119 results about "Lesion mapping" patented technology

The invention relates to the field of image segmentation, in particular to a multi-modal nuclear medicine image segmentation method and system and a storage medium. Comprising the steps that image type classification is carried out on multi-modal medical examination images, focus area positioning is carried out, and CT (computed tomography) focus images and PET (positron emission tomography) focus images are obtained respectively; carrying out lesion geometric mutation form prediction on the CT fault lesion image, and carrying out multi-modal image fusion boundary correction on the CT fault lesion image and the PET lesion image to obtain multi-modal lesion image fusion boundary correction data; according to the multi-modal lesion image fusion boundary correction data, performing multi-modal lesion region segmentation combination optimization on the CT image and the PET image to obtain a lesion region segmentation combination optimization strategy; and carrying out image segmentation model construction on the lesion region segmentation combination optimization strategy based on a strategy gradient algorithm to obtain a lesion image segmentation model. According to the method, the image segmentation technology is optimized, so that the image segmentation technology is more perfect.

The invention provides a disease monitoring method and system based on medical image data recognition, and relates to the technical field of image processing. The method comprises the following steps: acquiring multi-level prior information related to a target disease from a medical knowledge graph, encoding focus attributes and time sequence attributes in the multi-level prior information into knowledge prior vectors, inputting the knowledge prior vectors into a generator, and enabling the generator to output focus image data based on the knowledge prior vectors; when the lesion image data output by the generator and the knowledge prior vector are subjected to difference evaluation, the discriminator also calls lesion boundaries and staging limitation rules in the medical knowledge graph to perform compliance judgment on the form and staging state of the lesion image data; updating network parameters of the generator based on a difference evaluation and compliance determination result output by the discriminator to obtain target lesion image data; and combining the focus image data with the original medical image, inputting the combined image data into a preset disease monitoring model, and generating a potential focus detection result.

The invention discloses a sliding table motion control method and device for a laparoscopic surgical robot, and relates to the technical field of surgical robos.The method comprises the steps that a lesion three-dimensional model is established based on a lesion CT image, a target area and boundary coordinates of the target area are obtained, and a coordinate transformation matrix is calculated for space registration to obtain a unified coordinate system; selecting a moving mode according to the user control instruction; recording operation data of users, classifying the users to obtain corresponding force speed curves, force displacement sequences and habitual force, and establishing a comprehensive control model and an error compensation model; when entering a fine adjustment mode, inputting a real-time distance, a user type and a force displacement sequence into a comprehensive control model, determining a regulation and control scheme, and reversely correcting output displacement according to a real-time positioning error, so that the auxiliary surgical instrument reaches a target position; high-precision motion control over the sliding table (2) of the laparoscopic surgery robot is achieved, the surgery time is remarkably shortened, the target position of an auxiliary instrument is accurately and rapidly positioned, and the surgery efficiency is improved.

The invention discloses a digital image enhancement method and system based on an endoscope, and aims to solve the problem of imaging limitation of a traditional endoscope in multi-focus positioning and targeted intervention. According to the digital image enhancement method based on the endoscope, endoscope image acquisition, wireless transmission, abnormal anchor point identification and controllable terminal targeted intervention operation can be carried out. The enhancement system for realizing the digital image enhancement method based on the endoscope identifies a suspected lesion point through image analysis, the control unit sends a release instruction to enable the endoscope to release the controllable terminal and establish communication, the scheduling system navigates the controllable terminal to an abnormal anchor point according to position information, and the abnormal anchor point is sent to the control unit. A focus image after digital image enhancement is collected and transmitted, and a judgment unit analyzes whether a target focus with an intervention value exists or not. According to the method, the digital image in the human body cavity can be enhanced, and the focus recognition accuracy and intervention efficiency are improved.

The invention discloses a focus segmentation method fusing PET and CT bimodal images, and relates to the technical field of computers. The method comprises the following steps: acquiring focus image data, separating the focus image data to obtain CT data and PET data, and inputting the CT data and the PET data into a bimodal medical image segmentation model to obtain a preliminary segmentation result; a difference region is generated based on the preliminary segmentation result, difference region detection is carried out on the difference region, and click signal data is generated; the difference region is a region which is not marked as a focus in the preliminary segmentation result; training the bimodal medical image segmentation model for a preset number of times through the click signal data, the CT data and the PET data to obtain a segmentation result; and determining a segmentation result generated by the last training as a lesion image segmentation result. The method can improve the medical image lesion segmentation precision.

