Liver tumor auxiliary diagnosis system based on artificial intelligence
By integrating multimodal data and a real-time feedback mechanism, an AI-based liver tumor auxiliary diagnostic system has been developed, which has solved the problems of misdiagnosis risk and long diagnosis cycle in liver tumor diagnosis, and achieved automated support for accurate classification and personalized treatment.
Patent Information
- Application Number
- CN202511118068.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-08-11
- Publication Date
- 2025-11-21
AI Technical Summary
Current liver tumor diagnosis relies on doctors' experience, carries the risk of misdiagnosis or missed diagnosis, has a long diagnosis cycle, lacks multimodal data fusion and real-time feedback mechanisms, and makes it difficult to achieve dynamic monitoring and personalized treatment plan optimization.
An AI-based liver tumor auxiliary diagnostic system is adopted, which integrates deep learning image analysis, multimodal data fusion and real-time feedback mechanism to achieve automatic identification, accurate classification and personalized treatment suggestion generation of liver tumors. It includes modules for data acquisition, image processing, tumor classification and staging, clinical data analysis and real-time feedback and monitoring.
It reduces subjective bias in manual image interpretation, improves diagnostic accuracy, meets emergency needs, generates personalized treatment plans, enables full-cycle tracking and timely adjustments, reduces the risk of misdiagnosis, and improves diagnostic efficiency and personalized adaptability of treatment.
Smart Images
Figure CN120998463A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of medical tumor diagnosis, in particular to an auxiliary diagnosis system for liver tumor based on artificial intelligence. BACKGROUND
[0002] Liver tumor, especially hepatocellular carcinoma, is one of the high-incidence and high-risk malignant tumors in the world, and its early diagnosis and precise treatment directly affect the prognosis of patients. Currently, the diagnosis of liver tumor in the clinic mainly relies on medical imaging such as CT, MRI and ultrasound and examination as well as the clinical experience of doctors. However, the above methods have some problems.
[0003] The diagnosis result is more dependent on the personal experience of doctors, and different doctors are prone to bias in judging the image features, which can easily lead to misdiagnosis or missed diagnosis. Since artificial analysis of images requires frame-by-frame review, the diagnosis cycle of complex cases can take several hours, which is difficult to meet the clinical demand for rapid diagnosis. The image data and clinical data, such as liver function indicators and medical history, are analyzed separately, lacking comprehensive evaluation of multi-modal data, which affects the comprehensiveness of diagnosis. The existing diagnosis system cannot adjust the diagnosis model in real time combined with patient treatment feedback, and cannot realize dynamic monitoring of tumor progression and timely optimization of treatment plan.
[0004] Although some AI-assisted diagnosis systems have been applied to medical image analysis, most of them are limited to tumor recognition of single image, lack the integration ability of image, clinical and treatment in the whole chain, and have not formed a closed-loop mechanism of diagnosis, feedback and optimization. Therefore, an intelligent diagnosis system capable of multi-modal data fusion, dynamic monitoring and personalized decision support is needed to solve the above problems. SUMMARY
[0005] In view of the above situation, in order to overcome the technical defects of the present, the present application provides an auxiliary diagnosis system for liver tumor based on artificial intelligence, which is convenient to use. The system is based on artificial intelligence for auxiliary diagnosis of liver tumor, and realizes automatic recognition, accurate classification, clinical staging and personalized treatment suggestion generation of liver tumor by integrating deep learning image analysis, multi-modal data fusion and real-time feedback mechanism, and provides comprehensive diagnostic support for the clinic.
[0006] The technical scheme adopted by the present application is as follows: The auxiliary diagnosis system for liver tumor based on artificial intelligence provided by the present application comprises a data acquisition module, an image processing module, a tumor classification and staging module, a clinical data analysis module, a diagnosis decision support module and a real-time feedback and monitoring module.
[0007] The data acquisition module is used for acquiring multi-source data of patients, and the multi-source data includes but is not limited to medical image data and clinical data.
[0008] The image processing module adopts deep learning technology to automatically process and analyze liver images, and the image processing module comprises:
[0009] Image preprocessing, image noise is removed through Gaussian filtering, adaptive threshold denoising and other algorithms, and image preprocessing is performed on fuzzy images for super-resolution reconstruction;
[0010] Region segmentation, using a convolutional neural network (CNN) model, automatically segmenting liver anatomical regions and tumor suspicious regions;
[0011] Feature extraction, morphological features and imaging features of the tumor region are extracted through a residual network (ResNet), and a standardized feature vector is generated.
[0012] The tumor classification and staging module is based on the feature extraction results of the image processing module, and combines machine learning and clinical rules to realize accurate tumor evaluation. The tumor classification and staging module comprises:
[0013] Benign and malignant classification, using a support vector machine (SVM) and a deep learning classification network (MobileNet), classifying tumor feature vectors, and outputting benign, malignant and corresponding confidence;
[0014] Clinical staging, based on the TNM staging standard, integrating tumor size, lymph node metastasis and distant metastasis information through a decision tree algorithm, automatically determining tumor staging and labeling key evidence;
[0015] The clinical data analysis module integrates image analysis results and clinical data through multi-modal data fusion technology, which includes establishing an association model of image features, clinical indicators and medical history; using a Bayesian network algorithm to generate a comprehensive diagnosis result and output key influencing factors.
