Brain decoding or brain machine interface (BMI), to read out a stimulus given to or a mental state of human participants from measurable brain activities by means of machine learning techniques, has made a great success in recent years. Due to the large variation of brain activity between subjects, however, previous decoding/BMI studies mostly put focus on developing an individual-specific decoder. For making brain decoding or BMI more usable, in this study, we explored to seek subject-independent features within a large-scale brain activity database collected by fMRI measurements. Our feature extraction relied on deep learning, which is one of the most powerful techniques in data mining and has achieved a great success in data science very recently. The deep learning with a layered neural network architecture successfully acquired subject-independent but task-specific features in the hidden layer from the large-scale fMRI dataset, leading to higher decoding accuracy than those by other conventional methods. Our result indicated that the deep learning is useful for developing subject-independent BMI as well as elucidating neural correlates common to the individuals. In this study, we used the following open fMRI database:
Data were provided by the Human Connectome Project, WU-Minn Consortium (Principal Investigators: David Van Essen and Kamil Ugurbil; 1U54MH091657) funded by the 16 NIH Institutes and Centers that support the NIH Blueprint for Neuroscience Research; and by the McDonnell Center for Systems Neuroscience at Washington University.