102 results about "Lesion feature" patented technology

The invention discloses a CT image analysis method and system based on a neural network, and relates to the technical field of CT image analys.The method comprises the steps that an original CT image is obtained after user authorization, a Laplace operator is adopted to strengthen a focus boundary, and a circular region of interest is intercepted to remove edge sensitive information; extracting edge and texture information in the standardized image; focus area features are focused step by step; executing characteristic distillation balance based on category sample distribution, and outputting a focus characteristic graph with local perception enhancement and sample balance characteristics; segmenting the lesion feature map into serialized units, embedding position codes, inputting the serialized units into a plurality of layers of encoders, and fusing an image structure and text indication information through a dynamic adjustment mechanism; performing linear classification on the global semantic vector to output a diagnosis result, generating a focus thermodynamic diagram, and superposing the focus thermodynamic diagram to an original image for visualization; and performing dynamic optimization based on doctor feedback. The accuracy of feature analysis is improved; the overall operation efficiency of the system is improved.

The invention relates to the technical field of medical images, in particular to a medical image intelligent detection and auxiliary diagnosis system based on deep learning, and the system comprises a data access module which is used for obtaining medical image data and clinical text data of a patient; the data fusion module is used for generating a focus feature vector and a text feature vector, and performing cross-modal alignment and fusion to generate a fusion feature vector; the diagnosis analysis module is used for executing focus detection, focus segmentation and focus classification tasks, generating a diagnosis result and generating a diagnosis label based on the diagnosis result; the decision generation module is used for mapping the generated execution result to a preset medical knowledge base and performing deep reasoning to generate an auxiliary diagnosis decision; the decision auditing module is used for performing confidence scoring on the auxiliary diagnosis decisions and selecting the auxiliary diagnosis decision with the highest confidence score as the final auxiliary decision; and the data visualization module is used for carrying out visualization processing on the auxiliary decision and the diagnosis result.

The invention relates to a diabetic retinopathy fundus photography grading report system combined with a clinical guide, based on an artificial intelligence deep learning technology, and belongs to the field of fundus lesion analysis. The system accurately identifies and segments various lesions such as microhemangioma, bleeding, exudation and the like and symbolic structures such as optic discs, macular regions and the like by automatically analyzing fundus photographic images. The system adopts ICDR international standards to grade diabetic retinopathy, and provides diagnosis and treatment suggestions for lesion characteristics in combination with clinical guidelines of American ophthalmology institute in 2019. Through big data training, the system can generate detailed reports in real time, the early screening rate is remarkably improved, misdiagnosis and missed diagnosis are reduced, and the diagnosis speed and accuracy are improved. The system comprises a plurality of modules, such as an image pre-classification module, a deep learning focus recognition module, an ICDR grading module and a report generation module, efficient and accurate diabetic retinopathy diagnosis and grading are cooperatively achieved, and the clinical management level is improved.

The invention discloses a thyroid ultrasound diagnosis report intelligent generation method based on a multi-mode large language model, and relates to a thyroid ultrasound diagnosis report intelligent generation method. The objective of the invention is to solve the problems of lack of term standardization and insufficient complex focus feature analysis in the prior art. According to the method, a full-flow technical system of double-flow coding, cross-modal alignment, dynamic man-machine cooperation and multi-dimensional evaluation is constructed. Multi-scale feature fusion of a thyroid global form and a nodule ROI region is realized through ResNet-50 and ConvNeXt double-flow coding networks, image-text semantic alignment is optimized by adopting a CLIP symmetry loss function, and training resource consumption is reduced in combination with an LoRA parameter fine tuning technology. A dynamic man-machine collaborative closed-loop mechanism is innovatively introduced, model parameters are iteratively optimized through doctor correction data, and a four-dimensional clinical evaluation system comprising ROUGE-L, BLEU-4, CIDEr and expert blind evaluation is established. The invention belongs to the technical field of medical artificial intelligence auxiliary diagnosis.