The invention relates to the technical field of image segmentation, in particular to a tumor prediction method and system based on image segmentation processing, and the method comprises the following steps: dividing a lesion image into sub-regions according to space, extracting probability construction mapping, generating a prediction map, analyzing gradient change, judging a difference region, generating a density mutation map, extracting a gray scale trend, and generating a time consistency mapping table. And fusing the inter-frame probability reconstruction distribution to generate a high-sensitivity image, and aggregating a high-value region to extract a boundary index and mark the boundary index as a suspicious lesion position list. According to the method, the image sequence is divided into the sub-regions, and the prediction map is constructed in combination with the pixel probability and position relationship, so that the region recognition accuracy is improved, the probability gradient in the region is extracted to recognize the abnormal density mutation region, the time gray change direction is tracked, and dynamic consistency classification is realized; adjacent image pixel prediction results are fused to reconstruct local probability distribution, prediction stability is enhanced, abnormal areas are aggregated in the high-sensitivity image, boundaries are marked, and recognition sensitivity and reliability are improved.

The invention discloses a skin lesion image segmentation method and device based on deep learning, and belongs to the technical field of medical image processing and computer vision. The method comprises the following steps: acquiring a skin lesion image data set, and preprocessing the skin lesion image data set; constructing an RMS-Unet model based on the improved U-net network, wherein the model comprises a residual structure-convolution module, a channel-space attention module, a feature fusion module and a multi-scale convolution module; and training the RMS-Unet model by using the skin lesion image data set, and performing skin lesion image segmentation by using the trained RMS-Unet model. The RMS-Unet model adapts to the diversity and complexity of the skin lesion image by solving the problem of noise interference in the skin lesion image and the inherent induction deviation of a segmentation task, and effectively improves the segmentation precision of the complex skin lesion image while the parameter quantity and the model complexity are not remarkably increased.

The invention relates to the technical field of medical image processing, and particularly discloses a lesion image generation method for ultrasonic diagnosis, which comprises the following steps of: 1, acquiring initial state data of a lesion; 2, obtaining intervention factor data; 3, performing feature coding on the initial state data and the intervention factor data to obtain an initial feature vector and an intervention feature vector; and 4, inputting the initial feature vector and the intervention feature vector into a pre-trained time sequence generation model, and generating a continuous image sequence of the focus changing along with time. By constructing the time sequence generation model fusing the multi-modal feature coding and the space-time attention mechanism, the continuous image sequence of the focus changing along with time can be accurately generated, the disease progress or treatment response process can be accurately simulated, and a powerful reference basis is provided for clinical prognosis evaluation.

The invention discloses an ulcerative colitis image classification method based on a multi-scale attention fusion network, belongs to the technical field of image information mining and classification, and combines a staged feature extraction backbone network, a multi-scale attention fusion module (SCHL) and a classification module. The ulcerative colitis lesion image classification method has the advantages that significant advantages are shown in ulcerative colitis lesion image automatic classification tasks, an OfficientNet backbone network has staged feature extraction capacity, multi-level information from shallow texture to deep semantics in lesion images can be effectively captured, an SCHL module performs modeling by fusing channel attention, space attention and level high-frequency information, and the ulcerative colitis lesion image classification efficiency is improved. The model is guided to focus on a key lesion area, background interference is inhibited, feature discrimination is enhanced, the recognition ability of the model to a fine-grained lesion form and the generalization ability of the model to a complex clinical image are remarkably improved, and higher accuracy and robustness are achieved in refined classification of ulcerative colitis.

The invention discloses an oral disease detection method based on an artificial intelligence large model, and relates to the technical field of oral medicine. Comprising the following steps: acquiring a tooth focus picture and corresponding oral cavity internal state description, respectively extracting picture feature information of the tooth focus picture and text feature information of the oral cavity internal state by adopting a feature extraction block BERT and a multi-scale feature extraction block MFEM, and identifying and eliminating false negative samples therein; a total loss function and a dynamic label distribution strategy are introduced into a YOLOv8 network, an improved YOLOv8 network is obtained, a data set is divided into a training set and a test set, the training set and the test set are sequentially input into the improved YOLOv8 network for verification, the verified YOLOv8 network is obtained, a tooth focus picture and corresponding text information are input into the verified YOLOv8 network, and focus area type detection and text description are obtained. According to the invention, detection and classification of various diseases in the oral cavity can be effectively improved, and corresponding focus area description and suggestion are given.

