Research On Image Detection Technology Of Esophageal Cancer Based On Deep Learning

Esophageal cancer is a common malignant tumor in life,which seriously affects human health.In China,the incidence of esophageal cancer ranks the forefront of world and a large number of new cases of esophageal cancer emerge each year.Up to now,the main process of recognition and diagnosis of esophageal cancer is that doctors could observe the image on the screen by using the electronic gastroscope,which reflects the in-vivo image data on the screen.As growing numbers of patients,the pressure and intensity of the doctors is increasing.On that account,this paper studies the method of constructing the object detection and recognition system by computer vision,which aims at relieveing the pressure and intensity of the doctors.With the development of machine learning and deep learning,deep convolutional neural network has been widely used in image classification and object detection.The object detection algorithms based on deep learning have been studied,and the Faster RCNN object detection algorithm has been analyzed.For each stochastic gradient descent mini-batch sampling,it implemented randomly network training through the setting of ratio value parameter for positive and negative samples of Faster RCNN algorithm.In this paper,the online hard example mining(OHEM)mechanism is introduced into the Faster RCNN algorithm,so as to eliminate this ratio of parameter by automatically selection.It can be seen from the contrast experimental analysis in esophageal cancer image dataset that the introduction of the online hard sample mining mechanism into the Faster RCNN algorithm is able to efficiently improve the detection accuracy to a certain extent.Finally,an esophageal cancer image detection system is implemented on the Linux system using the improved Faster RCNN algorithm we have proposed,which detects the image automatically through the browser.