54 results about "Lesion type" patented technology

A non-invasive tool for skin disease diagnosis would be a useful clinical adjunct. The purpose of this study was to determine whether visible / near-infrared spectroscopy can be used to non-invasively characterize skin diseases. In-vivo visible- and near-infrared spectra (400-2500 nm) of skin neoplasms (actinic keratoses, basal cell carcinomata, banal common acquired melanocytic nevi, dysplastic melanocytic nevi, actinic lentigines and seborrheic keratoses) were collected by placing a fiber optic probe on the skin. Paired t-tests, repeated measures analysis of variance and linear discriminant analysis were used to determine whether significant spectral differences existed and whether spectra could be classified according to lesion type. Paired t-tests showed significant differences (p<0.05) between normal skin and skin lesions in several areas of the visible / near-infrared spectrum. In addition, significant differences were found between the lesion groups by analysis of variance. Linear discriminant analysis classified spectra from benign lesions compared to pre-malignant or malignant lesions with high accuracy. Visible / near-infrared spectroscopy is a promising non-invasive technique for the screening of skin diseases.

The invention discloses a cervical cell pathological slice automatic judging method with high and low resolution combination, which is characterized in that the method comprises the following steps: extracting a single cell mass region on a low-resolution cervical pathological slice image; extracting a single cell mass region from the low-resolution cervical pathological slice image; identifying suspicious abnormal cell clusters from low-resolution cell clusters; mapping suspected abnormal cell mass regions into high resolution slice images; semantically segmenting pathological cells from thehigh-resolution region of suspicious cell clusters and interpreting the type; according to the type, quantity and confidence of the segmented lesion cells, establishing the feature set of the slice, and then classifying the lesion type of the whole slice. The invention takes cell cluster as processing and recognition unit, utilizes classification neural network model to quickly identify suspiciousabnormal cell cluster on low-resolution digital slice, then divides out pathological cell on high-resolution suspicious area and judges its type, at the same time, improves recognition accuracy and recognition efficiency.

The present invention relates to a diabetic retinopathy sign detection method and device. The method comprises: receiving a fundus image to be detected; using a convolutional neural network (CNN) model to perform processing of the fundus image to be detected, and obtaining lesion area samples corresponding to the fundus image to be detected; according to the lesion area samples, constructing SIFT (Scale-Invariant Feature Transform) feature descriptors corresponding to the fundus image to be detected; and according to the SIFT feature descriptors, using an SVM (Support Vector Machine) classifier to determine the lesion type of the fundus image to be detected. The CNN model is employed to perform rough classification of the diabetic retinopathy to five out a lesion area, the SVM classifier is employed to perform fine classification of the lesion type of the lesion area to reduce interference generated by difference between the data size and the data when the neural network is directly configured to perform lesion classification and improve the classification precision.

The invention relates to a robotic laser beautifying and treatment system. The robotic laser beautifying and treatment system comprises an intelligent vision system, a man-machine interaction system and a tail end laser execution system, wherein the intelligent vision system is used for acquiring and processing skin images; the man-machine interaction system is used for operating the images by a user; and the tail end laser execution system is used for emitting laser in a designated area according to operation of the user and image processing results. The robotic laser beautifying and treatment system can provide an optimal treatment strategy, including laser parameters, irradiation time, and the like according to a skin type and lesion type given by a doctor through combination of a skinsurface vision technique by using an artificial intelligence method. According to the treatment strategy, the laser parameters are accurately controlled by adopting the mechanical operation, so that the influence of the laser on non-focus points is reduced; and the positioning accuracy of a focus point is improved by utilizing the robot-assisted treatment mode, and the working efficiency of doctors is improved.

The invention discloses a thyroid cytology multi-type cell detection method based on deep learning, and the method comprises the steps: obtaining a cell pathology slide image of a thyroid cell, obtaining a plurality of target images from the cell pathology slide image through sliding window scanning with selectable step length, the adjacent target images have an overlapping area, the target images are processed to obtain at least one feature image and confidence images, and the number of the confidence images is the same as the number of lesion types of lesion cells of the thyroid; and processing the at least one confidence coefficient image to obtain a plurality of confidence coefficients corresponding to different lesion types, obtaining a classification result matched with the target image based on the plurality of confidence coefficients, and processing the at least one confidence coefficient image based on the classification result to obtain an area matched with the classification result in the target image. Therefore, the detection efficiency of the detection system can be improved, and the marking cost can be reduced when the detection system is trained.

