Abstract：Objective This study aimed to present a novel unsupervised method for MR brain image segmentation based on self-organizing maps (SOMs) and genetic algorithms (GAs). Methods In particular, the proposed method was based on five stages consisting of image preprocess, extracting first and second order statistical features, feature selection using evolutionary computation, voxel classification using SOM, and entropy-gradient (EG) clustering. Results Both simulated and clinical datasets were evaluated by different methods. Qualitative analysis showed that the components of the white matter, gray matter and cerebrospinal fluid were well preserved, and globally all the regions were correctly classified. Quantitative evaluation results showed that the genetic algorithms can achieve optimized feature set than principal component analysis (PCA). EG had a statistical significance difference with K-means (P<0.01). Our algorithm outperformed the CGMM method in WM and CSF delineation, and was found to be the most effective among other techniques. Conclusion The complete procedure does not use any a priori knowledge regarding voxel class assignment, but reveals a fully unsupervised, automated method for MRI segmentation to directly identify different tissue classes, which can provide better robustness, superiority, and pervasiveness in clinical applications.
