Deep Net Predicts Sex From Brain Rhythms
Artificial neural network accurately classifies sex of subjects based on electroencephalography recordings.
Deep convolutional neural networks are powerful computational tools capable of learning from raw data, extracting subtle differences in apparently similar patterns. These artificial neural networks are especially well-suited for image classification and may even achieve superhuman performance. A study published Feb. 15 in Scientific Reports further highlights the versatility of deep learning, demonstrating its ability to predict sex from brain rhythms.
The researchers used electroencephalography (EEG) data -- recordings of electrical activity in the brain through the use of scalp electrodes -- collected from 1,308 subjects at different laboratories. The deep convolutional neural network predicted the sex of the subjects based on their EEG recordings with an accuracy of 81 percent. According to the authors, past efforts have failed at reliably extracting sex from visual or quantitative assessment of EEG, and this is the first study to explore deep learning for sex classification.
In a narrow context, the utility of this approach remains to be seen. But more broadly, the authors anticipate that deep learning might substitute or complement human-guided feature extraction and knowledge discovery in specialties including clinical neurophysiology, cardiology, intensive care medicine, psychiatry and neuropsychology.