The invention discloses a slide scanning image acquisition and analysis method. The method comprises the steps: acquiring and marking a slide sample; performing mark recognition and preview on the slide sample to obtain a preview image; identifying the preview image to obtain a scanning area of the preview image; sequentially scanning each visual field area of the scanning area to obtain a difference image of the visual field areas; calculating the out-of-focus distance value of the difference image through a convolutional neural network model, and determining the optimal focal plane positionof each visual field area; sequentially moving each visual field area to the optimal focal plane position and scanning to obtain a microscopic image under each visual field, and splicing the microscopic images into a digital microscopic image; and marking a suspicious lesion cell region in the digital microscopic image and marking a possible lesion type through a cell identification classificationnetwork. The method provided by the invention does not completely depend on the experience and business level of a single doctor and does not completely depend on the recognition precision of the recognition algorithm, and is high in speed and high in precision.

The invention provides a liver space-occupying lesion identification method and device and an implementation device, which relate to the technical field of medical images. The method comprises the steps of acquiring a first CT image; recognizing the first CT image of each phase by a first recognition model, and recognizing the suspected focus position of the first CT image of each phase; carryingout the lateral and longitudinal combination of suspect lesion positions to obtain a detection frame position, and cutting that first CT image of each phase longitudinally according to the detection frame position to obtain a second CT image; identifying the lesion type of the suspected lesion position in the second CT image through the second recognition model, and judging the target lesion position from the suspected lesion position; segmenting the first CT image by the third recognition model, and obtaining the third CT image; and according to the detection frame on the third CT image, obtaining the position coordinates of the target focus. The lateral features between the four-phase images and the longitudinal features of the single-phase images are comprehensively considered, which greatly improves the accuracy of the recognition results.

The invention provides an intelligent identification method of colposcope images. The method is characterized by comprising the steps that the colposcope images are collected, and suspected lesion areas are selected from the colposcope images; image features of the suspected lesion areas are extracted; multiple standard colposcope images similar to the image features of the suspected lesion areas are searched in a standard colposcope image library; meanwhile, a confirmed lesion type and lesion area of each colposcope image are identified; the selected images of the suspected lesion areas and images of standard colposcope image confirmed lesion areas are displayed in a compared mode in a same screen. According to an intelligent identification of the colposcope images, feature extraction is conducted by only selecting the images of the suspected lesion areas, and image identification is more targeted and higher in speed. Comparison with the standard colposcope images in the same screen is conducted by means of a computer aided diagnosis technology, and the conformance of diagnoses of different doctors on the same image can be effectively improved; an intelligent identification device can also serve as a colposcope learning and training tool to assist in improving an overall cognitive level of the colposcope images in medical profession.

The invention discloses a chest X-ray disease detection device and method based on a dual-channel separation network, and the device comprises a processor which is configured to carry out the following operations: carrying out the preprocessing of a training data set of a chest X-ray image, and dividing the training data set into a data enhancement part and a normalization regularization part; training a dual-channel separation depth network, extracting and fusing features of each level in different channels, minimizing a loss function in a classification layer, and performing complete training of the network; classifying the input chest X-ray images by using the trained network to obtain lesion types and probabilities contained in the images; and using the trained network to locate the disease in the input chest X-ray image. The method can greatly improve the recognition accuracy of chest lesions in a chest disease recognition task, and can achieve the precise positioning of a diseaseposition in a visualization task.