The present application provides a method for detecting pseudo-lesions of the fundus, comprising: performing image feature analysis of the lesion area according to preliminary lesion markings, locating and evaluating the cell density and the degree of vascular leakage, and determining the risk of disease aggravation; predicting the tortuosity and vascular fragility of new blood vessels and evaluating the probability of complications in patients according to morphological processing and feature analysis of multi-phase vascular images; predicting the risk of serious complications according to the lesion markings and vascular feature analysis of the patient, combined with blood test results, and re-identifying and excluding pseudo-lesion images that are mistaken for lesion images; comparing the feature changes of pseudo-lesions and real lesion areas according to time series analysis of fundus disease images, evaluating the disease development path, and performing early intervention on the disease; segmenting the new blood vessel area based on the use of sensitive index filtering technology, and evaluating the risk of disease aggravation through vascular feature statistics.

The invention provides a path planning method, a path planning device and electronic equipment, and relates to the technical field of image processing, and the method comprises the steps: obtaining a lesion image of a target object, and determining a target region in the lesion image; traversing each target area, determining a plurality of initial paths from the path planning start area to the path planning end point area, and determining a path evaluation value corresponding to each initial path; and based on the path evaluation value, screening a target path from the path planning start area to the path planning end point area from the plurality of initial paths. According to the process, a target path from a path planning starting area to a path planning end point area is screened based on a path evaluation value, path planning is carried out in an intelligent calculation mode, objectivity and standardization of path planning can be guaranteed in the planning process, the overall quality of path planning can be improved, and the path planning efficiency is improved. Therefore, scientific auxiliary information can be provided for the operation, and the subsequent treatment effect is improved.

The invention relates to the technical field of chloasma diagnosis and treatment effect evaluation, in particular to a chloasma diagnosis and treatment effect evaluation system based on multi-dimensional analysis. The system comprises an image segmentation processing module and a sample subset feature image selection module. According to the invention, the image segmentation processing module adopts a region-level image segmentation mode to carry out image segmentation processing on a collected lesion image, independent feature segmentation is carried out in combination with chloasma pathological features to obtain single feature images under different features, and meanwhile, sample subset feature images are marked through the sample subset feature image selection module, so that the sample subset feature images are obtained. And obtaining feature image screening items of each sample subset, updating features in a chloasma recovery stage in real time, and obtaining differences of the two features in combination with an initial lesion image to serve as a reference basis for evaluating the chloasma diagnosis and treatment effect, so that the chloasma diagnosis and treatment effect is evaluated in a multi-dimensional manner, the subjectivity of manual evaluation is reduced, and the evaluation efficiency is improved. Effects of different types of chloasma treatment schemes are objectively evaluated, and the evaluation accuracy is improved.

The invention relates to the field of image processing, in particular to a fundus disease image segmentation method, system and device. The method comprises the following steps: transforming the shape, position and color of a focus area of a real fundus disease image, and fusing with a focus-free image to obtain a preliminary synthesized focus image and a segmentation label; respectively inputting the features into a high-frequency detail optimization model consisting of a dual-path feature extractor and a decoder, acquiring global context features by the model by using a mask auto-encoder, extracting local detail features through a convolutional neural network, and judging whether to fuse along a channel according to a spatial pixel quantity relationship between local and global features of different scales; the target features of different scales are transmitted to a decoder through jump connection, and an optimized composite image is obtained through transposition convolution up-sampling; and taking the optimized multi-category synthetic image and the tag as an expansion data set, training a focus segmentation network, and segmenting an actual disease image by using a trained model, thereby effectively solving the problem of scarcity of fundus disease samples.

The application discloses a liver lesion image description method based on a semantic segmentation network, comprising the following steps: building a Unet semantic segmentation network based on a lightweight network GhostNet, and constructing a liver lesion segmentation model; acquiring a historical liver lesion ultrasound image dataset, inputting the liver lesion segmentation model for training, and obtaining an optimal segmentation model; and describing the features of the lesion image based on the segmentation result. The application segments the liver lesion by constructing a lesion segmentation model, and describes based on the segmentation result, thereby improving the accuracy and reliability of liver disease diagnosis, reducing misjudgment and subjective bias; by combining the advantages of deep learning and traditional image processing, the characteristics of both are fully utilized, the performance and stability of the algorithm are improved; automatic processing and analysis of the ultrasound image are realized, the lesion description result provides accurate disease positioning and type judgment, and a better treatment scheme is provided for doctors, thereby improving the treatment effect and treatment experience of patients.