The invention relates to the field of fruit tree disease and insect pest detection, in particular to a fruit tree disease and insect pest detection method and system based on machine vision, and the method comprises the steps: obtaining an image metabolome feature matrix and a preliminary difference pixel based on a preprocessed multispectral image of an original machine vision image obtained by a camera, carrying out the topological skeleton extraction, and carrying out the dimension fusion; outputting a core focus feature set; each discrete feature is used as a network node, a mutual information value between any two discrete features is calculated, and a focus area is obtained; and based on a lesion region containing lesion boundary coordinates, area and morphological parameters, extracting the ROI of the lesion region from the multispectral image, and carrying out disease and pest identification matching to obtain a detection result. According to the method, the essential attributes of the lesion are comprehensively captured by fusing the multi-dimensional features of the spectrum, the texture, the space coordinates and the morphological topology, a multi-dimensional fusion feature system is formed, and the problems that similar pest and disease damage forms are difficult to distinguish, and early lesion features are weak and difficult to recognize are effectively solved.

The invention provides a pneumonia medical data analysis and processing method, system and device and a storage medium, and belongs to the field of medical image processing. Multi-scale features of the lung CT image are extracted layer by layer through convolution, global pooling operation down-sampling is carried out based on the multi-scale features, and saliency features of lung lobe and lesion areas are captured from local details to global semantics; performing convolution and multi-scale pooling on the saliency features to generate a feature map fusing global and local information; and recovering the feature map through layer-by-layer up-sampling to obtain a segmentation result of the lung lobe and the focus. Quantitative indicators are calculated based on regions and pixels of the segmentation results of the lung lobes and lesions. And performing model processing based on the quantitative index, the segmentation result of the lung lobe and the focus and clinical data of the lung lobe and the patient to obtain pneumonia focus characteristic data. And a doctor can conveniently make an objective and reasonable treatment scheme according to a result of lesion segmentation and feature data calculation based on the pneumonia lesion feature data.

The invention discloses a diabetic retinopathy image classification method based on a multi-feature fusion network model, and the method comprises the steps: S1, obtaining an initial fundus image data set, and constructing a sample training set based on the initial fundus image data set, a first enhanced image data set, and a second enhanced image data set; s2, constructing a retinopathy image classification model, and training the retinopathy image classification model based on the sample training set to obtain a trained retinopathy image classification model; the retinopathy image classification model comprises a first multi-feature fusion enhancement module, a second multi-feature fusion enhancement module, a third multi-feature fusion enhancement module, a fourth multi-feature fusion enhancement module and a classification module; the classification module performs classification based on the input data to obtain a retinopathy image classification result. By designing a plurality of multi-feature fusion enhancement modules and double attention modules, hierarchical fusion of lesion features is realized, the recognition precision of tiny lesions is improved, dynamic calibration of a lesion area feature map is realized, and finally, high-precision and robust DR automatic classification is realized.

The invention provides an early-stage intelligent diagnosis method for rheumatoid arthritis based on a multi-modal medical image and deep learning, and the method comprises the steps: carrying out the image cleaning and standardization, automatic bone joint positioning and segmentation, and quality control of a multi-modal original data set, and obtaining a joint slice library; performing intra-modal self-supervised pre-training and cross-modal alignment representation, and then performing focus level detection and quantification to obtain a lesion feature vector of each joint; performing joint diagram construction and quality perception multi-modal fusion to obtain fusion feature representation, and performing weak supervision multi-instance learning and multi-task loss calculation based on the fusion feature representation to obtain an uncalibrated model; and carrying out model calibration and explainable output to obtain an intelligent diagnosis model. According to the method, the diagnosis time point can be advanced to the reversible inflammation stage, interpretable evidence of patient-level decision and joint-level quantification is provided, the engineering feasibility of small samples and multi-center generalization is considered, and the method has high clinical transformation potential.

