Diagnosis Of Celiac Disease Based On Strip Principal Component Analysis And Residual Recalibration Network

With the advent of the era of massive data,computer-aided decision-making systems(CAD)are integrated into the fields of image analysis,speech recognition,and emotion calculation with a low time consumption and high sensitivity.Traditional CAD systems excessively rely on prior key point selection,feature extraction and the classification of classifier.However,the feature extraction algorithm often plays an important role on the success or failure of the final CAD system.In recent years,deep learning and principal component analysis data processing technologies have achieved good benefits in data processing.However,for the complexity and diversity of the capsule endoscopy images that lie in the human small intestine,the current processing algorithms have certain limitations:The feature information lied in hidden layer that extracted by deep learning algorithm is redundant.How to ensure that the high-value information that is beneficial to the final task is efficiently selected in the neural network?(2)Computerized capsule endoscopy image analysis remains challenging when considering that the camera angle and illumination with respect to the small intestine wall are not controllable.In order to solve the above problems,this article made the following innovations:(1)A new local attention learning module is proposed for the diagnosis of celiac disease.This attention module not only focuses on the global channel information of the celiac disease images,but also takes the salient features of the local space into account.During network training,the weights of feature channels and spatial information are assigned to each convolution layer feature map to enriching the current valuable feature information and weaken unimportant ones.(2)A new strip principal component analysis algorithm based on non-greedy L1-norm maximization is proposed for the diagnosis of celiac disease.The villous atrophy of celiac disease is often accompanied with mucosal folds,cracks,and mosaic appearance.For t he special pathological pattern and the local images of celiac disease are usually highly correlated,we found that the strip principal component is more suitable for the analysis of celiac disease through control experiments.