Faces are complex and dynamic stimuli representing the most important visual stimuli in our social environment. Adult face expertise is characterized by the rapid and accurate discrimination of individuals and the ability to extract extensive information about individuals from even ephemeral looks at a face. Adult expertise to individuate faces does not reach mature levels until late adolescence. Our studies aim to track the development of the brain networks for face processing from school age children through adolescence and into adulthood. We use sophisticated functional magnetic resonance imaging (FMRI) to detect the function of various face sensitive brain regions and how these regions are organized into coherent functional systems. This work is supported by the Eunice Kennedy Shriver National Institute of Child Health & Human Development (5R01 HD060595).
It has long been known that people are better at recognizing faces of individuals from their own race compared to faces from other races. This is known as the "Other Race Effect" in face processing. What is less well understood is how brain processing may differ for own and other race faces. Our studies in this area use functional connectivity analyses to explore the brain networks used for own- and other-race face processing and attempt to uncover the functional properties that explain the Other Race Effect. The results of these studies have profound implications for everyday issues such as race relations and eyewitness testimony. Our studies are conducted with parallel experiments conducted in San Diego and China. This work is supported by the Eunice Kennedy Shriver National Institute of Child Health & Human Development (5R01 HD060595) and the UC San Diego Kavli Institute of Brain and Mind (Innovative Research Program).
Related to the cross race face studies described above, we are exploring the factors that influence human expertise in perceiving faces of different races and different facial emotional expressions (e.g., fear, happy, anger, surprise, disgust, sadness, and neutral) from the multimodal perspectives of computational modeling, human behavioral psychophysics, and functional brain imaging. This works involves the use of sophisticated 3D morphable models paired with model-free statistical analysis approaches to identify and classify the factors that distinguish the critical pattern differences within and between faces of different races and expressions of different emotions. The results of this approach defines the "face space" of races and emotional expressions. We submit these face space models to computational modeling and human psychophysics studies to estimate the critical "break points" that lead to labeling a face as one race or another, or expressing one emotional expression or another. These findings are transformed into an FMRI study allowing us to evaluate the brain architecture supporting these break point differences. Eventually, we expect this work to provide a comprehensive model of brain face processing for these important social cognitive questions of race and emotion processing. This work is supported by the UC San Diego Kavli Institute of Brain and Mind (Innovative Research Program).
Approximately 13% of all births in the U.S. are preterm, with 50% of children born before 37 gestational weeks at risk for neurological, cognitive, and/or behavioral disorders. Thus, preterm birth is a significant medical and public health issue. Recent work in the UCSD Center for Human Development has begun exploring the links between structural brain changes and academic performance in young school-age children. There has been little work exploring the possible longer-term consequences in older school-age children born premature. We are developing our research program to evaluate functional brain changes that may be associated with preterm birth using traditional FMRI methods and functional connectivity measures. Pilot testing has begun.
Research projects on visual face processing require high quality face images that allow for the presentation of individual faces under varying viewing conditions, such as different viewing angles, different lighting conditions, and different emotional expressions. In addition, many computational projects require 3D face images standardized for easy morphing between individuals or emotional expressions. While some high quality databases for 3D face images exist, they are typically limited in the number of images available. Our 3DFaceBase project will provide a large database of well over 100 3D images of faces across a representative sample of males and females, and different races and ethnicities. Most images have a complete complement of emotional expressions. Face stimuli come packaged in sets of 45 2D images of faces from different angles and lighting conditions useful for standard face processing studies. In additional, all scans have been aligned to a standard 3D mesh allowing for easy morphing. We will provide tools for morphing and rendering developed for Matlab and commercially available industry-standard programs (AutoDesk 3D Studio Max). A beta version of the site is available now at 3DFaceBase.org. This work is supported by an ARRA Supplement Grant from the Eunice Kennedy Shriver National Institute of Child Health & Human Development (5R01 HD060595-02S1).
FMRI data are inherently noisy. In addition to the main signal of interest, such as task related or resting state activity, the BOLD signal contains noise from physiological sources (pulsations, respiration), movement, and machine artifacts. FMRI data preprocessing is a methodical procedure aimed are reducing as much as possible these noise factors. Traditional methods often apply general linear model procedures to accomplish this, but often do not sufficiently account for physiological noise. Physiological factors can be measured at the time of scanning with external equipment, but the use of such equipment is often not feasible in developmental studies where children may not accept the additional testing burden required to obtain well calibrated physiological data. We have developed a semi-automated procedure based on Independent Component Analysis (ICA) that can remove noise due to physiological factors and motion. In addition, our method may provide the unique ability to equate task-negative or default mode activation across groups, a feature that is not possible to achieve using general linear models alone. Soon, we will offer our tools, developed primarily for the use with AFNI, FSL, and Matlab, free on this website. This work is supported by the Eunice Kennedy Shriver National Institute of Child Health & Human Development (5R01 HD060595).