Research On Small Intestinal Disease Detection Algorithm Based On Capsule Endoscope

Small intestine is an important part of the human digestive system,and its health affects human health.The small intestine is in a special position in digestive system and its structure is complex.The traditional invasive gastrointestinal endoscopy has some limitations in checking the health of small intestine,and it will cause great pain to patients.Therefore,most hospitals recommend patients to use safe and painless Wireless Capsule Endoscope.However,the huge medical image data obtained by Wireless Capsule Endoscope is huge,which not only increases the workload of medical personnel for disease screening,but also delays the diagnosis and treatment of diseases Therefore,it is of great clinical significance to study a method for effectively and automatically detecting small bowel diseases from wireless capsule medical images With deep learning methods in the field of deepening research,and showing strong capabilities.An automatic detection method for small intestinal diseases based on deep learning was proposed in this paperFirstly,an intestinal disease detection method based on Faster Regional Convolution Neural Network(Faster RCNN)is presented.The method consists of feature extraction,candidate region suggestion and target location.In feature extraction,the ResNet50 network is used to get the features of small intestine image;on this basis,the candidate lesion areas are obtained in the form of sliding windows on the feature map;finally,target localization network is used to screen and locate the candidate lesion area.This method can effectively detect small intestinal mass(abnormal protuberance in the small intestine)and ulcerSecondly,a detection method of small intestinal hemorrhage based on sparse features is proposed.A Sparse Auto-Encoder with convolution layer was constructed to extract the sparse features of small intestine sample images without supervision Several classifiers were used to classify the sparse features of small intestine and the best classifier was selected.This method can effectively recognize small intestinal bleeding from a large number of small intestinal imagesFinally,a joint network of Faster RCNN and Sparse Auto-Encoder is constructed The joint network uses Faster RCNN to screen the occupancy and ulcer lesions in the samples,and then sends the remaining non-occupancy and non-ulcer samples to the Sparse Auto-Encoder to detect small intestinal bleeding.The hierarchical detection of small intestinal diseases has been completed,and the advantages of Faster RCNN and Sparse Auto-Encoder in small intestinal disease detection have been brought into play The experimental results show that the network can not only detect a variety of small intestinal diseases,but also has high sensitivity.In order to verify the validity of the experiment,this paper carried out a series of performance test experiments using the Wireless Capsule Endoscope images provided by the Eastern Theater General Hospital as a dataset.The experimental results show that the deep learning network model used in this paper can effectively detect small bowel diseases in Wireless Capsule Endoscopic images.