Automated Atlas-Based Segmentation of Breast Region in 3-D Magnetic Resonance Imaging (MRI) Using FCM Method
Parinaz Eskandari, Shahriar B. Shokouhi · 2021
In this paper, an automated atlas-based segmentation approach for segmenting breast region in 3-D Magnetic Resonance Image (MRI) is proposed. This algorithm combines probabilistic atlas and atlas selection approaches and utilizes Fuzzy C-Means (FCM) classifier to classify atlases. Due to fuzzy nature of mentioned classifier, each pixel belongs to different classes by different membership values and sum of membership values which is derived from probabilistic theory, equals to 1. So, sum of the membership values for all classes of each data never equals to 0. Thus, there are not any empty classes. Proposed algorithm by utilizing 2240 MR images which are related to 40 women and are imaged in axial position, was evaluated. To represent similarities between automatic and manual segmentation results, average of Dice coefficients, average of Jaccard coefficients, average of FN and average of FP were calculated and obtained as 0.9166, 0.8481, 0.0816 and 0.0457, respectively.