Authors:

Abstract

Breast cancer is recognized as one of the leading causes of death in most western countries as well as in Sri Lanka. Recent studies have proved that this is a disease that not only among women but also found in men. However, breast cancers can be cured by detecting them in their early stages. Mammograms are the primary medical imaging mechanism to diagnose breast cancer. Due to various artefacts, mammogram images must be enhanced before the analysis. Although some sophisticated enhancement algorithms are available, most algorithms are either unsuitable for mammogram enhancement or highly complex. To address these issues, this study proposes a hybrid method for mammogram image enhancement. It uses an adaptive histogram equalization technique followed by Haar wavelet transformation for edge enhancement. Mammograms are further enhanced with a few morphological operations. The adaptive histogram equalization method is used considering the features of the mammogram images and its low time complexity. The proposed method is adaptive because it uses a probability distribution and a set of control parameters to predict the suitable intensity levels for the enhancement process. Haar wavelet enhancement maximizes the edge details in the mammograms to detect clean contours. The wavelet transformation is very precise with noisy images like mammograms, and it helps to enhance dense areas and emphasis the region of interest. Stages in the enhancement are carefully planned to have improved mammograms for the detection phase. Finally, the enhanced images are used for the detection of cancerous regions. The detection phase is carried out using state-of-the-art Viola-Jones features and a Cascade classifier. The effectiveness of the proposed enhancement technique was determined by calculating the absolute mean brightness error test (AMBE) and the cancer detection accuracy on the well-known Mammographic Image Analysis Society (MIAS) database. An impressive AMBE value of 16.317 and a cancer detection accuracy of 85.1% were recorded. Therefore, experimental results prove that the breast cancer detection rate can be increased considerably with the proposed method compared to the existing mammogram enhancement methods.

Keywords: Adaptive histogram equalization, Haar Cascade classifier, Mammograms, Wavelet transformation, Breast carcinoma

How to Cite: Amarasinghe, A. S., & Nawarathna, R. D. (2022). Wavelet and histogram based mammogram enhancement for abnormality detection. Ceylon Journal of Science, 51(4), 347–357. DOI: http://doi.org/10.4038/cjs.v51i4.8052