The invention relates to the technical field of oral cavity digital image processing, and provides an implant perimucosa red and swollen area detection method based on deep learning, and the method comprises the steps: S1, collecting three-dimensional model data of an implant site of a patient, and exporting a standardized visual angle rendering screenshot with a visual enhancement effect by using a matched software rendering function; s2, constructing a deep convolutional neural network based on a YOLOv8 architecture, training a model by adopting a transfer learning strategy, and realizing automatic extraction of implant perimucosa red and swollen focus features; s3, through a feature fusion module in the deep convolutional neural network, automatically retrieving suspected red and swollen sites on the feature maps with different resolutions, performing coordinate correction on the candidate region, and generating an accurate detection frame; and S4, automatically executing batch prediction on the test set based on the trained model, and outputting a detection frame and a quantitative index. And automatic identification and spatial positioning of the red and swollen mucosa area around the implant are realized, so that a visual basis is provided for clinical precise probing and diagnosis and remote early warning.

The application provides an ophthalmic image diagnosis method and system based on multi-modal imaging cooperation, and relates to the technical field of medical image diagnosis. First, the OCT, fundus camera and ultrasonic original image data of an ophthalmic examination object are acquired, and multi-modal dynamic correlation mapping results are obtained through dynamic correlation and trend correlation processing. Cross-modal lesion feature progressive mining and interactive verification are performed to obtain a cross-modal lesion correlation feature set. Multi-modal cooperative diagnosis reasoning and weight feedback optimization are used to generate an ophthalmic disease reasoning result containing disease types, lesion dynamic distribution and reasoning confidence. Finally, an ophthalmic diagnosis report with dynamic labeling and confidence explanation is generated. The application comprehensively utilizes the advantages of various image technologies to improve the accuracy of ophthalmic image diagnosis.

The invention relates to the technical field of medical image analysis, in particular to a dynamic DR digestive tract radiography automatic tracking and focus marking system which comprises a dynamic DR imaging module, an image real-time processing module, a contrast agent motion tracking module and a focus intelligent marking module. A deep learning algorithm is utilized to segment a contrast agent flowing area in real time and construct a dynamic three-dimensional model of a digestive tract, a contrast agent movement track is analyzed and tracked in combination with time-space domain features, and meanwhile, based on morphological anomaly detection and comparison with a focus feature library, focus positions such as stenosis, ulcers and space-occupying lesions are automatically marked; the problems of image blurring and tracking offset caused by alimentary canal peristalsis are solved by adopting a self-adaptive registration optimization algorithm of dynamic DR images and motion characteristics of alimentary canal contrast agents; by means of the method, the problems that traditional digestive tract radiography depends on manual interpretation of dynamic DR images, the focus tracking efficiency is low, and marking is prone to omission are solved.

The present application relates to a kind of breast lesion three-dimensional reconstruction methods, comprising the following steps: step S1, breast region self-rough scanning is carried out, and the depth image of breast region is acquired, and the three-dimensional model of breast and spatial position are determined;Step S2, fine scanning is carried out to lesion, and AI algorithm is used to realize the segmentation extraction of lesion feature in ultrasonic image;Step S3, with time, ultrasonic rotation angle or ultrasonic moving distance as interval, the image and position information of lesion feature during fine scanning are sequentially recorded;Step S4, the three-dimensional reconstruction of lesion is realized by three-dimensional reconstruction algorithm based on lesion boundary point cloud feature;Step S5, according to the recorded center position of lesion feature, the three-dimensional spatial position of lesion is calculated by weighted average method;Step S6, by breast center coordinates and lesion center coordinates, the relative position relationship of both based on mechanical arm coordinates is determined, and breast lesion three-dimensional reconstruction method is realized.The method can help doctor to determine the shape, size and its position in breast of lesion.