The invention discloses a breast rheid model applied to mammary gland impedance scanning detection technique training, and a preparation method of the model. The model mainly consists of a conductive rubber film layer, sepharose gel and an electrolyte solution, wherein the sepharose gel and the electrolyte solution are wrapped by the conductive rubber film layer in a sealed manner so as to form an overall hemispherical breast rheid model. The model is capable of vividly simulating form and impedance distribution state of a breast and has the characteristics of high form simulation degree, approximation to real impedance parameters, simplicity in preparation process and the like, different model phantoms can be prepared according to structural types and lesion types of the breast, and meanwhile the model can be repeatedly used for a long time and is beneficial for long-term operation training.

The embodiment of the invention discloses a medical image recognition method and device, electronic equipment and a storage medium. The method comprises the following steps: firstly, segmenting a tissue region of a target part in the medical image; obtaining a plurality of candidate tissue regions and tissue types corresponding to the candidate tissue regions; the method comprises the steps of obtaining a plurality of candidate tissue areas, obtaining associated tissue types of preset lesion types, selecting target tissue areas of which the tissue types belong to the associated tissue types from the plurality of candidate tissue areas, and finally performing preset lesion type recognition on the target tissue areas to obtain a preset lesion type recognition result. The scheme is based on acomputer vision technology; when the lesion type of the medical image is identified, the medical image is segmented into a plurality of candidate tissue areas, the target tissue area corresponding tothe preset lesion type is selected for preset lesion type recognition, the recognition process does not need manual participation, and compared with the prior art, the dependence of lesion type recognition of the medical image on manual work is reduced.

The invention provides an AMD grading system based on a macular attention mechanism and uncertainty, and relates to the technical field of macular disease detection. An optic disc segmentation image and a blood vessel segmentation image are obtained by using a segmentation network model, a macular area image is obtained according to the obtained optic disc segmentation image and the blood vessel segmentation image, a multichannel image is obtained by using an attention network sub-module, features are extracted by using a Bayesian deep learning classification network model, and through multiple times of dropout Monte Carlo, four groups of probability values and one group of noise corresponding to the four lesion types are output; and the classification network module gives out accidental uncertainty and model uncertainty while finally outputting a model classification result. And the safety performance of the model is ensured.

The invention provides a visible-light-terahertz-light-based plant health identification method and device. The method comprises: S1, acquiring a terahertz moisture absorption grayscale image and a visible-light Lab image A channel grayscale image of a complete leaf of a plant; S2, determining whether the terahertz moisture absorption grayscale image and the visible-light Lab image A channel grayscale image are in a normal grayscale threshold range by using an adaptive threshold segmentation method and determining whether the leaf is in lesion state based on a corresponding relationship between leaf imaging results and plant lesions; and S3, collecting a terahertz absorption coefficient map of a diseased leaf virus spore and determining a specific lesion type of the plant. According to the invention, on the basis of combination of the special moisture absorption characteristic of the terahertz spectrum, the association characteristic of the plant leaf moisture and the plant health, and the characteristic of association between the plant leaf color and the plant health, the terahertz moisture absorption grayscale image and the visible-light Lab image A channel grayscale image are obtained and whether the two images are in a lesion state is determined, thereby finding out the lesion early.

The invention relates to the field of image recognition of computer vision and the field of artificial intelligence, and provides a common colon disease classification method based on a deep neural network and an auxiliary system, and the auxiliary system comprises a model training module and an auxiliary diagnosis module. The method comprises the steps of obtaining a multi-classification data setand a binary classification data set; training based on the binary classification data set to obtain a normal abnormal deep neural network classification model; training based on the multi-classification data set to obtain a lesion type deep neural network classification model; fusing the two models to obtain a deep neural network disease classification model of the colonoscope image, in the model, firstly inputting the image in the test set into a normal abnormal deep neural network classification model, if the image is judged to be normal, directly outputting a judgment result, and if the image is judged to be abnormal, inputting into a lesion type deep neural network classification model, further determining the lesion type, and outputting a judgment result.

