Abstract: Fuzzy clustering algorithms suffer from some weakness. The main weakness including the inclination to be trapped in local optima and vulnerable to initialization sensitivity. This study proposed a new approach called (FFCM) to solve Fuzzy C-Means (FCM) initialization problem using firefly algorithm to find optimal initial cluster centers for the FCM, thus improve all applications related fuzzy clustering such as image segmentation. The new approach (FFCM) has been evaluated in MRI Brain segmentation problem using simulated brain dataset of McGill University and MRI real images from IBSR center benchmark datasets. The experiments indicate encouraging results after applying (FFCM) and compared the outcomes with FCM random initialize cluster center.