82 results about "Lesion mapping" patented technology

The invention provides a method for detecting an X-ray mammary gland lesion image based on feature pyramid network under transfer learning. The method comprises the steps: 1, establishing a source domain data set and a target domain data set; 2, establishing a deformable convolution residual network layer by a deformable convolution and extended residual network module; 3, establishing a multi-scale feature extraction sub-network based on a feature pyramid structure through a feature map up-sampling and feature fusion method in combination with a deformable convolution residual network layer;4, establishing a deformable pooling sub-network sensitive to the focus position; 5, establishing a post-processing network layer to optimize a prediction result and a loss function; and 6, migratingthe training model to a small sample molybdenum target X-ray mammary gland focus detection task so as to improve the detection precision of the network model on the focus in the small sample image. According to the method, a transfer learning strategy is combined to realize focus image processing in a small sample medical image.

The invention relates to the technical field of medical equipment, and discloses a focus image recognition method and a focus image recognition system based on a deep learning model. The invention provides a new method and a new system for automatically identifying conditions of medical images by replacing doctors based on the deep learning model. Regarding acquired to-be-detected medical images,Segmentation of a smallest organ tissue image and image recognition of a corresponding organ tissue, prediction of a deep learning model and marking on a predication result are carried out. The invention can replace doctors find out underlying conditions of medical images and automatically mark lesion tissues on the medical images. Therefore, a reminder is given to doctors for timely diagnosis. Working intensity of the doctors can be reduced. Whether diseases occur or not can be timely diagnosed. The treatment opportunity of conditions is prevented from being delayed, and the discovery of early-stage diseases is particularly facilitated.

The invention belongs to the technical field of medical image processing, and discloses a three-dimensional MRI semi-automatic lesion image segmenting method and system. The method includes: determining a range of a lesion slice and a position of an initial slice by observing an MRI three-dimensional image, and dividing an initial lesion region and an initial normal region on the initial slice; segmenting the initial slice lesion, using a classifier to classify voxels in an extended region, and obtaining a final lesion region and a final normal region of the initial slice after multiple iterations; projecting the final lesion region and the final normal region on the initial slice onto an adjacent slice to obtain an initial lesion region and an initial normal region of the adjacent slice;and segmenting other section lesions, repeating slice lesion segmentation and region projection, segmenting the lesions within the lesion slice range, and performing combination to obtain the entire lesion region. After the above operation, the method can obtain a three-dimensional MRI lesion image.

The invention relates to a method and a system for automatically updating a medical ultrasonic image auxiliary diagnosis system. The method comprises the following steps of: acquiring correction information and a corresponding original focus image, re-training and updating a nodule key frame extraction module and a key frame nodule property judgment module according to the obtained correction information and the corresponding original focus image, and enabling the medical ultrasonic image auxiliary diagnosis system to use the updated system to generate a new diagnosis report. According to thesystem, a closed loop with a continuously updated model is formed, a feedback capture process is provided, correction information given by a doctor is obtained, the system is continuously updated, andthe accuracy of auxiliary diagnosis can be continuously improved.

The invention provides an image registration method and device, a surgical robot and a surgical robot system. According to the specific implementation scheme, the method comprises the steps of obtaining a first transformation matrix used for representing an initial mapping relation between a space coordinate system of a physical space and an image coordinate system of a focus image through coarse registration; through a fine registration process, eliminating distance deviation and angle deviation existing in a coarse registration result, and obtaining a second transformation matrix, a third transformation matrix and a fourth transformation matrix; and generating a target transformation matrix between the space coordinate system and the image coordinate system according to the first transformation matrix, the second transformation matrix, the third transformation matrix and the fourth transformation matrix. Therefore, the image registration precision is improved, and the safety of subsequent operation by using the image registration result is improved. For a computer processor, the operation precision and efficiency of the computer processor are also improved.

The invention relates to a medical ultrasonic image computer-aided diagnosis method. The method comprises the following steps: directly acquiring a video signal from medical ultrasonic equipment to obtain ultrasonic images; extracting key frame images containing a focus in the ultrasonic images and carrying out property judgment; performing comprehensive judgment in combination with all the ultrasonic images confirmed as the focus obtained by the scanning, outputting a scanned diagnosis report, and acquiring correction information and corresponding original focus images; re-training and updating a nodule key frame extraction module and a key frame nodule property judgment module according to the obtained correction information and the corresponding original focus images; and generating a new diagnosis report by using an updated model. The method can adapt to a current ultrasonic diagnosis scene, effectively assists a doctor in identifying a focus in the video frame images, forms a report and outputs the report, and improves the diagnosis efficiency and accuracy. The method can also acquire correction information given by a doctor, so that the system is continuously updated, and theaccuracy of auxiliary diagnosis can be continuously improved.

