< Back to previous page

Project

Coding of body and body parts by macaque inferior temporal neurons.

Perception of bodies is an important process which animals (including non-human primates and humans) perform effortlessly every day. The information which is decoded from visual inspection of bodies can be instrumental for survival – e.g. distinguishing bodies of dangerous animals from conspecifics. Furthermore, analysis of body posture can provide various non-verbal signals useful for decoding the emotional state, intentions, attitude, etc.

Despite the importance of body perception, the neural mechanisms underlying it are not fully understood yet. That is why the main purpose of this thesis was to shed more light on how body perception is performed in the brain.

First, we localized the brain regions involved in representing bodies in four rhesus monkeys by means of functional Magnetic Resonance Imaging (fMRI). We found two body patches in the middle and anterior part of the Superior Temporal Sulcus (STS) bilaterally, which we label as the (midSTS and antSTS body patches, respectively). Those patches were very close to the face patches discovered previously in the inferotemporal (IT) cortex. We then showed 200 stimuli of different classes (such as bodies, faces, and objects) and studied how they were represented in the fMRI activation pattern. Surprisingly, both the midSTS and antSTS body patch showed a representational structure dominated by the face-non-face distinction, which was probably due to their close proximity to the face patches plus the limited spatial resolution of fMRI.

Second, we targeted the left midSTS body patch of two of the monkeys we used in the fMRI study with electrodes for single unit recordings. The monkeys had to passively view 100 stimuli of different classes (half of those employed in the imaging study) while we were recording the action potentials of single neurons. The population of neurons responded on average higher to bodies than to other stimuli (including artificial objects and faces), however individual neurons had heterogeneous responses, responding only to few body exemplars but never to all bodies. Thus, we demonstrated a high within-class selectivity for the midSTS body patch neurons.

Third, we studied the tolerance of midSTS body patch neurons to different image transformations which preserve the stimulus identity. We showed that most of the neurons were tolerant to location and scale changes. Most neurons were also preserving their selectivity to silhouette and outline version of the images, which preserve the shape but eliminate the cues for texture and shading. However, the midSTS body patch neurons were sharply tuned for a particular in-plane orientation of the stimuli, i.e. the cells were not tolerant to changes in planar rotation.

Finally, we attempted to reveal the body features which drive the response in the midSTS body patch neurons using the Bubbles technique (Gosselin & Schyns, 2001). We showed trial unique stimuli in which randomly placed Gaussian apertures sampled different parts of the image. Then we reverse-correlated the images with the responses they elicited. We demonstrated that the neurons responded to relatively small fragments usually related to the limbs of the bodies or the outer segments of the non-bodies. Those responses were tolerant to translation and size changes. Moreover, the features exciting the cells were usually similar across different effective stimuli.

In conclusion, fMRI mapping showed two distinct body-category selective patches in macaque IT cortex. Single unit recordings in one of these patches showed a greater average response to images of bodies compared to faces and objects. The population of neurons in this midSTS body patch could discriminate between images of bodies and other objects (including faces), however the individual neurons showed profound within-class selectivity. This selectivity most probably originates from the fact that midSTS body patch neurons respond to oriented shape features, such as extremities and curved segments, which are present in some but not all bodies.

Date:8 Oct 2009 →  6 Oct 2015
Keywords:Visual perception², Neurophysiology, Functional magnetic resonance imaging, Electrophysiology, Visual cortex, Coding of body and body parts, Object recognition
Disciplines:Laboratory medicine, Palliative care and end-of-life care, Regenerative medicine, Other basic sciences, Other health sciences, Nursing, Other paramedical sciences, Other translational sciences, Other medical and health sciences
Project type:PhD project