DARPA uses brain analysis to identify images

The work takes advantage of the fact that the brain does a lot of work in filtering visual information before making the conscious mind aware of what has been seen. This can take the form of weeding out distractions, generating associations, and identifying features that may be worthy of further attention. All of these process happen even if images are only visible for fractions of a second.

The DARPA program funds Honeywell to work with academic researchers to listen in on the brain's internal conversation to help filter out those images that aren't worth detailed analysis. Trained analysts are fitted with electrodes that monitored their brain activity, and then shown stacks of 50 images in rapid succession.

Each image was visible for only 50-100 milliseconds, meaning the analysts could view many thousand images in the span of an hour (details of the work are available on pages 26-36 of this PDF). The researchers found that distinct patterns of electrical activity were apparent starting 250 milliseconds after a target image was visible. Images that trigger this signal can be tagged as meriting more careful analysis.