The invention discloses a multi-modal cerebral apoplexy lesion segmentation method based on small sample learning, and the method comprises the steps: obtaining an original brain training sample which comprises a multi-modal CT medical image and an annotation image, and the original training sample comprises CT, CBF, CBV, MTT, TMax and ischemic cerebral apoplexy lesion tags; the method comprises the following steps: preprocessing a multi-modal medical image, carrying out image augmentation through modes of image deformation, image scaling, generative adversarial and the like, and expanding a small sample image data set; registering the multi-modal image after data augmentation, taking a reference image as a CT image, and performing pixel-level fusion on the multi-modal image after registration; and the fused multi-modal image data is transmitted to a segmentation network constructed based on Transform to carry out image segmentation. The invention further discloses a system using the method. According to the method, the influence on a medical image data segmentation task caused by insufficient data samples is improved, more focus image information is obtained through multi-modal image fusion, and the accuracy of image segmentation is improved.

The invention provides a gastrointestinal disease model construction method and a diagnosis system. The diagnosis system comprises an image acquisition unit, an image labeling unit, an image databaseconstruction unit, an image recognition unit and a data output unit, wherein the image database construction unit comprises a normal image feature library and a focus image feature library, and the image recognition unit comprises a focus type recognition model. Firstly, a normal image feature library of a normal gastroscope image and a plurality of lesion image feature libraries of a lesion gastroscope image are constructed; a focus type identification model is used; color feature and texture feature fusion is carried out on a gastroscope image to be detected, matching and recognition are carried out on the gastroscope image to be detected, a normal image feature library and a focus image feature library in sequence, whether the gastroscope image to be detected contains focuses or not andfocus types are output, and therefore the focuses with various stomach types can be recognized rapidly and accurately.

The invention belongs to the technical field of image processing, and discloses a liver lesion image processing method and system, a storage medium, a program and a terminal. An information processingserver acquires basic information and a CT image of an examiner who has undergone chest CT from a security port; through a gRPC, the information processing server transmits the image to the algorithmserver for calculation, and an algorithm server transmits a result back to the information processing server; the information processing server writes the inspector information and the calculation result into a MySQL and MongoDB database, and a RESTful standard interface is established to expose the data; and a front-end webpage displays liver and possible focus information. On the basis of the existing classic neural network unet architecture, on the basis of continuously feeding back an increased liver lesion data set, through comprehensive improvement of the model, the liver recognition accuracy can be improved to 97%, and the liver occupation recognition accuracy is improved to 98%.

The invention discloses a lung disease lesion unsupervised segmentation method based on knowledge distillation, and belongs to the field of medical image processing and computer vision. The method comprises the following steps: firstly, constructing and training an auto-encoder to obtain a pre-trained teacher network with rich CT image semantic knowledge, then training a student network with the same architecture as the teacher network from the pre-trained teacher network by only distilling the knowledge of normal CT images, and finally, performing lesion segmentation on the difference of features extracted by the lesion-containing image by using teacher and student networks. Meanwhile, besides traditional pixel-level distillation, affinity-level distillation considering the relation between pixels is further designed in the method, so that effective knowledge is fully distilled. Experiments prove that the lesion segmentation precision can be effectively improved on different data sets. The method is easy to construct, an unlabeled lung disease focus segmentation result can be obtained only depending on normal data, and generalization and operation efficiency are considerable.

The invention provides an image recognition method and device, electronic equipment and a storage medium. The method comprises the following steps: firstly, acquiring an initial image of medical scanning corresponding to a target part, and then preprocessing the initial image to obtain at least one of a post-processing image of a blood vessel in the target part and an abnormal region in which a blood vessel center line parameter exists; and performing focus identification based on the initial image and the post-processing image of the blood vessel to obtain a focus area, and determining and identifying a focus image in the initial image according to the focus area and the abnormal area to obtain a focus identification result. According to the method and device, focus identification is performed on the initial image by adopting multiple ways, so that omission of manual identification is made up, and the accuracy of an identification result is improved.