The invention relates to the technical field of medical image analysis, in particular to a lung pathological image classification and segmentation method. The method comprises the following steps: inputting a pathological image of a full slice; segmenting the pathological image by using a sliding window to obtain image blocks; analyzing foreground image blocks in sequence by using a focus type classification model, and identifying focus types of tissue regions in the foreground image blocks; outputting a focus type classification result; and splicing the foreground image block lesion region segmentation results according to the relative positions of the corresponding foreground image blocks in the pathological image, and filling a background image block region with the background to obtainpathological image lesion region segmentation results. According to the invention, the specific boundary of the lesion area is segmented while accurate lesion type identification is carried out on the lung pathological image.

The invention discloses an eye fundus image recognition method and device, equipment, a storage medium and a program product, and relates to the technical field of artificial intelligence such as computer vision, deep learning and intelligent medical treatment. A specific embodiment of the method comprises the following steps: acquiring a position of a fundus focus, a position of a macular fovea center, a lesion level of a retina and a lesion probability value of a macular region in a fundus image; establishing the correlation between each feature and the lesion type of the macular region based on the position of the fundus focus, the position of the fovea center of the macular, the lesion level of the retina and the lesion probability value of the macular region; performing feature screening based on the correlation between each feature and the lesion type of the macular region; and inputting the screened features into a pre-trained macular region classification decision tree to obtain the category of the macular region. According to the embodiment, computer-assisted eye fundus image recognition is utilized, so that the labor cost is greatly reduced.

The invention discloses a mammary gland MRI lesion area detection method based on multi-dimensional information fusion, and the method comprises the steps: carrying out the feature learning of a preset image through employing a convolutional neural network, and enabling the convolutional neural network to learn the difference between a lesion area and a normal tissue in the preset image; Acquiringa mammary gland MRI image to be detected, wherein the mammary gland MRI image comprises a same-period continuous tomography image and a same-fault different-period image; According to the trained convolutional neural network, lesion area selection is carried out on the continuous tomography images of the same period, and more than one candidate window is obtained; Encoding the images of differentperiods of the same fault by adopting a recurrent neural network to obtain relevant information between the signal intensity and the lesion type of the images of different periods of the same fault;And mapping more than one candidate window into the related information, and classifying through a preset classifier to obtain a final lesion area detection result. According to the technical scheme provided by the invention, the lesion area in the mammary gland can be more accurately detected.

The invention provides a fundus lesion recognition method and device. The method comprises the following steps: acquiring an OCT fundus image of a specific patient; determining boundary information of a retina tissue of the OCT eye fundus image by adopting a first neural network model, and segmenting the OCT eye fundus image into a plurality of image blocks according to the boundary information of the retina tissue; the plurality of image blocks are provided for a second neural network model to obtain the fundus lesion type of the specific patient, and the second neural network model calculates the feature vector of each image block and the attention weight of each image block for the set fundus lesion type; and performing weighted calculation to obtain a fused feature vector so as to obtain an identification result aiming at a set fundus lesion type. Wherein the second neural network model reinforces important information in the image block group in combination with the spatial domain significance and the scanning position significance, and filters unimportant information, so that low-cost model training and application are realized.

The invention relates to the technical field of medical image processing, in particular to an artificial intelligence system for identifying high myopia retinopathy, comprising: an image acquisition module for acquiring an optical coherence tomography image and a fundus reference of a patient to be identified; the first recognition module is used for inputting the optical coherence tomography image into a high myopia retinopathy recognition model, judging whether the optical coherence tomography image has lesion or not and obtaining a lesion type result; the second recognition module is used for inputting the fundus color photograph into a high myopia retinal lesion staging model, judging the retinal lesion staging of the fundus color photograph and obtaining a lesion staging judgment result; and the report generation module is used for generating a diagnosis and treatment suggestion report according to the lesion type result and the lesion staging judgment result. According to the method, the optical coherence tomography image and the fundus color photograph of the patient can be quickly and effectively recognized to recognize common retinopathy of high myopia.

The invention discloses an eyelid tumor digital pathological section image multi-classification method based on deep learning. The method comprises steps of scanning eyelid tumor pathological sections classified by known lesion categories to obtain an image construction training set; performing data enhancement and normalization processing, constructing a three-layer cascaded tumor digital pathological section diagnosis network, training the tumor digital pathological section diagnosis network by using an enhanced training set, performing prediction processing, generating a probability thermodynamic diagram, and performing lesion category detection. The method can effectively visualize the position and lesion type of the tumor in the full-field digital slice to assist in diagnosis, perform preliminary screening and prompting of lesion areas, change the attention of the network to channels and improve the network performance.