The present application relates to the technical field of lesion recognition, in particular to an automatic lesion recognition ultrasonic system for real-time monitoring, which comprises a delay path discrimination module, a gray level trend detection module, a texture feature screening module, a feature fusion sorting module and a region highlight labeling module. Based on continuous ultrasonic frames, the collected deep tissue echo path is analyzed, and the echo arrival time of each pixel point in the continuous frame is detected. The present application supports multi-type data fusion judgment through a comprehensive judgment process supported by multi-dimensional parameter collaborative screening, penetration behavior, gray level trend and texture aggregation. The regional abnormal priority sorting mode improves the hierarchy of lesion feature discrimination, provides partition directional recognition for local structural abnormalities and early micro-variation, and converts the feature judgment result into a high confidence region label by a weight aggregation method. The image output process automatically completes the real-time visual presentation of the high-risk area, improving the clarity and pertinence of the lesion region presentation.

The invention belongs to the technical field of medical image analysis, and provides a multi-feature fusion endoscope image quality online evaluation method and system. The method comprises the following steps: extracting stable anatomical regions such as a gastric horn through a U-Net semantic segmentation model, completing image confidence analysis and judging a trigger confidence signal; if triggering, performing gastral cavity inflation state analysis to obtain an image stretching coefficient; positioning a target focus, and extracting morphological regularity and texture fluctuation values to construct a current morphological parameter set; and performing deformation attribution analysis to obtain a form and texture residual error value, constructing an inflation-deformation mapping model after triggering a mapping signal, and dynamically correcting a current parameter to obtain a standard form parameter set. The system comprises a credible analysis module, a state recognition module, a two-dimensional analysis module, a mapping judgment module and a model construction module. Interference of the inflation state on image lesion features is eliminated, and image quality and lesion feature accuracy are improved.

The invention relates to the technical field of image classification, and discloses a myocardial ring slice prediction method and system, a terminal and a storage medium, and the method comprises the steps: carrying out the global average pooling processing of a myocardial ring slice through an extrusion-excitation model, outputting a channel scalar value, and carrying out the channel-feature matching, and outputting a channel-feature weight; performing enhancement or suppression processing on the channel-feature weight according to the feature type of the myocardial ring slice, and then performing feature fusion to obtain a feature fusion image; and performing enhancement processing on the feature fusion image to obtain an enhanced data set, and obtaining a prediction classification result of the target object through a 3D residual network. Based on the single data of the heart image, the extrusion-excitation model adapts to the weights of different feature channels, the ability of capturing the complicated three-dimensional structure and lesion features of the cardiac muscle is improved in combination with the optimized 3D residual network, and efficient feature extraction and accurate image classification are achieved.

The application discloses a liver cancer early screening method combined with dynamic enhancement MRI sequence, and relates to the technical field of image analysis, and the method comprises the following steps: acquiring an MRI multi-sequence image, dividing the MRI multi-sequence image into multiple connected domains; based on the gradient direction and the gray distribution of each connected domain, the comprehensive suspected lesion characteristics of each connected domain are calculated; based on the comprehensive suspected lesion characteristics and the abnormal fluctuation level of the connected domain, the abnormal degree characteristic value of each connected domain is calculated; the dynamic development trend of the lesion in the time sequence change sequence formed by the abnormal degree characteristic value is analyzed, and the comprehensive judgment characteristics of each connected domain are calculated; based on the comprehensive judgment characteristics, the MRI multi-sequence image is enhanced, the enhancement parameter map is obtained, and the preset prediction model is iteratively trained based on the enhancement parameter map, so that the trained model can predict and process the image to be processed. The application realizes early and accurate screening of liver cancer.

The invention provides a cardiovascular image processing system and method based on deep learning. The method comprises the following steps: acquiring a standardized cardiovascular image of a target patient; extracting a blood vessel edge texture feature and a lesion area texture feature of the target patient; determining a feature association relationship between each continuous image sub-block in the standardized cardiovascular image and a blood vessel edge texture feature, and determining a detail fusion feature map of the cardiovascular lesion area of the target patient according to all the feature association relationships and the lesion area texture features; determining a lesion trend characteristic spectrum of the target patient according to a historical cardiovascular image set similar to the cardiovascular image of the target patient and the cardiovascular image of the target patient; and determining a lesion labeling result of the cardiovascular image of the target patient according to the detail fusion feature map and the lesion trend feature map. According to the scheme, cross-time-sequence dynamic risk assessment can be carried out on the high-heterogeneity lesion features of the cardiovascular system of the patient based on the single-time static cardiovascular image.

