Abstract：Objective To explore brain MRI image segmentation algorithm based on hidden markov random field (HMRF) and conjugate gradient (CG) algorithm. Methods Hidden Markov random field was used to simulate the objective function minimization
during brain MRI image segmentation, and the conjugate gradient algorithm was used to solve the objective function parameters.Next, MRI images of IBSR brain gallery clinical examples and artificially synthesized images of BrainWeb brain gallery were
selected for simulation experiments. Finally, the accuracy of segmented images was evaluated by Dice similarity coefficient and specific value. Results Based on this study, the edge of white matter, gray matter and cerebrospinal fluid in the segmentation image
was complete and clear, and the dice similarity coefficient and specificity value were higher than other algorithms. Among them, the average dice similarity coefficient and specificity value of IBSR brain map database were increased by 5.87%~6.67% and
3.17%~8.16%, respectively. Moreover, the average dice similarity coefficient of brain segmentation results of BrainWeb Library with different noise levels increased by 38.15%~1.42% and the specificity increased by 3.17%~8.16% compared with the Improved
Markov random field algorithm. Conclusion In this study, the proposed algorithm can accurately divide the brain MRI image regions, with strong anti noise and robustness.
居敏，薛丽君，朱建新. 基于隐马尔可夫随机场和共轭梯度算法的脑部MRI图像分割算法研究[J]. 中国医疗设备, 2020, 35(12): 95-98.
JU Min, XUE Lijun, ZHU Jianxin. Research on Brain MRI Image Segmentation Algorithm Based on Hidden Markov Random Field and Conjugate Gradient Algorithm. China Medical Devices, 2020, 35(12): 95-98.