The invention relates to an intelligent gastroscope image processing device, including a gastroscope device housing; a probe is arranged on one side of the gastroscope device housing; a 3D image display and a split screen touch display are arranged on the front side of the gastroscope device housing; the connected end of the split screen touch display is provided with an image editing unit; a PLCis arranged inside the housing; the PLC is internally provided with a coding classification unit; the input end of the PLC is provided with an image processing system and a data processing system; theinput end of the image processing system is provided with an image pickup module; the image pickup module is internally provided with an A coding unit; and the connection end of the image processingsystem is provided with a first database. The invention can perform color contrast and lesion feature contrast on the photographed picture through the setting of the image processing system, and the setting of a picture search engine can search lesion pictures uploaded on the network, conveniently assisting a doctor to quickly judge the condition, shortening the time for a patient to do the gastroscope and reducing the pain.

The invention provides an OCT image recommendation method and device based on artificial intelligence, electronic equipment and a storage medium, and the method comprises the steps: obtaining fundus OCT images, and carrying out the preprocessing of all the obtained fundus OCT images, and obtaining an OCT to-be-detected image set; inputting images in the OCT to-be-detected image set into a network detection model for focus detection to obtain a focus image set; constructing a lesion vector, and screening the lesion image set based on the lesion vector to obtain an effective lesion image set; calculating a focus characteristic index of each image in the effective focus image set; and evaluating images in the effective lesion image set based on the lesion feature indexes to obtain an OCT recommended image. According to the method, the focus characteristic indexes of the OCT images are calculated, and the vectors are constructed according to the focus characteristic indexes for operation to quickly recommend similar OCT images, so that the reading and diagnosis efficiency of ophthalmologists is improved.

The invention discloses a lesion image generation method and device, electronic equipment and a storage medium, and the method comprises the steps: carrying out the image recognition processing of each historical lesion image in at least one historical lesion image, and obtaining at least one mask image in one-to-one correspondence with the at least one historical lesion image; performing superposition processing on each mask image and the corresponding historical focus image to obtain at least one superposition image in one-to-one correspondence with the at least one mask image; training an improved cyclic adversarial neural network according to the at least one superimposed image, the at least one mask image and the at least one historical lesion image to obtain a lesion image generation network; acquiring an original image of a to-be-generated focus; randomly selecting one or more images from the at least one mask image, and carrying out superposition processing on the selected one or more images and the original image to obtain a to-be-processed image; and inputting the to-be-processed image into the focus image generation network to generate a focus image.

The invention provides a wireless capsule endoscope lesion automatic detection method and device and a storage medium. The method mainly comprises the steps that a video file is converted into an image sequence with a timestamp, and a data set with labels is obtained through key frame expert labeling and target tracking of a similar interval sequence; a focus detection model is trained through a convolutional neural network, the trained model is loaded to predict a focus detection result of the to-be-detected video, and the focus detection result comprises timestamps, focus positions, focus categories and confidence coefficients of all images with focuses; obtaining a manual correction result of an expert for a'difficult example 'sample in the detection result, and obtaining an error detection area; and transforming the error detection area, re-synthesizing the error detection area and the background into a new data set, and finely adjusting the original model to gradually optimize the model effect. According to the method, the high-performance focus detection model is obtained under the condition that expert annotation data is as few as possible, so that experts are assisted to complete focus automatic detection, and the film reading efficiency and the detection performance are improved.

The invention discloses an image data generation method and device, relates to the technical field of image processing, and aims to solve the problem that a data generation method in the prior art cannot provide effective training data samples for an artificial intelligence medical image analysis process. The method mainly comprises the following steps: acquiring a focus-free image; generating a predicted lesion mask according to a preset lesion contour; and inputting the focus-free image and the predicted focus mask into an image generation model to generate a predicted focus image, the imagegeneration model being generated by training a marked focus training picture. The method is mainly applied to the process of expanding the image data.

The invention provides a thyroid image processing method and device, electronic equipment and a storage medium. The method comprises the following steps: acquiring a to-be-detected image, wherein the to-be-detected image comprises an image obtained by detecting a thyroid part; inputting the to-be-detected image into a preset deep learning model to obtain a first detection result of the preset deep learning model on the to-be-detected image; and when the first detection result shows that the thyroid nodules exist in the to-be-detected image, inputting a lesion image region representing that the thyroid nodules exist in the to-be-detected image in the first detection result into the tested machine learning model to obtain a second detection result of the machine learning model on the lesion image region. In the scheme, the to-be-detected image is detected in combination with the preset deep learning model and the machine learning model, and the machine learning model does not need to detect the image without the thyroid nodule and the area which is not the lesion image area, so that the operand can be reduced, the operation efficiency can be improved, and the thyroid nodule detection accuracy can be improved.