This invention relates to the field of AI analysis technology for medical ultrasound images, specifically to an intelligent ultrasound image analysis method and system based on artificial intelligence. The method includes: acquiring ultrasound image sequences of a target patient and performing standardized preprocessing; extracting deep features through an anatomical structure recognition network to generate an anatomical structure description containing boundary contour coordinate sequences, tissue type classification identifiers, and spatial location depth information; using a lesion feature discrimination model to perform multi-dimensional feature fusion analysis on the three types of information to obtain preliminary lesion localization and attribute descriptions; then verifying spatiotemporal continuity based on adjacent frames of the image sequence to check the consistency of lesion evolution and generate a verification report; comparing the report with a medical knowledge base template to output an auxiliary diagnostic conclusion with diagnostic opinions and confidence ratings. This invention can enrich the dimensions of anatomical structure representation, improve the accuracy of lesion discrimination, ensure the temporal consistency of analysis results, and enhance the reliability of auxiliary diagnostic conclusions.

The present application belongs to the technical field of intelligent agriculture, and discloses a monitoring method for a fruit bottom colorizing and light reflecting film, comprising the following steps: S1, forming a grid map for an orchard; S2, collecting fruit bottom images to obtain array type multi-angle image data; S3, performing edge processing on the image data to form fruit images; S4, performing fruit surface color analysis on the fruit images to form color feature data; S5, performing edge and dispersion analysis on dark color patches in the color analysis process to form lesion feature data; S6, inputting the color feature data and the lesion feature data of S4 and S5 into a preset fruit maturity and lesion grade library to obtain preliminary maturity and sunscald grades; S7, performing feature weighting judgment to output maturity grades and / or sunscald lesion probabilities; S8, if the lesion probability is greater than a preset threshold, automatically recording a collection time stamp, a lesion probability value, an image slice and a map location, and triggering an early warning of adjusting the light reflecting film to weaken the radiant light; if the lesion probability is less than or equal to the preset threshold, normally recording data.

The application discloses an agricultural multi-source image low-delay transmission and intelligent processing method and system, and belongs to the technical field of image processing, and comprises the following steps: a double-branch parallel network is constructed; target data containing RGB images and near-infrared spectral data are synchronously collected through an industrial-grade device, the target data are input into the double-branch parallel network, visual feature vectors and spectral feature vectors are respectively extracted, and the visual feature vectors and the spectral feature vectors are input into a cross-modal attention layer; weights are dynamically calculated with the aid of a cross-modal attention mechanism, global semantic feature vectors fused with double-modal information are generated, and local attention graph elements pointing to key spectral bands of lesions are simultaneously output; an edge node-cloud two-stage transmission architecture is adopted, the global semantic feature vectors and the graph elements are preprocessed by the edge node and then transmitted to the cloud; a trained generative model is deployed on the cloud, and a visualized lesion feature map is reconstructed according to the received preprocessed data; and the method realizes efficient fusion of double-modal features and low-delay transmission, and improves the lesion feature recognition accuracy.

Embodiments described relate to a medical device including an invasive probe that, when inserted into an animal (e.g., a human or non-human animal, including a human or non-human mammal), may aid in diagnosing and / or treating a lesion of the animal (e.g., a growth or deposit within vasculature that fully or partially blocks the vasculature). The invasive probe may have one or more sensors to sense characteristics of the lesion, including by detecting one or more characteristics of tissues and / or biological materials of the lesion. The medical device may be configured to analyze the characteristics of a lesion and, based on the analysis, provide treatment recommendations to a clinician. Such treatment recommendations may include a manner in which to treat a lesion, such as which treatment to use to treat a lesion and / or a manner in which to use a treatment device.

