Segmentation of Relevant Region in Breast Histopathology Images using FCM with Guided Initialization

Abstract

This study proposes a modified initialization approach for the conventional FCM, namely FCM with guided initialization. The FCM with guided initialization was implemented to segment the relevant regions in the breast histopathology images. The initialization method to select initial centers is based on the Cyan (C) channel histogram. Area Overlap Measure (AOM) and Combined Equal Importance (CEI) were used to evaluate the performance of the proposed FCM with guided initialization. The obtained AOM and CEI for the overall dataset achieved promising results: 0.89 in AOM and 0.88 in CEI. When comparing the number of iterations required to complete the proposed FCM clustering algorithm, the FCM with guided initialization is found to be effective in reducing the search space by showing a lower number of iterations.