The invention discloses an automatic interpretation system for a cell pathology smear. The system comprises an imaging module; an image acquisition module; an image storage and management module; an image preprocessing module; the intelligent interpretation module receives the image output by the image preprocessing module, predicts and classifies a normal sample and a lesion sample for the image, and further predicts and interprets the lesion type when the image is the lesion sample; and the report writing module automatically generates an interpretation conclusion text of the sample corresponding to the image. According to the method, the artificial intelligence auxiliary diagnosis technology is successfully applied to the rapid staining cytopathology, the accuracy and consistency of diagnosis can be remarkably improved, and the workload of cytopathology doctors is reduced; according to the method, the existing convolutional neural network is improved, the multi-channel attention mechanism characteristics are utilized, uncertain factors introduced by manual sampling are solved, high-accuracy full-scene and multi-classification tasks can be achieved, and finally the accuracy of classification interpretation results can be improved.

The invention provides a lesion classification method and device, and a training method of a lesion classification model. The lesion classification method comprises the steps that a first lesion areain a first medical image and a second lesion area in a second medical image are acquired, wherein the first medical image and the second medical image are images obtained by shooting an examination part of the same examination object at different angles, and the first lesion area and the second lesion area are image areas corresponding to the same lesion in the first medical image and the second medical image respectively; feature extraction is respectively performed on the first lesion area and the second lesion area; the features of the first lesion area and the features of the second lesionarea are fused to obtain a first fused feature of the lesion; and according to the first fusion feature, the lesions are classified. Therefore, the resolution capability of the lesion types can be improved, and the lesion classification accuracy is improved.

The invention provides a premature infant retinopathy plus lesion classification system based on uncertainty, and relates to the technical field of deep learning. The method comprises the following steps: via a vascular segmentation module, firstly, segmenting a blood vessel of a fundus image; using a Bayesian deep learning classification network model to extract features via a classification module; performing dropout Monte Carlo for multiple times; obtaining three groups of probability values corresponding to three lesion types and a group of image noise, calculating the mean value and variance of each group of probability values, taking the lesion type with the maximum mean value as a final classification result, and taking the mean value of the image noise as accidental uncertainty andvariance sum as model uncertainty. When the method is actually put into use, the credibility of the image classification result can be judged through two kinds of uncertainty instead of selecting a diagnosis result given by a blind belief model; for doctors and patients, whether artificial ophthalmology experts are needed for re-diagnosis or not is considered to be very helpful, and the method issafer and more reliable in actual clinical use.

The invention discloses a crop lesion detection method and device. The detection method comprises a lesion classification step and a lesion recognition step. In the lesion classification step, training of a decision tree classifier is completed according to lesion type labels and image features of all sample images; and a plurality of feature template images are generated according to the decisiontree classifier; the lesion identification step comprises the step of calculating the matching degree between the to-be-detected image and each feature template image by using a semi-naive Bayes classifier; and judging possibility of lesion of to-be-detected crop according to the matching degree between the to-be-detected image and each feature template image. According to the lesion detection method, in the lesion classification step, a decision tree classifier is used for determining crop characteristics corresponding to various different lesion types, so that a template image is generated;in the lesion identification step, the matching degree of the to-be-detected image of the to-be-detected crop and each template image is calculated by using the semi-naive Bayes classifier, so that the lesion type of the to-be-detected crop and the possibility data of the lesion are obtained, and the reliability is high.

