Computer-aided Recognition Algorithm For High Incidence Of Esophageal Cancer Image In Xinjiang

Objective: The Kazak nationality is the high-risk group in Xinjiang Uygur Autonomous Region.This paper studied the algorithm and technologies of computer-aided diagnosis for esophageal cancer based on X-ray images,to verify the feasibility and rationality.Meanwhile the use of computer-aided diagnosis technology of medical X-ray images can help doctors identify lesions based on the type of reducing the workload of doctors and improve the diagnostic quality.Methods: The experiments were conducted in the MATLAB platform.Firstly,the regions of interest were selected under the guidance of the radiologist.And the preprocessing methods,including median filter and histogram equalization,were applied on the X-ray images.Secondly,using Ostu image segmentation method to proceeding segmentation for preprocessing image.Through gray level co-occurrence matrix,Hu invariant moments,gray-level histogram and wavelet transform were employed to extract the image features.Thirdly,using principal component analysis method reduced the dimensionality of the features with strong classification ability,to eliminate the redundancy of information between different features.In addition,using the BP neural network method to classify the images,this paper define two classifier for the image classification,using the independent-samples T test and principal component analysis to classify the normal esophagus images and the non normal esophagus images;using the one-way analysis of variance and principal component analysis to classify the fungating type,infiltrating type and ulcerative type esophageal images.And the classification performance was evaluated by the classification accuracy and kappa number.Results:(1)preprocessing image method can improve the quality of the images and obtain the clearly edge of images.(2)This paper was proposed to segmentation method,it can segment intact images,and retain integralityoriginal image;(3)Experimental results show that29 features are extracted based on the algorithm above from normal esophagus and non normal esophagus images;(4)When using the independent-samples T test and principal component analysis to classify,according to using the independent-samples T-test to select 19 the features with the best classification performance.The top-five principal components are selected for their cumulative percent reached to 88.085%.Other method,it uses the one-way analysis of variance and principal component analysis to classify.Firstly,selecting the one-way analysis of features,a total of nineteen were obtained.Then,using the principal component analysis calculated the feature,so it got to five numbers and cumulative percent is 87.537%.(5)Then using BP neural network method to classify normal esophagus and non normal esophagus images.The experimental results show that when the hidden layer of seven,feature selected method of the average classification accuracy outperforms other methods,recognition rate of normal esophagus,the non normal esophagus were 97.130%,98.620%,the recognition effect is better.When using BP neural network method to classify three types of esophageal images,the experimental results show that the average classification accuracy of the fungating type,infiltrating type and ulcerative type esophageal images reached to 98.000%,96.000%,98.500%,respectively.Conclusion: Preprocessing and image segmentation method are the key steps in the medicine image processing.This paper selected preprocessing and threshold segmentation algorithm is not only to image quality improved,but also keep a complete set of the target area.Then after processed image,using principal component analysis combined with independent-samples T test and the one-way analysis of variance to select the classification ability strong feature value through using different feature extraction methods to get the characteristic value.Finally using the BP neural network method to classify and select the normal esophagus,fungating type,infiltrating type and ulcerative type esophageal images,the proposed methods achieve high classification performance and the algorithm is reasonable and feasible,which can provide the valuable reference to radiologists and lay a foundation for computer diagnosis system of esophageal cancer for Kazak nationality.