Parallel implementation for 3D medical volume fuzzy segmentation.

• This research introduces the hybrid optimization of both CPU and GPU for image processing. • A novel 3D medical volume segmentation is proposed. • Modified FCM versions are used to increase the accuracy of volume segmentation. • Volumetric neighborhood has been considered to validate the actual 3D FCM. In the past, 2D models were the main models for medical image processing applications, whereas the wide adoption of 3D models has appeared only in recent years. The 2D Fuzzy C-Means (FCM) algorithm has been extensively used for segmenting medical images due to its effectiveness. Various extensions of it were proposed throughout the years. In this work, we propose a modified version of FCM for segmenting 3D medical volumes, which has been rarely implemented for 3D medical image segmentation. We present a parallel implementation of the proposed algorithm using Graphics Processing Unit (GPU). Researchers state that efficiency is one of the main problems of using FCM for medical imaging when dealing with 3D models. Thus, a hybrid parallel implementation of FCM for extracting volume objects from medical files is proposed. The proposed algorithm has been validated using real medical data and simulated phantom data. Segmentation accuracy of predefined datasets and real patient datasets were the key factors for the system validation. The processing times of both the sequential and the parallel implementations are measured to illustrate the efficiency of each implementation. The acquired results conclude that the parallel implementation is 5X faster than the sequential version of the same operation. [ABSTRACT FROM AUTHOR]