[0016] The diagnostic decision support module generates a personalized treatment plan based on the comprehensive diagnosis result and the medical knowledge base, which includes:
[0017] Plan generation, combining tumor staging and patient liver function reserve, using reinforcement learning algorithm to recommend treatment plan;
[0018] Effect evaluation, predicting the objective response rate, progression-free survival and adverse reaction risk of each plan through historical case big data.
[0019] The real-time feedback and monitoring module is used to build a closed-loop interaction mechanism between the system, doctors and patients. It realizes real-time interaction through medical display screens, doctor workstations and client-side real-time push of image analysis results and diagnosis suggestions, and can periodically collect patient follow-up images and clinical data, evaluate tumor changes through comparison algorithms, and automatically update staging and treatment recommendations to dynamically monitor.
[0020] Preferably, in the data acquisition module, medical image data supports DICOM format standardization import, and clinical data supports HL7 protocol automatic docking with a hospital information system (HIS) and a laboratory information system (LIS).
[0021] Preferably, in the image processing module, a convolutional neural network adopts a pre-training model and supports fine-tuning on local hospital data to adapt to image characteristics of different devices.
[0022] Preferably, in the tumor classification and staging module, a tumor image feature database is built in, used for real-time comparison and accuracy calibration of the model.
[0023] Preferably, in the clinical data analysis module, a medical knowledge base is real-time synchronized with the latest clinical guidelines to ensure that the diagnosis basis conforms to the current clinical specifications.
[0024] Preferably, in the real-time feedback and monitoring module, a doctor can mark misdiagnosis cases, the system automatically includes the cases in a model optimization data set, and automatically triggers model incremental training.
[0025] The system has the following beneficial effects:
[0026] The system can realize automatic segmentation and feature extraction of tumor regions through deep learning, reducing subjective bias of manual film reading. The system combines multi-modal data of images and clinical data to significantly improve the comprehensive diagnosis coincidence rate compared with traditional manual diagnosis. The image analysis whole process from data import to staging result output significantly improves the efficiency compared with manual film reading, meeting the needs of emergency and batch screening. The system can generate a treatment plan based on individual characteristics of a patient, and predict the effect combined with historical cases, so that the treatment suggestion is suitable for the actual situation of the patient, reducing the risk of ineffective treatment. The system can realize whole-cycle tracking of diagnosis, treatment and follow-up through real-time feedback and dynamic monitoring, timely discovery of tumor progression and adjustment of the plan, and improvement of long-term prognosis. BRIEF DESCRIPTION OF DRAWINGS
[0027] The accompanying drawings are included to provide a further understanding of the application, and constitute a part of the specification, illustrate the application, and are used to explain the application without limiting the application. In the drawings:
[0028] Fig. 1 FIG. 1 is a system structure schematic diagram of the liver tumor auxiliary diagnosis system based on artificial intelligence;
[0029] Fig. 2 FIG. 2 is a liver image processing flowchart of the liver tumor auxiliary diagnosis system based on artificial intelligence;
[0030] Fig. 3The tumor staging and treatment scheme logic diagram of the liver tumor auxiliary diagnosis system based on artificial intelligence. DETAILED DESCRIPTION
[0031] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative labor fall within the protection scope of the present application.
[0032] It should be noted that the words "front", "back", "left", "right", "up" and "down" used in the following description refer to the directions in the drawings, and the words "in" and "out" refer to the directions towards or away from the geometric center of a particular component.
[0033] As shown in the description accompanying drawings, Figs. 1-3 The technical solutions adopted by the present application are as follows:
[0034] The liver tumor auxiliary diagnosis system based on artificial intelligence provided by the present application comprises:
[0035] A data acquisition module is configured to acquire multi-source data of a patient, including medical image data, liver images acquired by a CT scanner, an MRI device, and an ultrasonic diagnostic instrument, and clinical data, patient information acquired by an electronic medical record system and a laboratory detection device, including basic information, hematological indexes, and medical history.
[0036] An image processing module is configured to automatically process and analyze the liver images by using a deep learning technology, including:
[0037] Image preprocessing, image noise is removed by using algorithms such as Gaussian filtering and adaptive threshold denoising, and fuzzy images are super-resolution reconstructed to ensure image clarity; regional segmentation, a convolutional neural network (CNN) model is used to automatically segment anatomical regions of the liver, such as the left lobe, the right lobe, the hepatic portal, and the target region suspected of tumor; feature extraction, morphological features such as size, shape, edge smoothness of the tumor region and imaging features such as density and signal intensity are extracted by using a residual network (ResNet) to generate a standardized feature vector.
[0038] A tumor classification and staging module is configured to realize accurate evaluation of the tumor based on the feature extraction results of the image processing module in combination with machine learning and clinical rules, including:
[0039] Benign and malignant classification, using support vector machine (SVM) and deep learning classification network (MobileNet), classifies tumor feature vectors, outputs "benign" and "malignant" and corresponding confidence; clinical staging, based on TNM staging standard, integrates tumor size, lymph node metastasis, and distant metastasis information through decision tree algorithm, automatically determines tumor staging, and labels key basis.
