Improve the Quality of Mammogram Images by Image Processing Techniques

Subject Areas : Intelligent Multimedia Processing and Communication Systems (IMPCS)

Mahdi  Hariri ¹ (Assistant Professor, University of Zanjan, Zanjan, Iran)
Hassan  Najafy ² (1Graduated Masters Department Electrical engineering,Islamic Azad university of Zanjan Branch, Iran)

Received: 2021-10-02 Accepted : 2022-11-10 Published : 2022-03-21

Keywords: Mammogram images, Curvelet Transform, Feature extraction, Contourlet Transform, quality improvement,

Abstract :

Abstract: Due to the spread of breast cancer, its early detection from mammogram images using computerized methods have been considered an effective method to reduce the death rate of patients.In this research, a method based on image processing techniques is presented to improve the quality of mammography images. Therefore, in this research, we try to improve the quality of mammography images with image processing techniques to create a medical system. The research has two stages of pre-processing, including equalizing the dimensions and adjusting the histogram of the images, and a stage of feature extraction using Contourlet and Curlet transforms from mammography images, which provides three categories of morphological and histological, statistical, and frequency features to improve diagnosis. And it increases the accuracy of diagnosis. The proposed improvement method was implemented on the MIAS dataset and a subset of the extracted features was selected for the input of the classifier. Comparing the performance of the proposed method on different classifications, this method shows an accuracy rate of 86.3, which is a better result than other methods.MethodThis research is looking for a method that can improve the accuracy of the final diagnosis. Therefore, after the pre-processing stage, which includes rescaling and adjusting the texture of the image, highlighter transforms in the frequency domain such as Curvelet and Contourlet are used to highlight and increase the differentiation of areas with masses in the image for decision-making.Local features based on image zoning are used for image segmentation, and these methods are also used to increase the contrast of mammogram images concerning their surroundings. The improvement methods used in this research use features based on the wavelet domain.ResultsThe input image to the system is subjected to the feature extraction process and three main categories of frequency, morphology, and histology features are extracted from it. This process is done through the cycle, size equalization, histogram adjustment, contourlet transform, and curvelet transform. Due to the sensitivity of the systems, it has been tried to extract the features with various levels and known matrices such as the matrix of events and gray groups.The classification results evaluated the said method. The best results on the data set were the proposed method, which reached an accuracy rate of 86.3 and showed a good improvement on the displayed data set.

References:

Wang, G. and Liao, T.W.(2002). Automatic Identification of Different Types of Welding Defects in Radiographic Images. NDT & E International,35:519-528.

[10] Shafeek, H.I., Gadelmawla, E.S., Abdel-Shafy, A.A. and Elewa, I.M. (2004). Assessment of Welding Defects in Pipeline Radiographs Using Computer Vision. NDT & EInternational,37:291-299.

[11] Gonzalez, R.C. and Wintz, P.(1987). Digital Image Processing. Addison-Wesley, 2nd Edition.

[12] Castleman, K.R.(1979). Digital Image Processing. Englewood Cliffs N.J., Prentice Hall, 90-94.

[13] Addison, Paul S. "Introduction to redundancy rules: the continuous wavelet transform comes of age." (2018): 20170258.

[14] Candes, E. J., and Donoho, D. L, “Continuous curvelet transform”, II. Discretization and frames: Applied and Computational Harmonic Analysis, 19, 198-222, 2003.