Mass detection based on pooled mass probability map of 3D reconstructed slices in digital breast tomosynthesis

Abstract

In this paper, we propose a novel approach for automated detection of breast masses in three-dimensional (3D) reconstructed slices on digital breast tomosynthesis (DBT). The 3D reconstructed slices provide quasi-3D information with limited resolution along the depth direction due to insufficient sampling in depth direction. This problem could cause an error of general 3D segmentation approaches which have to process information with different resolution at the same time. In order to resolve the problem, this paper proposes an effective mass detection method based on pooled mass probability map. The proposed pooled mass probability map contains slice plane information by fusing mass probabilities of initially detected regions along slices. Extensive and comparative experiments have been conducted using clinical data set to validate the effectiveness of proposed mass detection approach. Experimental results demonstrate the feasibility of proposed pooled mass probability map based approach for detecting masses on 3D reconstructed slices.