Top left (Medicine): 3D model of infants learned from RGB-D sequences. Right (Society): Voters make subconscious judgements about leadership qualities based on body shape. Bottom left (Psychology): the BMI of personalized avatars is varied to study how subjects perceive their body shape.
Our bodies and our health are intertwined. The question we ask is how we can leverage our models of the human body to detect and treat disease? To answer this, we collaborate with doctors and psychologists to relate body shape and movement to health.
Specifically, we explore how body shape is perceived by people with anorexia nervosa (AN). It was thought that AN patients might perceive their bodies in a distorted way. Using virtual reality and body scans of AN patients we explored this by varying the shape of personal and other avatars to test their perception. We found that anorexics perceive body shape veridically but prefer unhealthy weights.
Through this work, we have developed a range of body shape modeling, animation, and VR technologies that can be clinically deployed. For example, our virtual caliper allows people to create a realistic 3D human shape using only the controllers of a VR game system.
We also studied how we see body shape and describe it with language. We had subjects rate 3D bodies along many dimensions and then built a statistical model relating words and shape. With this "Body Talk" system, we could recover 3D shape from the descriptions of people. This opened up research on body shape and subjective judgements.
Our body shape is also closely related to our risk of various diseases such as diabetes and cardiovascular disease. Body fat carried around the internal organs is particularly dangerous and the question is whether this can be determined non-invasively from visible body shape? To answer this, we have scanned people with MRI and our optical body scanners. By fitting our models to the MRI data, we relate external shape to the internal body properties.
Our movement is also an important clue to health. For example, cerebral palsy (CP) can be diagnosed in infants based on their movement but this requires a trained expert. Early treatment of CP can improve outcomes and automating early diagnosis would make it more widely available. Consequently, we have extended our adult body models and tracking methods to infants in a clinical setting.
Our bodies are also political. Humans hold subconcious biases about others based soley on body shape. Such biases could influence decision making in politics, hiring, and justice. We have used our 3D body models to relate perceived leadership attributes to 3D body shape and have used this to analyze subconcious bias.
