Seven-layer deep neural network based on sparse autoencoder for voxelwise detection of cerebral microbleed

• Published in: Multimedia tools and applications Vol. 77; no. 9; pp. 10521 - 10538
• Main Authors: Zhang, Yu-Dong, Zhang, Yin, Hou, Xiao-Xia, Chen, Hong, Wang, Shui-Hua
• Format: Journal Article
• Language: English
• Published: New York Springer US 01.05.2018; Springer Nature B.V
• Subjects: Artificial neural networks; Brain; Computer Communication Networks; Computer Science; Computer simulation; Data Structures and Information Theory; Multimedia Information Systems; Neural networks; Special Purpose and Application-Based Systems; Voxelwise detection; Accuracy paradox; Deep neural network; Sparse autoencoder; Cerebral microbleed
• ISSN: 1380-7501; 1573-7721
• DOI: 10.1007/s11042-017-4554-8

In order to detect the cerebral microbleed (CMB) voxels within brain, we used susceptibility weighted imaging to scan the subjects. Then, we used undersampling to solve the accuracy paradox caused from the imbalanced data between CMB voxels and non-CMB voxels. we developed a seven-layer deep neural network (DNN), which includes one input layer, four sparse autoencoder layers, one softmax layer, and one output layer. Our simulation showed this method achieved a sensitivity of 95.13%, a specificity of 93.33%, and an accuracy of 94.23%. The result is better than three state-of-the-art approaches.