Multi-Level Fusion in Ultrasound for Cancer Detection based on Uniform LBP Features

• DOI: 10.32604/cmc.2021.013314
• Publication Date: 2021-01-04T08:12:18Z
• Authors: Diyar Qader Zeebaree; Adnan Mohsin Abdu...; Dilovan Asaad Zebari; Habibollah Haron; Haza Nuzly Abdull...
• Keywords: Local Binary Patterns; Speckle noise; Feature (linguistics); Breast ultrasound

Collective improvement in the acceptable or desirable accuracy level of breast cancer image-related pattern recognition using various schemes remains challenging. Despite combination multiple to achieve superior ultrasound image by reducing speckle noise, an enhanced technique is not achieved. The purpose this study introduce a features-based fusion scheme based on enhancement uniform-Local Binary Pattern (LBP) and filtered noise reduction. To surmount above limitations aim study, new descriptor that enhances LBP features threshold has been proposed. This paper proposes multi-level for auto-classification static images cancer, which was attained two stages. First, several were generated from single pre-processing method. median Wiener filters utilized lessen enhance texture. strategy allowed extraction powerful feature overlap between benign malignant classes. Second, mechanism production diverse different images. feasibility LBP-based texture categorize demonstrated. effectiveness proposed tested 250 comprising 100 150 images, respectively. method achieved very high (98%), sensitivity specificity (99%). As result, process can help decision produced improved results terms accuracy, sensitivity, specificity.