3D Brain Image Segmentation Model using Deep Learning and Hidden Markov Random Fields
El-Hachemi Guerrout, Ramdane Mahiou, Anfel Melouk, Ines Harmali · 2020
In this paper, we present a new brain image segmentation model (DL-HMRF) using deep learning and hidden Markov random fields (HMRFs). That segments the brain images into different tissue types (Grey Matter, White Matter, and Cerebro Spinal Fluid) in order to help physicians to determine easily the final decision. The underlying method is based on hidden Markov random fields to extract image features for using them as inputs of our neural network model. IBSR images with their ground-truth are used to train and evaluate our model. The quality of segmentation is measured by using the Dice coefficient metric. Setting the parameters of the DL-HMRF model is a task in itself. We have generated six models with different parameters to select the one that gives the best segmentation. Through tests, the selected model shows excellent results in terms of the quality of segmentation and execution time.