[0040] Clinical data analysis module: integrate image analysis results and clinical data through multi-modal data fusion technology, including:
[0041] Data correlation, establish the correlation model of image features, clinical indicators and medical history; comprehensive evaluation, use Bayesian network algorithm to generate comprehensive diagnosis results, including tumor nature, stage, potential risk, and output key influencing factors.
[0042] Diagnosis decision support module: based on comprehensive diagnosis results and medical knowledge base to generate personalized treatment plan, including:
[0043] Plan generation, combined with tumor staging and patient liver function reserve, using reinforcement learning algorithm to recommend treatment options such as surgical resection, radiofrequency ablation, and targeted drugs; effect evaluation, through historical case big data to predict the objective response rate, progression-free survival and adverse reaction risk of each scheme.
[0044] Real-time feedback and monitoring module is used to build a closed-loop interaction mechanism between the system, doctors and patients, including:
[0045] Real-time interaction, through medical display screen and doctor workstation client to push image analysis results and diagnosis suggestions in real time, support doctors to manually adjust tumor marking area and feedback correction suggestions; dynamic monitoring, regularly collect patient follow-up images and clinical data, evaluate tumor changes through comparison algorithm, and automatically update staging and treatment recommendations.
[0046] Specific use: hospitals collect patient liver images through CT / MRI equipment, synchronize patient basic information from HIS system, laboratory data from LIS system, and record hepatitis history; the system automatically preprocesses image data: removes device noise through Gaussian filtering, and corrects blurred areas caused by respiratory motion through image registration; standardizes clinical data processing;
[0047] Image processing module calls pre-trained U-Net model to segment liver region in preprocessed CT images; extracts features in this region through ResNet, generates feature vectors and stores them;
[0048] Tumor classification and staging module inputs feature vectors into SVM classification model, outputs and determines after comparison with built-in database; combined with image display and clinical data, determines according to TNM staging standard;
[0049] The clinical data analysis module fuses the decision information to generate a comprehensive report, synchronously outputs key basis, and the diagnosis decision support module makes a recommendation for diagnosis based on the decision basis; meanwhile, the historical case data are combined to comprehensively predict and evaluate the scheme;
[0050] The doctor checks the report through the workstation, checks the scheme, confirms the diagnosis result after no objection, and meanwhile the system automatically stores the case into the local database; after the patient is operated, the system compares the original tumor area through image registration, and updates the report.
[0051] It is to be understood that the terminology used herein such as first and second, and the like, is only used to distinguish one entity or action from another entity or action, and does not necessarily require or imply that there is any such actual relationship or order between such entities or actions. Moreover, the terms "comprising", "including", or any other variant thereof are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can also include other elements not expressly listed or inherent to such process, method, article, or apparatus.
[0052] Although the embodiments of the present application have been shown and described, it is to be understood that various changes, modifications, substitutions and alterations can be made to the embodiments without departing from the principles and spirit of the present application, and the scope of the present application is defined by the appended claims and their equivalents.
Claims
1. An artificial intelligence-based auxiliary diagnosis system for liver tumors, characterized by: It comprises a data acquisition module, an image processing module, a tumor classification and staging module, a clinical data analysis module, a diagnosis decision support module, and a real-time feedback and monitoring module; the data acquisition module is used for collecting liver image data and clinical data, the image processing module is used for liver image preprocessing and tumor feature extraction, the tumor classification and staging module is used for tumor benign and malignant determination and clinical staging, the clinical data analysis module is used for multi-modal data fusion and comprehensive diagnosis, the diagnosis decision support module is used for generating individualized treatment plans, and the real-time feedback and monitoring module is used for dynamically tracking tumor progression.
2. The liver tumor auxiliary diagnosis system based on artificial intelligence according to claim 1, characterized in that: The medical image data in the data acquisition module supports DICOM format standardized import, and the clinical data supports HL7 protocol and automatic docking with hospital information system (HIS) and laboratory information system (LIS).
3. The liver tumor auxiliary diagnosis system based on artificial intelligence according to claim 1, characterized in that: The convolutional neural network of the image processing module adopts a pre-trained model and supports fine-tuning on local hospital data to adapt to the image characteristics of different devices.
4. The liver tumor auxiliary diagnosis system based on artificial intelligence according to claim 1, characterized in that: The tumor classification and staging module has a built-in tumor image feature database for real-time comparison and accuracy calibration of the model.
5. The liver tumor auxiliary diagnosis system based on artificial intelligence according to claim 1, characterized in that: The medical knowledge base of the clinical data analysis module synchronizes the latest clinical guidelines in real time to ensure that the diagnosis basis conforms to the current clinical specifications. 6.The liver tumor auxiliary diagnosis system based on artificial intelligence according to claim 1, wherein: The real-time feedback and monitoring module supports doctor annotation of misdiagnosis cases, and the system automatically includes the case in the model optimization data set and automatically triggers model incremental training.