The neural fingerprint of human image processing

Our understanding of the human brain is hampered by the limitations we face with respect to the research techniques we can apply. Common non-invasive measurements of brain activity including electroencephalography and functional Magnetic Resonance Imaging (fMRI) provide only coarse information on ongoing neural activity because of a restricted spatiotemporal resolution, whereas noninvasive brain stimulation techniques such as transcranial magnetic stimulation provide insufficient spatial resolution to systematically alter the activity in particular subsets of neurons. Thus, the functional organization of the human brain remains very difficult to investigate at the scale of cortical columns, and to resolve this major hurdle in cognitive neuroscience a sub-millimetric investigation with millisecond precision in human subjects is needed. We propose a groundbreaking approach to unravel neuronal selectivity underlying cognitive processing by implementing invasive recordings in neurosurgical patients. Recordings from intracranial micro-electrodes will be combined with microstimulation to study the functional network underlying image processing. Intracranial data will be used to constrain deep neural network based encoding models, and the best models will be used to generate predictions. We complement microscale investigations with fMRI to integrate our findings in a larger network perspective.