Also can be used as a “Room 101″ Interrogation Technique….
GEEK
A team of Yale researchers, led by a then-undergraduate student, have made an astonishing step forward in brain science. The (perhaps unsettling) breakthrough allows scientists to use a medical imaging machine and a well-trained algorithm to visually reconstruct faces seen by test subjects. As seen below, their technique returns some results with a truly astonishing level of accuracy. Oddly, their results seem to have been possible specifically because the brain processes faces in such a unique and distributed way. This study takes the field’s greatest and most intractable problem and leverages it to truly impressive effect.
Faces have historically been very difficult to see via the brain. Ever since brain scientists first found our visual processor (the occipital lobe, at the back of the head), they have tried to read and interpret its activity to reconstruct visual data. They reasoned that a detailed-enough model for how each “pixel” we see appears in the visual cortex would allow a one-to-one reconstruction — but that’s only true some of the time. When viewing images like buildings or furniture, simple and inherently unemotional objects, we see mostly with our eyes. When we view a human face, on the other hand, we “see” it in both the visual and emotional brains, evaluate it on a visual and a personal level, and look at it through rotating lenses of trust, safety, sex, and more.
To account of this, to see what participants saw – not just what information their eyes took in — the team pared down a set of training photos to a much smaller list of Eigenfaces, simplified facial maps that covered the majority of variation in the training faces despite being far less numerous and complex. The program then recorded and cross-referenced patterns of neural excitations in response to its Eigenfaces — and it took those readings both in and out of the occipital lobe. When the experimental “test faces” were shown to the same participants, their neural responses were reconstructed into a new Eigenface that fit best, if you will.
This research used functional MRI scans (fMRI), which tracks neuro-excitation in real-time. The results are the first direct reading of visual information from outside of the visual cortex, and as can be seen in the figure above, it provided powerful results. Both the “occipital” and “non-occipital” constructions are inferior to the image using whole-brain information. By looking at areas like the ventral temporal cortex, they hope to get a more detailed view of perception than ever before. The technology aims to read the image in your mind’s eye, as opposed to that seen by your physical eyes alone.
Modern functional MRI machines can see with truly exquisite detail.
Oddly enough, reconstructions of this quality would probably be impossible in a topic less complex than facial analysis. We see faces with a uniquely discerning eye and process them in more detail than we do virtually anything else. This attention provides more excitation for researchers to study, and easily identifiable responses that can be easily picked from regular resting brain activity. Faces are checked against memory, emotion, reason, intuition, and more. That’s a lot of data for scientists to exploit.
As a result, we have here a model for constructing subjective experiences along with more objective visual information. This could help in seeing how a mentally ill or impaired person sees things, from schizophrenics to autistic people and beyond. Neuro-imaging already allows scientists to watch brain activity in patients of all types, but the information gained is purely locational — a burst of activity in a certain part of the brain might imply fear or panic in an autistic person, for instance. With a more holistic approach to visual processing, though, we could see how that abstract brain activity affects perception by the conscious mind.
Medical imaging scanners are becoming more and more powerful, seeing increasing levels of detail, but that detail is only going to be useful when plugged into a larger conception of the brain’s interconnected nature. In the past, we’ve engaged almost entirely in the act of “brain reading.” This is one of the first truly successful attempts to read the mind instead.
