Breast cancer Diagnosis using Waveletscattering transform and hierarchy Multi-Layer Percep۱۲۸tron Neural Network

Background: Breast cancer has been the most commonly diagnosedcancer and the leading cause of cancer death in womenin the last decade. Despite the fact that many experimentshave been done in this regard, breast cancer has been one ofthe deadliest diseases the in all courses. Mammography is thewell-known screening examination for breast cancer, which decreasesthe death rate by ۳۰ to ۷۰ percent.However, because of the type of the breast tissue and using theX-ray radiation with low density, the images that achieved bymammogram are in low quality. Although, there are many differentmethods for diagnosis of breast cancer by mammogramimages, these methods have not been very successful in diagnosisof benign or malicious masses because of not using properclassifier and also proper extractors in order to extract the tissuefeatures perfectly.Materials and Methods: In this paper, we use features basedon image tissue such as scattering wavelet transform. using variouscharacteristics to have images with high qualities increasethe amount of the input data and in result makes the classificationvery challenging. Therefore, it is necessary to reduce thesize of the features appropriately and to improve the accuracyof the classification.Results: To achieve this, after the extracting of the features, inorder to reduce the problems size and to improve the accuracyof the classification, we use the principle component analysisand the linear discriminant analysis.at the end, we use the hierarchyMulti-Layer Perceptron Neural Network classificationin order to improvement of the accuracy in compare to the pastachievements.Conclusion: for further evaluation of the proposed method, weuse the Mini-MIAS and DDSM image collection and reaches tothe accuracy of ۹۷.۵۷ percent.

کلیدواژه ها:

Breast Cancer ، Scattering ، Principle ComponentAnalysis ، Multi-Layer Perceptron Neural Network