The present invention relates to the technical field of minimally invasive instrument control, and in particular, to a minimally invasive instrument control system and method for surgical operating rooms. The system comprises a master hand operation module, an operation position monitoring module and a tactile feedback module. The present invention controls the contact between a tactile operating arm and a lesion through a tactile feedback module, obtains contact feedback, and sends the contact feedback to the master hand operation module, controls the vibration feedback of a vibration instrument built into the master hand operation module, and feeds back vibrations in real time to the hand through a hand joint encoder held by a doctor by referring to the vibration mode matched with the lesion area, determines the hardness of the lesion area currently contacted by the tactile operating arm, and cooperates with the intraoperative patient lesion image displayed in real time by a 3D image real-time display module to predict the state of the lesion operation point, provide tactile feedback for doctors performing remote surgery, further improve the accuracy of the surgery, and assist doctors in performing surgical operations.

The present application relates to the technical field of pest and disease monitoring, and specifically discloses a method and system for monitoring and early warning of agricultural pests and diseases based on image processing, which continuously collects under-exposed, normally exposed, and over-exposed images of farmland crops, and aligns and combines the three into a high dynamic range image that fully retains the details of highlights and shadows. Subsequently, target detection is performed on the high dynamic range image to generate a binary mask map of the suspicious lesion area, and the binary mask map is further used to extract the lesion area images from the three original images with different exposures. By performing deep pest and disease feature extraction and multi-source feature dynamic compensation aggregation on the three groups of multi-source suspicious lesion images containing complementary information, a comprehensive and integrated representation of the pest and disease characteristics is formed, and precise identification of the pest and disease types is achieved. This method can improve the accuracy and robustness of agricultural pest and disease monitoring under complex natural lighting conditions through the comprehensive utilization of multi-exposure image data.

A method for mapping one or more cutaneous conditions of a cutaneous surface may include: identifying a body map; identifying one or more body regions corresponding to the body map; identifying at least one representation corresponding to the one or more body regions; identifying one or more icons corresponding to the at least one representation; and defining the one or more icons on the at least one representation.

According to the ultrasonic imaging equipment and the lesion distribution presentation method provided by the invention, the volume data is firstly obtained, then the target tissue and the target lesion are identified from the volume data to obtain the three-dimensional contours of the target tissue and the target lesion, and then the three-dimensional schematic diagram is generated based on the two three-dimensional contours. The three-dimensional schematic diagram is used for presenting the form of the target tissue and the form and the position of the target focus in the target tissue, so that a user can intuitively know the distribution condition of the focus in space, the distribution of the focus is easier to understand and master compared with an ultrasonic image, and the working efficiency of a doctor is improved. And generating a two-dimensional diagram corresponding to at least one target section according to the three-dimensional contour of the target tissue and the focus, wherein the two-dimensional diagram corresponding to the target section comprises a target tissue graph used for presenting the form of the target tissue and a target focus graph used for presenting the form and the position of the target focus. Therefore, the user can know the distribution of the focus through the two-dimensional schematic diagram.

The present application relates to the technical field of image processing, and more particularly to a lesion segmentation method of adaptive dynamic text prompt, comprising: acquiring a lesion image, and generating an adaptive dynamic text prompt based on the lesion image; constructing a multi-modal enhanced fusion text prompt adapter based on a FiLM global channel recalibration module, a spatial cross-attention interaction module and a zero initialization gate nonlinear integration module cascade; the global channel recalibration module of FiLM utilizes a feature linear modulation mechanism to perform channel-level weighting on a visual feature map; before the feature enters spatial interaction, according to the text semantic enhancement band response, the activation of the background noise channel is inhibited. The present application overcomes the problems of insufficient utilization of multi-modal prompts, dependence on artificial static labeling of text prompts, serious dependence on prior information in the reasoning stage and difficulty in multi-modal feature fusion in a low-contrast environment in the existing medical image segmentation technology.