The invention provides a lung CT image segmentation device based on spatial neighborhood analysis. On the basis of extracting two-dimensional image features of a single CT layer, contextual three-dimensional image features between neighborhood CT sequences are fused in parallel through three-dimensional convolution, and expression of three-dimensional focus area image features is achieved while the full 3D convolution operand and the parameter scale are reduced; meanwhile, channel domain two-dimensional image feature components corresponding to each neighborhood layer slice sequence are remapped from the context fusion feature map by utilizing a self-attention mechanism, so that the feature decoding process of a single CT layer is guided, and the focus image segmentation accuracy is improved; in order to improve the adaptability and interpretability of the algorithm, interpretable priori knowledge is introduced as an additional image segmentation judgment rule, so that the segmentation result is calibrated and checked, and a basis is provided for clinical auxiliary diagnosis.

The invention discloses a magnetic resonance navigation method and device for a microwave thermal ablation operation of a liver, and relates to the technical field of medical instruments and magnetic resonance, and the method comprises the following steps: collecting a magnetic resonance anatomical image of a to-be-ablated area of the liver; identifying a microwave probe image and a lesion image from the anatomical image, and positioning tracking time coordinates of a microwave probe and a lesion according to the microwave probe image and the lesion image; and predicting actual coordinates of the current moment through the tracking moment coordinates of the microwave probe and the lesion, and generating a navigation path of the microwave probe according to the actual coordinates so as to carry out navigation on the microwave probe according to the navigation path. According to the method, real-time and accurate navigation information can be provided for doctors, the requirements of the doctors are met, system delay is effectively overcome, real-time feedback is achieved, and the operation experience of the doctors and patients is improved.

The invention relates to a focus image classification and identification method based on fundus images, and belongs to the field of medical image processing. According to the method, a YOLOv5 network structure and a target frame are fused in a weighted mode to form a frame, data sets of collected eye fundus images are deepened through a Mosaic method, the deepened image data sets and loss function training and image classification are combined, then the images enter the target frame weighted fusion frame, and finally binary images of different eye fundus images are output according to classification probabilities. And outputting the focus binary images in a classified manner. According to the method, the focus feature extraction capability of the network model is improved.

The invention discloses an image recognition system for digestive tract lesions, which includes a memory, an image preprocessing module, an image feature extraction module, a machine learning module, and an image recognition module, wherein the storage data communication end of the memory is connected to the image preprocessing module The data input end of the image preprocessing module, the data output end of the image preprocessing module is connected to the data input end of the image feature extraction module, the first data output end of the image feature extraction module is connected to the data input end of the machine learning module, and the second data of the image feature extraction module The output end is connected to the first data input end of the image recognition module, and the data output end of the machine learning module is connected to the second data input end of the image recognition module. The invention improves the efficiency and accuracy of image recognition of digestive tract lesions.

The invention relates to a focus area image classification method based on a convolutional neural network. The method specifically comprises the following steps: (1) collecting focus images to manufacture an image database, and marking the types of the images in the image database; (2) enhancing an original image in a training sample set by adopting an image enhancement algorithm to improve the texture and detail definition of the image; (3) constructing a lesion classification network, training the classification network by using the training sample set, and determining network parameters toobtain a classification model; and (4) inputting a lesion image to be tested into the lesion classification network to obtain the category to which the lesion belongs, and completing the classification of the lesion image. According to the method, global and color features and texture and detail features are extracted based on O-stream and P-stream models through the lesion classification networkdouble-stream CNN, and the final classification effect is effectively improved.

The invention provides a method and device for generating a focus on a CT image and a storage medium. The method comprises the following steps: acquiring continuous original CT images and focus generation data; establishing a first-stage cube taking the focus as the center; generating a focus contour mask based on the cut CT image, the focus generation data and a pre-trained first generative adversarial network; generating a focus CT image based on the cut CT image, the focus generation data, the focus contour mask and a pre-trained second generative adversarial network; and inserting the focus CT image into an original CT image to generate a complete CT image. According to the invention, through the two generative adversarial network models, the lesion contour is generated firstly, then the complete lesion is generated, and the problems of overfitting and obvious square edges existing in other lesion generation methods are solved.

The invention discloses an auxiliary diagnosis method for congenital heart disease based on multi-view synergy relationship. The steps of the present invention are as follows: 1. Medical ultrasound data enhancement and data preprocessing to obtain the medical image to be detected; 2. Input the multi-frame ultrasound images with different viewing angles to the SSD detector trained by the convolutional neural network for accurate Positioning to obtain the precise positioning result of Top1; 3: The above multi-view focus image frame C i and color ultrasound original image frame O i Combine to build data group {C i ,O i}, where i represents the i-th sample group. 4: Send the data group into the MUVDN network for training and obtain the trained MUVDN binary classification network. The invention has higher robustness. It reduces the impact of artifacts and noise on diagnosis in a single view, and effectively improves the accuracy of network classification.

