Search

CN-121982392-A - Endoscopic gastrointestinal tract image detection and recognition system and method based on artificial intelligence

CN121982392ACN 121982392 ACN121982392 ACN 121982392ACN-121982392-A

Abstract

The invention discloses an endoscopic gastrointestinal tract image detection and recognition system and method based on artificial intelligence, which belong to the technical field of medical image detection and comprise an image acquisition module, an image preprocessing module, a feature extraction module, a lesion detection module, a lesion recognition module, a model training module, a result output module, a data storage module, a man-machine interaction module and a quality evaluation module, wherein the image acquisition module is used for acquiring image data in the gastrointestinal tract and transmitting the image data to the image preprocessing module, and the image preprocessing module is used for preprocessing the image data. Through the cooperation of multiple modules, a set of full-flow intelligent detection and identification system from image acquisition to result output is constructed, the modules are clear in division and high in data transmission efficiency, the accurate processing and analysis of the endoscopic gastrointestinal tract images can be realized, and the defects of the traditional manual diagnosis and the existing AI diagnosis technology are effectively overcome.

Inventors

  • WANG LECHAO
  • ZHONG GUOYUN

Assignees

  • 东华理工大学

Dates

Publication Date
20260505
Application Date
20260120

Claims (10)

  1. 1. The endoscopic gastrointestinal tract image detection and recognition system based on artificial intelligence is characterized by comprising an image acquisition module, an image preprocessing module, a feature extraction module, a lesion detection module, a lesion recognition module, a model training module, a result output module, a data storage module, a man-machine interaction module and a quality evaluation module, wherein the image acquisition module is used for acquiring image data in a gastrointestinal tract and transmitting the image data to the image preprocessing module, the image preprocessing module is used for preprocessing the image data, the feature extraction module is used for extracting deep semantic features of an image by adopting an improved convolutional neural network model, the lesion detection module is used for detecting a lesion area based on the deep semantic features and determining position information, the lesion recognition module is used for carrying out feature refinement extraction on the lesion area and recognizing a lesion type, the model training module is used for training and optimizing a related model, the result output module is used for integrating detection recognition results and visualized output, the data storage module is used for storing various data, the man-machine interaction module is used for realizing interaction between a user and a system, and the quality evaluation module is used for evaluating detection recognition performance and feedback optimization of the system.
  2. 2. The endoscopic gastrointestinal tract image detection and recognition system based on artificial intelligence according to claim 1, wherein the preprocessing operation of the image preprocessing module comprises image denoising, image enhancement, image size standardization and image normalization, wherein the image denoising adopts an adaptive median filtering algorithm, and the image enhancement adopts a Retinex algorithm.
  3. 3. The endoscopic gastrointestinal tract image detection and recognition system based on artificial intelligence according to claim 2, wherein an attention mechanism module and a residual error connection module are added to the improved convolutional neural network model based on a traditional convolutional neural network, the attention mechanism module adopts a mode of combining a channel attention mechanism and a space attention mechanism and is used for highlighting important characteristic information, and the residual error connection module is used for solving a gradient vanishing problem in a deep network training process.
  4. 4. The endoscopic gastrointestinal tract image detection and recognition system based on artificial intelligence according to claim 3, wherein the target detection algorithm adopted by the lesion detection module is an improved YOLOv algorithm, and the detection capability of a micro lesion area is improved by optimizing a backstone network of the YOLOv algorithm to increase the number of feature fusion layers.
  5. 5. The endoscopic gastrointestinal tract image detection and recognition system based on artificial intelligence according to claim 4, wherein the classification model in the lesion recognition module adopts an integrated learning model combining a support vector machine and a random forest, and realizes accurate recognition of multiple lesion types by fusion analysis of refined features of a lesion region.
  6. 6. The system for detecting and identifying the gastrointestinal tract image under the endoscope based on the artificial intelligence of claim 5, wherein the training data set and the verification data set constructed by the model training module are marked by a professional doctor, marking information comprises position coordinates and lesion types of a lesion area, the model is trained in a cross verification mode in the model training process, and model parameters are optimized through an adaptive learning rate adjustment algorithm.
  7. 7. The endoscopic gastrointestinal tract image detection and recognition system based on artificial intelligence according to claim 6, wherein the visual output mode of the result output module comprises the steps of carrying out bounding box marking and color highlighting marking on a lesion area on an original image, and simultaneously generating a text report containing lesion positions, lesion types, confidence degrees and suggested treatment schemes.
  8. 8. The endoscopic gastrointestinal tract image detection and recognition system based on artificial intelligence of claim 7 is characterized in that the data storage module adopts a distributed storage architecture, supports efficient storage and quick reading of massive image data and model data, encrypts the stored data, and ensures safety and privacy of the data.
  9. 9. The endoscopic gastrointestinal tract image detection and identification method based on artificial intelligence is characterized by comprising the following steps of S1, image acquisition, S2, image preprocessing, S3, feature extraction, S4, lesion detection, S5, lesion identification, cutting and refining feature extraction on a lesion area, identification of a lesion type through a classification model, S6, result output, integration of detection identification results, generation of a visual report and output, S7, model optimization, evaluation of detection identification performance, and retraining of an optimization model by using newly added labeling data if a preset threshold is not reached.
  10. 10. The method for detecting and identifying the endoscopic gastrointestinal tract image based on artificial intelligence according to claim 9, wherein the training process of the improved convolutional neural network model in the step S3 comprises the steps of dividing the preprocessed image data into a training set and a verification set, initializing model parameters, setting training super-parameters, inputting the training set data into the model for forward propagation, calculating loss function values, updating the model parameters through a backward propagation algorithm, verifying model performance through the verification set data, repeating the training process until the model converges or reaches a preset training number, and the evaluation indexes adopted in the step S7 comprise accuracy, recall rate, precision and F1 value.