The invention discloses a medical image encryption method based on a trap door network and a diffusion model, and relates to the technical field of image encryption. The method comprises the following steps: encrypting a focus image scanned by medical equipment of a patient by using a session key through key negotiation between the patient and a doctor, and sending the encrypted image to an image filing and communication server; and when the doctor needs to look up the lesion image, decrypting the ciphertext image through the previously established session key, and recovering the corresponding original lesion image. The image encryption / decryption network of the invention is composed of a Diffusion model and a trap door network. The trap door network enhances the security of the medical image encryption method, and the forward generation and sampling process based on the Diffusion model ensures the security of the encrypted image and the high-quality reconstruction effect of the decrypted image.

The application belongs to the technical field of image recognition, and discloses a monkeypox virus classification method and system based on image comparison and a storage medium, which comprises an individual reference modeling module, collects normal skin and initial lesion images of a healthy side of a patient, generates a healthy texture template and an initial lesion reference library, performs gray scale normalization on the healthy texture template and the initial lesion reference library through adaptive histogram stretching, and outputs a comparison reference image group; a symptom density quantification module, acquires a current monitoring image of the patient, segments a lesion area in combination with the comparison reference image group, extracts layered features of the lesion area to generate a three-dimensional symptom density vector, extracts texture roughness of a surface of the lesion area, and constitutes a lesion quantification data set; the accuracy of disease monitoring is improved, and a clinical nursing process is optimized.

The invention relates to the technical field of medical treatment, and discloses a gastrointestinal endoscope diagnosis method based on AR glasses. The method comprises the steps that an initial detection image which is shot by a gastrointestinal endoscope and corresponds to a target object is obtained and sent to the AR glasses to be displayed; according to a preset planning path, utilizing a navigation algorithm to generate guidance indication information, and displaying the guidance indication information in the initial detection image displayed by the AR glasses so as to control the gastrointestinal endoscope to move according to a display picture of the AR glasses; performing focus recognition on all the real-time detection images through a preset focus recognition model to obtain a focus recognition result; according to the lesion identification result, determining a lesion marking area and a suggested biopsy position, and determining the lesion marking area as a lesion image; the focus image is sent to AR glasses for display, and a gastrointestinal endoscope diagnosis report is generated according to the focus image. According to the focus characteristics of the white light endoscopic image and the narrow-band wide image, determination of the diseased region, the marked focus region and the suggested biopsy region is achieved, and deep combination of gastrointestinal endoscope diagnosis, the AR technology and artificial intelligence is achieved.

The present invention discloses an endoscope-based digital image enhancement method and system, which aims to solve the imaging limitations of traditional endoscopes in multi-lesion positioning and targeted intervention. The endoscope-based digital image enhancement method can perform endoscopic image acquisition, wireless transmission, abnormal anchor point identification and controllable terminal targeted intervention operations. The enhancement system for implementing the endoscope-based digital image enhancement method identifies suspected lesions through image analysis, the control unit issues a release instruction, causes the endoscope to release the controllable terminal and establish communication, the scheduling system navigates the controllable terminal to the abnormal anchor point according to the position information, acquires and transmits the digitally enhanced lesion image, and the judgment unit analyzes whether there is a target lesion with intervention value. This method can enhance the digital image in the human body cavity and improve the accuracy of lesion identification and intervention efficiency.

The invention discloses a focus image recognition method and system based on a deep learning model, and relates to the technical field of medical equipment. Comprising the following steps: acquiring medical images of various lesion types, and preprocessing the medical images to generate sample images; marking a focus area, extracting image feature parameters, and constructing a sample training data set; and based on the data set, training a segmentation identification model, preprocessing a medical image to be identified, identifying and segmenting a focus area by using the trained segmentation identification model, inputting a feature extraction model to obtain an image feature parameter prediction value, and calculating a biochemical discrimination influence index in combination with biochemical index data of a patient. And generating a comprehensive lesion identification coefficient according to the lesion texture complexity index, the edge influence index and the biochemical discrimination influence index, and comparing the comprehensive lesion identification coefficient with a set threshold to complete lesion image identification. And the accuracy and reliability of focus identification are improved.