The invention provides a breast lump segmentation method based on channel-guided double-pooling multi-scale space attention, and aims to solve the problems of small area, low contrast, fuzzy boundary and the like of breast lumps in an X-ray image, the network extracts full-view image features and compresses space dimensions through a deep residual encoder; and realizing fusion of low-level details and high-level semantic features in the decoder by means of jump connection. A double-pooling gating mechanism is built in the decoder, channel guide weights are generated in parallel through global average pooling and maximum pooling, lesion significant features are screened in a self-adaptive mode, and redundant backgrounds are restrained; the multi-scale space attention module captures multi-scale space information through multi-branch large-kernel separable convolution, generates a space attention graph, combines the space attention graph with channel weights element by element, and accurately focuses a lesion area and a boundary. Experiments show that the Dice coefficients of the method on INbreast, CBIS-DDSM and private In-home data sets respectively reach 90.94%, 80.60% and 84.50%, the method is superior to a mainstream method, the segmentation precision and generalization ability are improved, and reliable support is provided for early screening and computer-aided diagnosis of breast cancer.

The invention discloses a generative artificial intelligence cerebral artery lesion detection method based on plain-scan brain CT. The method comprises the steps that 1, a plain-scan brain CT image to be processed is acquired; 2, converting the plain scanning brain CT image into a CTA image through an adaptive noise elimination network; and 3, carrying out multi-modal lesion detection on the basis of the CTA image generated in the step 2. According to the method, a CTA image is generated through an adaptive noise elimination network (ANE-NET), and a real-time lesion feature analysis engine (RTAL-FE) is embedded in the generation process, so that real-time classification prediction of lesion types is realized. And furthermore, a detection result is output through a dynamic weight decision model (DWD-M), so that the generation quality and the detection efficiency are remarkably improved. The problems that a traditional method is low in generation quality and lags behind detection are solved, and the method is particularly suitable for low-dose and non-invasive cerebrovascular disease screening scenes and has important clinical application value.

The invention discloses a medical image report generation method and device and a storage medium. The method comprises the following steps: acquiring a target image examination part corresponding to a target examination object and a first image description text; and if it is detected that the first image description text has a writing error, correcting the first image description text based on a reference correction rule matched with the image description text in a correction rule library, and generating a second image description text. And identifying a plurality of medical entity words contained in the second image description text, and classifying the plurality of medical entity words to obtain classification tags corresponding to the plurality of medical entity words. According to the method, multiple pieces of structured data used for representing lesion features are generated based on the classification labels corresponding to the multiple medical entity words, and the multiple pieces of structured data are filled into the medical image report template corresponding to the target image examination part, so that the medical terms in the image description text can be more accurately filled into the template; and a high-quality image report is obtained.

The invention relates to the technical field of biomedical engineering, and provides a tongue body motion form reconstruction and neuromuscular lesion recognition method, which comprises the following steps: S1, acquiring a multi-modal original signal by using a flexible electrode array attached to a lingual surface, and recording a system timestamp; s2, performing targeted filtering, artifact removal and feature extraction processing on the multi-mode original signal; s3, reconstructing a local dynamic curved surface of the lingual surface by adopting a preset model, and inferring an overall position and posture sequence of the tongue body in combination with an IMU signal; s4, a joint optimization objective function is constructed, and a tongue body three-dimensional dynamic form sequence is obtained through iterative solution; s5, performing motion unit action potential MUAP decomposition and feature extraction on the surface electromyogram signal sEMG; and S6, constructing lesion feature vectors based on the multi-dimensional abnormal features, and analyzing and generating a tongue lesion risk map TDM through a machine learning model. The defects that real motion reconstruction, form inference and neuromuscular lesion positioning of the tongue cannot be achieved through an existing tongue detection technology are overcome.

