A Pathological Brain Detection System Based on Extreme Learning Machine Optimized by Bat Algorithm.

Siyuan Lu,Xin Qiu,Jianping Shi,Na Li,Zhi-Hai Lu,Peng Chen,Meng-Meng Yang,Fang-Yuan Liu,Wen-Juan Jia,Yudong Zhang
DOI: https://doi.org/10.2174/1871527315666161019153259
2017-01-01
Abstract:AIM:It is beneficial to classify brain images as healthy or pathological automatically, because 3D brain images can generate so much information which is time consuming and tedious for manual analysis. Among various 3D brain imaging techniques, magnetic resonance (MR) imaging is the most suitable for brain, and it is now widely applied in hospitals, because it is helpful in the four ways of diagnosis, prognosis, pre-surgical, and postsurgical procedures. There are automatic detection methods; however they suffer from low accuracy.METHOD:Therefore, we proposed a novel approach which employed 2D discrete wavelet transform (DWT), and calculated the entropies of the subbands as features. Then, a bat algorithm optimized extreme learning machine (BA-ELM) was trained to identify pathological brains from healthy controls. A 10x10-fold cross validation was performed to evaluate the out-of-sample performance.RESULT:The method achieved a sensitivity of 99.04%, a specificity of 93.89%, and an overall accuracy of 98.33% over 132 MR brain images.CONCLUSION:The experimental results suggest that the proposed approach is accurate and robust in pathological brain detection.
What problem does this paper attempt to address?