Mohammad Mehdi Movahedi - Department of Medical Physics and Engineering, School of Medicine, Shiraz University of Medical Sciences, Shiraz, Iran
Ali Zamani - Department of Medical Physics and Engineering, School of Medicine, Shiraz University of Medical Sciences, Shiraz, Iran
Hossein Parsaei - Department of Medical Physics and Engineering, School of Medicine, Shiraz University of Medical Sciences, Shiraz, Iran
Ali Tavakoli Golpaygani - Department of Biomedical Engineering, Standard Research Institute, Karaj, Iran
Mohammad Reza Haghighi Poya - Department of Medical Physics and Engineering, School of Medicine, Shiraz University of Medical Sciences, Shiraz, Iran

Background: Breast cancer is the second cause of death among women. Ultrasound (US) imaging is the most common technique for diagnosing breast cancer; however, detecting breast lesions in US images is a difficult task, mainly, because it provides low-quality images. Consequently, identifying lesions in US images is still a challenging task and an open problem in US image processing. This study aims to develop an automated system for the identification of lesions in US images Method: We proposed an automatic method to assist radiologists in inspecting and analyzing US images in breast screening and diagnosing breast cancer. In contrast to previous research, this work focuses on fusing information extracted from different frames. The developed method consists of template matching, morphological features extraction, local binary patterns, fuzzy C-means clustering, region growing, and information fusion-based image segmentation technique. The performance of the system was evaluated using a database composed of ۲۲ US videos where ۱۰ breast US films were obtained from patients with breast lesions and ۱۲ videos belonged to normal cases. Results: The sensitivity, specificity, and accuracy of the system in detecting frames with breast lesions were ۹۵.۷%, ۹۷.۱%, and ۹۷.۱%, respectively. The algorithm reduced the vibration of the physician’s hands’ while probing by assessing every ۱۰ frames regardless of the results of the prior frame; hence, lowering the possibility of missing a lesion during an examination. Conclusion: The presented system outperforms several existing methods in correctly detecting breast lesions in a breast cancer screening test. Fusing information that exists in frames of a breast US film can help improve the identification of lesions (suspect regions) in a screening test.

Automatic lesion detection, Breast lesion, Ultrasound imaging segmentation, Ultrasound video analysis