Jingwen Hu is the associate director of the University of Michigan Transportation Research Institute (UMTRI) and a research associate professor in UMTRI and the Department of Mechanical Engineering. His research primarily focuses on using a combination of experimental, computational, and epidemiological procedures to study impact/injury biomechanics and injury prevention.
Dr. Hu has recently developed parametric computational human models representing diverse populations. Such models have been used to study the injury mechanism and safety design optimizations for motor vehicle crashes in various vulnerable populations, such as children, the elderly, obese occupants, pedestrians, pregnant people, and wheelchair users.
We sat down with Dr. Hu to speak about his research and how his parametric computational human model can help advance concussion science.
Michigan Concussion Center: Can you talk about your career path and research?
Dr. Jingwen Hu: My undergraduate degree is in automotive engineering and I have always been fascinated by vehicle crash tests. Albert King, a professor of biomedical engineering at Wayne State University, inspired me to go into impact biomechanics by his quote, “vehicles are designed for humans, not dummies.” I was thinking if you don’t know the human being and how they get injured, you cannot design a better car. That is why I pursued my PhD in Biomedical Engineering focusing on impact biomechanics.
My main research area uses computational human body models to do injury assessment and design safety countermeasures for injury prevention. Human models are universal tools that can apply to injury prevention or safety design not only in motor vehicle crashes but also in sports and many other injurious events. Additionally, I am also a big sports fan and concussions are something I am passionate about. I believe that the parametric human modeling concept we used for vehicle crashes can directly apply to concussion assessment and prevention.
MCC: What is the parametric finite element human model?
JH: A finite element human model separates the body into thousands or even millions of small elements, where we assign different material properties to represent different body tissues. The model can predict the tissue biomechanical responses of the human body, including the brain, under various impact conditions. The assumption is that the model’s tissue responses, such as strain and stress, will correlate better to injury than head kinematic measures from video footage or sensors since they do not measure tissue-level deformations.
At the same time, there are large variations in skull and brain geometries in the population. The geometry variation will highly affect the tissue-level responses in the brain. That is why we cannot use one human body model to represent everyone. The parametric human model can account for the morphological and biomechanical variations among the population, which should give a much better prediction at the tissue level of a future concussion.
MCC: How does this compare to other models?
JH: One of the main difficulties in injury assessment and prevention is human diversity because the current safety designs are primarily based on one dummy or a single computational human model representing a mid-sized male. Finite element human models are time-consuming to build, which could take upwards of a year to build a single model. Therefore, it is almost infeasible to develop many different human models to represent a diverse population using the traditional approach.
Our parametric modeling approach uses automated mesh morphing tools to change the geometry of one human model to another person so we can represent a diverse population. We use medical image and body scan data and quantify 3-D geometry variations of a skeleton, internal organs, and external body surface among the population. This way, we can tell predict the human geometry based on age, sex, stature, weight, or other physical measures of a person and quickly morph a baseline human model into the predicted geometry. In the future, we could then develop many models for different athletes, and potentially predict the injury risk in their brains more accurately.
I believe that our approach is a paradigm change. Instead of using a single model, we can generate hundreds or thousands of human models for population-based simulations. Now that we have this capability, we could run the model millions of times under different impact conditions and use some machine-learning models to predict the strain/stress distribution in the brain of an athlete in a specific impact condition. Those will be instantly useful results and a future tool to help us put the subject-specific injury assessments into safety design. We could even evaluate specific impact distributions for certain types of athletes and begin designing specific safety features to best protect them.
MCC: How do you gather the subject’s geometric information?
JH: We have a partnership with the U-M Medical School and the Department of Radiology, where we receive thousands of CT scans from them. When we get the images, we use 3-D reconstruction software to extract the geometry of the skin and the skull’s outer and inner surfaces. We then use the statistical analysis to quantify the variations in those 3-D geometries and predict the geometry based on a given set of subject covariates.
MCC: How do you hope your model will move concussion science forward?
JH: I think this will be a good tool for both injury assessment and injury prevention. Current diagnosis procedures for concussion are largely based on a patient’s self-reporting, which is somewhat subjective and does not consider the biomechanical responses inside of the brain. We think this model could potentially be used as an objective tool to quantify the severity of a concussive impact.
Let’s say in the future we have some on-field sensors to quantify head impact accelerations. We could take that information along with the subject co-variates to quickly predict the strain and stress distribution in the brain. If we see some high strain areas that are over the threshold, we can say this impact is on the high alert. Even though the player may not behave out of the ordinary, we can be more cautious and further evaluate them. So this is a tool that can provide objective evaluation beyond those based on our current technology and procedure.
The other thing I’m thinking is that the model can help design safety equipment. Let’s take the helmet as an example. Currently, college ice hockey players and football players only have a few sizes that they can pick, and they may not necessarily fit an athlete well. We could fix that by creating a subject-specific design feature to fill in the gap and make sure that helmets fit better and are more comfortable. Furthermore, personalized design features can be tailored to better prevent brain injuries based on model prediction.
We can also use this model to run simulations to test hypotheses. For example, there are differences between male and female concussion rates, but we don’t know why. One potential reason could be neck strength. We could use the computational model to have different neck muscle models and then run the impact simulation to see the concussion risk. The data from the computational simulations could help fill in where we don’t have data.
MCC: What are the next-level barriers of concussion research in your area of expertise?
JH: One of the biggest hurdles is going to be figuring out how we can validate this model. We want to make sure we have a biofidelic model to make sense of its results, which requires field data. We can measure some of the head kinetics with sensors attached to the helmet, skin, or mouthguard. We need to gather this kind of information to check whether model-predicted responses can be correlated with a concussion. We will need to collect large data sets to say if this model indeed improved concussion prediction and be used as a reliable way to provide critical injury risk assessment.
MCC: What excites you most about being a member of the U-M Concussion Center?
JH: I’m an engineer by training, so my vision is focused on the engineering part of this multi-disciplinary problem. However, getting to know so many medical doctors and understanding their perspectives on concussions makes my vision and contribution to the theme of concussion prevention much broader. This is extremely valuable for me.