Description

Endoscopic gastrointestinal tract image detection and recognition system and method based on artificial intelligence Technical Field The invention relates to the technical field of medical image detection, in particular to an endoscopic gastrointestinal tract image detection and identification system and method based on artificial intelligence. Background Gastrointestinal tract diseases are common digestive system diseases, such as gastritis, gastric ulcer, intestinal polyp, gastric cancer, intestinal cancer and the like, early accurate detection and identification are critical to treatment and prognosis of the diseases, endoscopy is an important means for diagnosing the gastrointestinal tract diseases, the mucous membrane form inside the gastrointestinal tract can be directly observed through an endoscope, related image information is obtained, traditional endoscope image detection and identification mainly depends on naked eye observation and experience judgment of doctors, various defects exist, algorithms such as deep learning and the like show good application prospects in the field of medical image identification along with rapid development of artificial intelligence technology, and the artificial intelligence technology is applied to endoscopic gastrointestinal tract image detection and identification, so that doctors can be assisted to improve diagnosis efficiency and accuracy. However, the existing endoscope image detection and recognition technology based on artificial intelligence also has the problems of poor model generalization capability, insufficient recognition precision on tiny lesions, incapability of effectively processing image noise and illumination change and the like. Disclosure of Invention The invention aims to provide an artificial intelligence-based endoscopic gastrointestinal tract image detection and recognition system and an artificial intelligence-based endoscopic gastrointestinal tract image detection and recognition method, which can solve the problems that a model has poor generalization capability, the recognition precision on tiny lesions is insufficient, and image noise and illumination change cannot be effectively processed. In order to solve the problems, the following technical scheme is provided: The system comprises an image acquisition module, an image preprocessing module, a feature extraction module, a lesion detection module, a lesion recognition module, a model training module, a result output module, a data storage module, a man-machine interaction module and a quality evaluation module, wherein the image acquisition module is used for acquiring image data in the gastrointestinal tract and transmitting the image data to the image preprocessing module, the image preprocessing module is used for preprocessing the image data, the feature extraction module is used for extracting deep semantic features of an image by adopting an improved convolutional neural network model, the lesion detection module is used for detecting a lesion area based on the deep semantic features and determining position information, the lesion recognition module is used for carrying out feature refinement extraction on the lesion area and recognizing a lesion type, the model training module is used for training and optimizing a related model, the result output module is used for integrating detection recognition results and visualized output, the data storage module is used for storing various data, the man-machine interaction module is used for realizing interaction operation between a user and the system, and the quality evaluation module is used for evaluating the detection recognition performance and feedback optimization of the system. According to the technical scheme, a set of full-flow intelligent detection and identification system from image acquisition to result output is constructed through the cooperation of the multiple modules, the modules are definite in labor division and efficient in data transmission, the accurate processing and analysis of the endoscopic gastrointestinal tract images can be realized, and the defects of the traditional manual diagnosis and the existing AI diagnosis technology are effectively overcome. Further, the preprocessing operation of the image preprocessing module comprises image denoising, image enhancement, image size standardization and image normalization, wherein the image denoising adopts an adaptive median filtering algorithm, and the image enhancement adopts a Retinex algorithm. According to the technical scheme, the problems of image noise and illumination variation are solved in a targeted manner by combining a plurality of preprocessing algorithms, the size of a filtering window can be adaptively adjusted by the adaptive median filtering algorithm according to the density of noise in an image, gaussian noise and spiced salt noise are effectively removed, meanwhile, image detail information is reserved, the Retinex algorithm is used fo