The invention discloses a focus fusion method and device, and the method comprises the steps: obtaining a focus set of a breast image; executing at least one focus fusion operation until any two focuses in the focus set cannot be fused into one focus; wherein the step of executing the lesion fusion operation comprises the following steps: determining a first lesion and a second lesion from a lesion set; wherein the first lesion and the second lesion are different types of lesions to be subjected to lesion fusion operation, and the first lesion and the second lesion are lesions which are located in a lesion set according to a preset fusion sequence and are subjected to lesion fusion operation firstly; determining whether the first lesion and the second lesion are fused into a third lesion or not according to the overlapping condition of the area of the first lesion and the area of the second lesion; and if so, adding the third lesion into the lesion set, and deleting the first lesion and the second lesion from the lesion set. According to the technical scheme, the same lesion with various lesion types can be accurately identified, and the working efficiency of doctors is improved.

The invention discloses a method for recognizing a lesion area in a thyroid follicular tumor tissue pathological image, and predicts gene mutation. The method is an automatic auxiliary diagnosis technology based on a deep learning method, a lesion area in a thyroid follicular tumor pathological tissue slice image is automatically positioned by using big data and a deep convolutional neural network algorithm, and a pathological histological type and a gene mutation type of the lesion area are automatically recognized. According to the pathological tissue image, recognizing cases simultaneously carrying RAS and other driver gene mutations, and realizing histological classification of the follicular thyroid tumor and prediction of related gene mutations. According to the method, information is provided for clinicians, pathological diagnosis and clinical decision making are assisted, and development of digital pathology and precise medical treatment is promoted.

The invention discloses a multi-feature fusion image classification method based on deep learning. The method specifically comprises the steps of data set division, data enhancement, classification network model construction, model initialization and model training optimization. And the data enhancement part is used for enhancing data characteristics by randomly carrying out operations such as horizontal overturning, vertical overturning, brightness modification and horizontal overturning according to probability on the picture. In the construction process of the classification network model,the features extracted for the first time are randomly covered and then extracted again, and then the features extracted for the two times are fused, so that the features are diversified, and the classification accuracy is improved. The system can be used for classifying eye malignant tumor images, positioning lesion areas in the images as feature areas, giving out probability values of lesion types and assisting film reading doctors in judgment.

The invention provides an image recognition method, a readable storage medium and electronic equipment, and the method comprises the steps: directly obtaining the lesion type information of a lesion part through a constructed lesion classification model, generating a thermodynamic diagram of a lesion image based on the lesion type information, and determining the specific lesion position information, outputting lesion prompt information according to the lesion type information and the lesion position information; or obtaining the texture and color of the skin of the lesion area through the constructed lesion color classification model and lesion texture classification model, then determining lesion type information through the texture and color of the skin of the lesion area, and generating a thermodynamic diagram of the lesion image based on the lesion type information to determine specific lesion position information; and outputting lesion prompt information according to the lesion type information and the lesion position information. The lesion prompt information can be used as a preliminary judgment of skin disease problems in animal feeding, so that a feeder can know the health condition of the animal at any time.

The invention discloses a premature infant retina plus lesion detection method and system, and the method comprises the steps: firstly inputting a to-be-detected retina posterior pole image into an image quality detection model for quality evaluation, and inputting a qualified image into a blood vessel recognition model to obtain a blood vessel graph; inputting the blood vessel graph into a lesiondetection model, and outputting the probability of the corresponding lesion types of plus, pre-plus and no-plus; and converting the probability of the lesion type of the lesion examination into a corresponding i-ROP lesion score according to a preset score conversion equation to reflect the continuity of the retinopathy degree. According to the invention, the severity of plus lesion is objectively measured, and the severity is tracked over time to provide objective assessment of disease progression or regression; by analyzing the change of the score along with the time, the eyes which progress into plus lesion ROP can be recognized in advance under most conditions for early intervention, disease delay caused by insufficient diagnosis is prevented, the accuracy of lesion degree detection is high, and the ophthalmoscope examination frequency of doctors can be greatly reduced.

The embodiment of the invention discloses an image processing method and device, computer equipment and a storage medium. According to the embodiment of the invention, the medical image of the target object is identified through the lesion identification model, the lesion area of the medical image is determined, then the initial sign information of the lesion area is extracted, the sign information of the lesion area is complemented according to the initial sign information, and the completed sign information is obtained. The lesion type of the lesion area is further determined through the complete sign information, and the accuracy of an image processing result can be improved.