The present application relates to a kind of multi-center medical image lesion typing methods based on federal spectrum clustering, belong to medical image analysis technical field, including the following steps: S1: multiple medical institutions are regarded as federal learning client, and the medical image data and supporting clinical data of each client local are obtained and stored in each client local;S2: feature extraction and multimodal feature fusion are carried out in each client local, and multimodal lesion feature set is obtained;S3: center server and each client pass through iteration communication, and the multimodal lesion feature set is jointly clustered using federal spectrum clustering algorithm, until global clustering center converges;Only transmission encrypted statistical parameters in communication process;S4: according to clustering result, output lesion typing information.The present application solves the pain points of difficult data sharing and high privacy leakage risk in traditional centralized lesion analysis, and improves the lesion typing accuracy in heterogeneous data scenario.

The invention provides a Mama-based membranous nephropathy multi-mode pathological image quantitative analysis system and method, and belongs to the crossing field of biomedical engineering and artificial intelligence. The problems that in an existing membranous nephropathy diagnosis system, the diagnosis process is high in subjectivity, single-mode analysis is limited, lesion feature quantification is insufficient, and model calculation is complex are solved. According to the technical scheme, the system comprises an image preprocessing module, the image preprocessing module is in communication connection with a macroscopic lesion analysis module, a microstructure analysis module and a thickness quantification module, and the macroscopic lesion analysis module and the thickness quantification module are both in communication connection with a feature fusion and prediction module. The macroscopic lesion analysis module, the microstructure analysis module, the thickness quantification module and the feature fusion and prediction module are all in communication connection with the result visualization module, and multi-modal pathological image quantitative analysis of membranous nephropathy is realized; the method is applied to multi-mode pathological image analysis of membranous nephropathy.

The application discloses a kind of based on multimodal high-resolution anoscope image lesion classification system and method.The system includes image acquisition module, lesion area segmentation module, feature fusion module and classification module.Image acquisition module obtains the original image of the same patient anal region, acetic acid white image and iodine test image.Lesion area segmentation module is based on the unified lesion area mask of multimodal image feature generation.Under the mask constraint, feature fusion module extracts the lesion feature of each modality located in the mask covered area and carries out cross-modal fusion, and obtains fusion feature vector.Classification module outputs the classification result of LSIL, HSIL and condyloma acuminatum.The application suppresses background interference by mask-guided spatial hard truncation, realizes cross-modal semantic alignment by sharing backbone network, guarantees input reliability by quality screening gate, and improves classification accuracy and robustness.Clinical experiment classification accuracy reaches 95.8%.

This invention discloses a heterogeneous two-stream fusion method and system for grading diabetic retinopathy (DR), comprising obtaining the DR grading output using a heterogeneous two-stream architecture: processing the input fundus image into images of different resolutions; extracting global contextual features from the low-resolution image using a lightweight visual Transformer model distilled from composite knowledge, and extracting local lesion features from the high-resolution image using a convolutional neural network model; interactively fusing the global contextual features and local lesion features of the two-branch architecture through a symmetrical bidirectional cross-attention fusion module to obtain an enhanced fused feature representation; and finally inputting the fused features into a classifier to output the DR severity grading result. This invention aims to improve the accuracy and robustness of grading diagnosis through in-depth analysis of global information and local details, and can be applied to medical fields such as clinical computer-aided diagnosis and ocular image analysis.

This application provides a method and system for controlling an orthopedic device, as well as the orthopedic device and storage medium, applicable to scenarios of intelligent control of medical devices and spinal biological analysis. The method includes: collecting current anatomical structure data, current biological pressure data, and current soft tissue lesion data of the target object's spine; extracting spinal disease features from the current anatomical structure data, current biological pressure data, and current soft tissue lesion data to obtain current spinal lesion features; constructing an orthopedic plan based on the current spinal lesion features and a preset spinal treatment plan to obtain a preliminary orthopedic plan; generating preliminary control instructions based on the preliminary orthopedic plan and preset control instructions for the orthopedic device; and controlling the orthopedic device according to the preliminary control instructions. This application embodiment can automate the control of the orthopedic device, improve the accuracy of orthopedic device control, and save manpower.