The embodiment of the invention provides a 5G-based remote internet big data smart medical system and a medical assistance method, and aims to provide richer focus images for doctors. The medical system comprises a mobile terminal and a server, a focus rehabilitation function is preset in the server; and the server is used for receiving a focus image and historical treatment time uploaded by the mobile terminal, calculating the treatment time length from the historical treatment time to the current time; extracting a feature vector of the focus image, and determining a target sample feature vector matched with the feature vector from a focus rehabilitation function, taking the treatment time corresponding to the target sample feature vector as a starting point, advancing the treatment timelength to determine a sample feature vector corresponding to the historical treatment time, thereby extracting a sample lesion image corresponding to the sample feature vector; and sending the focusimage, the historical treatment time and the sample focus image to a mobile terminal of the target doctor.

The invention discloses an X-ray chest radiograph pseudo focus image generation method and device and a storage medium. The method comprises the steps: acquiring a real focus image and a normal chest radiograph image; generating boundary suppression mask information according to lesion area information in the real lesion image; according to the lesion area information, operating the normal chest radiograph image to generate skeleton area inhibition mask information; performing mask operation on a lesion area image in the real lesion image by using the boundary inhibition mask information and the skeleton area inhibition mask information to obtain a fused lesion image; and superposing the fused lesion image and the normal chest radiograph image to obtain a pseudo lesion image. Through the technical scheme of the embodiment, more training data can be provided for deep learning, and the robustness and accuracy of a deep learning network are improved.

The invention relates to a machine learning technology, and discloses an image recognition method based on a CT sequence, and the method comprises the steps: obtaining a tissue image sequence and a focus image sequence of a target pathological tissue; performing feature extraction on the tissue image sequence and the focus image sequence, and splicing the first feature image set and the second feature image set into a focus feature map; generating a prediction image label of the focus feature map; calculating a loss value between the prediction image label and a preset target pathological label, and updating the feature extraction model according to the loss value to obtain a target image recognition model; and performing image recognition on the to-be-recognized image sequence by using the target image recognition model to obtain a recognition result. The invention further provides an image recognition device based on the CT sequence and a computer readable storage medium. In addition, the invention also relates to a blockchain technology, and the to-be-identified image sequence can be stored in the blockchain node. The method and device can be applied to medical image recognition. According to the invention, the image identification efficiency and accuracy can be improved.

The embodiment of the invention provides an image processing method and device, computer equipment, a storage medium and a computer program product. According to the method, an image sequence including a target body part is received, a reference image corresponding to the image sequence is determined based on prior knowledge, further, according to position information and size information corresponding to an area range in the reference image, the image sequence is cut, and a focus image sequence is obtained; the lesion image is accurately cut from the image sequence by taking the reference image as a reference, the noise influence in the background image is reduced, and the lesion is completely cut from the image sequence.

The invention relates to a real-time detection and breast disease auxiliary analysis method based on a breast ultrasonic image. The method comprises the steps: collecting video data during breast ultrasonic examination, performing the preprocessing of the video data, inputting the preprocessed video data into a yolov5 network for training, and obtaining a real-time detection model; according to the obtained real-time detection model, screening out a breast lesion image for each case, performing preprocessing, and training a multi-view breast disease classification model based on an EfficientNetV2-L network; forming a real-time video by echoes of real-time breast disease detection, inputting the real-time video into the real-time detection model, and outputting a focus map with the confidence coefficient larger than a threshold value; for the suspicious focus map, performing focus size reminding, and if no focus exists, outputting a prediction report; and inputting the focus map into the breast disease classification model, predicting the type and confidence of the breast disease, and outputting related information. According to the method, doctors can be assisted in screening breast lesions, and the diagnosis efficiency and accuracy are improved.

The invention discloses an intelligent photodynamic therapy illumination system based on an LED array. The system comprises a digital image acquisition module, a computer with digital image processingand analyzing software, a single-chip microcomputer system and the LED array. The digital image acquisition tool performs image acquisition on a lesion area to obtain a lesion image with a scale. Thecomputer with the digital image processing and analyzing software is used for carrying out digital image processing on the acquired lesion image to obtain a light-emitting time length signal of eachLED. The single-chip microcomputer system controls the LED array to emit light according to the light-emitting time length signal of each LED. According to the system, the light-emitting area of a light source can be adjusted according to the shape and range of the lesion of a patient; the local illumination dose can be adjusted according to the lesion degree of the area corresponding to the lesion; and therefore the accuracy of the photodynamic therapy process is improved.