suppressed in combination with a spatial attention mechanism.