The invention relates to the technical field of disease and insect pest monitoring, and particularly discloses an agricultural disease and insect pest monitoring and early warning method and system based on image processing, and the method comprises the steps: continuously collecting underexposure, normal exposure and overexposure images of farmland crops, and enabling the underexposure, normal exposure and overexposure images to be aligned and combined into a high-dynamic-range image which completely retains highlight and dark part details. Then, target detection is carried out on the high dynamic range image to generate a binary mask image of a suspicious lesion area, and lesion area images are extracted from the three original images with different exposures by further utilizing the binary mask image; according to the method, deep pest feature extraction and multi-source feature dynamic compensation aggregation are performed on three groups of multi-source suspicious focus images containing complementary information, so that comprehensive characterization of pest features is formed, and fine recognition of pest types is realized. According to the method, through comprehensive utilization of the multi-exposure image data, the accuracy and robustness of agricultural pest monitoring under the complex natural light condition can be improved.

The invention relates to the technical field of lesion recognition, solves the technical problem that details in an image cannot be accurately recognized in the prior art, and particularly relates to a method for recognizing a leukoplakia vulvae lesion image, which comprises the following steps of: S1, acquiring a leukoplakia vulvae lesion image at a # imgabs0 # moment, performing primary processing on the focus image to obtain a first processing value # imgabs1 #; and S2, calculating a second processing value # imgabs3 # of the lesion image at the # imgabs2 # moment, and fusing the first processing value # imgabs4 # and the second processing value # imgabs5 # to obtain a preprocessed image. According to three-value positioning of the to-be-processed image, classification of pixels in the to-be-processed image can be completed according to a query value, a key value and a positioning value, the fine-grained extraction capability is improved by means of convolution and an attention mechanism, and the accuracy of the fine-grained extraction is improved. Meanwhile, the parameter number of the model can be optimized, the model operation recognition efficiency is improved, and the accuracy and the recognition speed of focus position recognition in the to-be-processed image can be improved by combining the average pooling mode with the classification threshold value.

The invention relates to the technical field of magnetic resonance imaging, in particular to an intra-operative low-field magnetic resonance intelligent navigation method based on preoperative high-field data guidance, which comprises the following steps: acquiring a high-field MRI image of a patient before an operation and extracting a three-dimensional segmentation mask and clinical features of a focus; adjusting according to the clinical features to obtain an intraoperative scanning protocol; and performing low-field magnetic resonance examination during the operation to generate an intra-operation navigation image, and marking focus information. Aiming at the problem that the accuracy of a focus image provided by an ultra-low field system in the operation process in the prior art is not enough, a high-field MRI image collected before the operation is introduced as a reference, scanning sequence parameters of the ultra-low field system used in the operation are dynamically adjusted based on clinical features embodied in the high-field MRI image, a better imaging effect is achieved, and meanwhile, the accuracy of the focus image provided by the ultra-low field system in the operation process is improved. Focus information is marked on the intraoperative navigation image based on the three-dimensional segmentation mask as prior information, so that more refined segmentation of focuses on the intraoperative navigation image is realized, and the surgical requirements are met.

The present application provides a method, apparatus, device and storage medium for generating an answer based on medical images, the method comprising: obtaining a target question, obtaining a first period medical image and a second period medical image according to the target question; performing feature extraction on the medical image through a lesion extraction model to obtain a first lesion feature and a second lesion feature; performing feature matching on the first lesion feature and the second lesion feature to obtain lesion matching information; updating the first lesion feature, the second lesion feature and the text embedding feature based on a first attention mechanism to obtain a lesion map of the first period and a lesion map of the second period; performing a difference comparison between the lesion map of the first period and the lesion map of the second period based on the lesion matching information to obtain a lesion difference map; performing feature extraction on the lesion map of the first period, the lesion map of the second period and the lesion difference map based on a second attention mechanism to obtain visual features; and obtaining a corresponding target answer according to the visual features.

The invention relates to the technical field of image processing, in particular to a gynecological tumor MRI image three-dimensional reconstruction method and system based on deep learning, and the method comprises the steps: processing a coarse label tumor region through a target detection model based on an original image, and generating a tiny focus positioning frame; using a positioning frame for segmentation to obtain a tiny focus image and a non-tiny focus image, and scaling down the tiny focus according to a proportion to obtain a focus image; according to the generated boundary region, selecting a sample to calculate a gray average value and a difference coefficient, and performing weighted merging to obtain a normal tissue image; compressing the tumor core area and the boundary area to obtain an overall tumor image; the focus image, the normal tissue image and the tumor image are spliced together to form a low-resolution image training model, and then features are segmented and extracted to construct a tumor three-dimensional model. According to the method, tiny lesions and boundary features are reserved, the segmentation precision is improved, a high-precision three-dimensional model is constructed, and clinical diagnosis requirements are met.