Research And Exploration Of Artificial Intelligence Technology In Identifying Endoscopic Sites

Chapter Ⅰ To explore the clinical application of Transformer model in location recognition in gastroscopic images.Purpose Transformer is a deep learning technology model,which has the ability to learn and analyze image features and obtain the functions of images recognition and prediction.The purpose of this research is to form an artificial intelligence system with the function of gastroscope images recognition and prediction by learning the location features of gastroscope images using Transformer model,and to evaluate the accuracy,specificity and overall effectiveness of Transformer model in upper gastrointestinal tract recognition.Research Content And Results 21782 gastroscopic images were collected from August 2018 to April 2020 in Nanfang Hospital of Southern Medical University and other hospitals,then the locations were marked.18640 gastroscopic images were selected to form a training set.The Transformer model is based on the training set to learn and extract the features of the gastroscope images to generate an artificial intelligence system.The system has the function of identifying and predicting the upper digestive tract in the gastroscope images.3142 gastroscope images are verification set(7282 site tags).The experimental group is an artificial intelligence system based on gastroscope images,and the control group is artificial routine labeling.Based on the verification set,the accuracy,specificity and overall effectiveness of the experimental group and the control group were evaluated,and the completion time was recorded.The results showed that the accuracy,specificity and overall effectiveness of the experimental group were 83.4%,66.9%and 77.8%,respectively,and the completion time was 9 minutes and 30 seconds.In the control group,the overall accuracy,specificity and overall effectiveness were 78.4%,68.9%and 73.8%,respectively,and the completion time was 10 hours,15 minutes and 30 seconds.Conclusion Compared with the control group,the artificial intelligence system formed by Transformer model based on gastroscope images location learning has obvious advantages in accuracy and overall effectiveness,and the time-consuming of the experimental group is significantly shorter.Chapter Ⅱ To explore the clinical application of Transformer model in real-time location recognition in gastroscope videos.Purpose Furthermore,based on gastroscope videos and images learning,the Transformer model is used to form gastroscope video intelligent system,gastroscope video&image intelligent system and gastroscope image intelligent system respectively.By the real-time recognition of upper digestive tract in the gastroscope videos of the test set,we can evaluate the accuracy,specificity and overall effectiveness of these systems.Research Content And Results Using 50 gastroscope videos recorded by Shenzhen second people’s Hospital from May to July in 2021 as research data,frames were drawn to form a "video image set",of which 40 were training set 1 and 10 were test set.The "gastroscope image set" containing 25525 gastroscope images was used as the control group to form the training set 2.Transformer model is based on training set 1,training set 2 and training set 1&training set 2 to form "video-based intelligent system","image-based intelligent system" and "video&image-based intelligent system" respectively.Through the test set,the accuracy and specificity of video intelligent system,image intelligent system,video and image intelligent system are compared.The results showed that the accuracy,specificity and overall effectiveness of video-based intelligent system in experimental group 1 were 84.3%,78.9%and 81.5%,respectively,while those in experimental group 2 which was video&image-based intelligent system,were 82.9%,81.5%and 82.2%,respectively.The accuracy,specificity and overall effectiveness of the intelligent system based on gastroscope images in the control group were 80.0%,76.8%and 78.4%,respectively.Conclusion The video learning effect of Transformer model was better than learning effect of gastroscope images.And the effect of video&image learning of Transformer model was better than that of video learning alone.