Extraction of vessel structure in thermal images to help early breast cancer detection. Issue 1 (2nd January 2020)

• Record Type: Journal Article
• Authors: Hamidpour, Seyedeh Shahrbanoo Falahieh; Firouzmand, Mohammad; Navid, Mitra; Eghbal, Manouchehr; Alikhassi, Afsaneh
• Is Part Of: Computer methods in biomechanics and biomedical engineering. Volume 8:Issue 1(2020)
• Page Start: 103
• Page End: 108
• Publication Date: 2020-01-02
• Subjects: Breast cancer -- screening -- mammography -- thermal imaging -- breast vessel extraction algorithm; Imaging systems in biology -- Periodicals; Imaging systems in medicine -- Periodicals; Biomechanics -- Data processing -- Periodicals; Biomedical engineering -- Periodicals; 616.0757
• DOI: 10.1080/21681163.2019.1598895
• Languages: English
• ISSNs: 2168-1163

Abstract

ABSTRACT: Breast cancer screening plays a significant role in early detection and consequently higher rate of survival and full treatment of the cancer. Thermal imaging is a physiological imaging which is going to be an adjunctive modality and has become noticeable field of research. Breast thermography is non-invasive because of using no radiation and avoiding painful breast compression. The considerable characteristics of breast cancer feature in thermal images consist of structure, density and symmetric of breast vessels. There are rare methods can be found to extract breast vessel structure in thermal images so it cannot be compared with other methods, but many studies have been carried out on vessel extraction in retinal images. Because of importance of vessel structure to diagnosis malignancy in thermal images, in this paper, it is suggested a new method to segment it. The breast vessel extraction algorithm has been proposed based on retinal vessel extraction approaches. But at first, the gradient of images is amplified, images are enhanced and then breast vessel extraction is done by using filtering, thresholding and morphological operators. This method can distinguish hot regions from vessels and results show acceptable sensitivity, specificity and accuracy of proposed algorithm which are 98%, 95% and 96.5%, respectively.