Hierarchical process using Brier Score Metrics for lower leg injury risk curves in vertical impact

Lower Extremity Injury Risk Curve Development for a Human Body Model in the Underbody Blast Environment

Computational human body models (HBMs) provide the ability to explore numerous candidate injury metrics ranging from local strain based criteria to global combined criteria such as the Tibia Index. Despite these efforts, there have been relatively few studies that focus on determining predicted injury risk from HBMs based on observed postmortem human subjects (PMHS) injury data. Additionally, HBMs provide an opportunity to construct risk curves using measures that are difficult or impossible to obtain experimentally. The Global Human Body Models Consortium (GHBMC) M50-O v 6.0 lower extremity was simulated in 181 different loading conditions based on previous PMHS tests in the underbody blast (UBB) environment and 43 different biomechanical metrics were output. The Brier Metric Score were used to determine the most appropriate metric for injury risk curve development. Using survival analysis, three different injury risk curves (IRC) were developed: "any injury," "calcaneus injury," and "tibia injury." For each injury risk curve, the top three metrics selected using the Brier Metric Score were tested for significant covariates including boot use and posture. The best performing metric for the "any injury," "calcaneus injury" and "tibia injury" cases were calcaneus strain, calcaneus force, and lower tibia force, respectively. For the six different injury risk curves where covariates were considered, the presence of the boot was found to be a significant covariate reducing injury risk in five out of six cases. Posture was significant for only one curve. The injury risk curves developed from this study can serve as a baseline for model injury prediction, personal protective equipment (PPE) evaluation, and can aid in larger scale testing and experimental protocols.

Default risk prediction and feature extraction using a penalized deep neural network

Online peer-to-peer lending platforms provide loans directly from lenders to borrowers without passing through traditional financial institutions. For lenders on these platforms to avoid loss, it is crucial that they accurately assess default risk so that they can make appropriate decisions. In this study, we develop a penalized deep learning model to predict default risk based on survival data. As opposed to simply predicting whether default will occur, we focus on predicting the probability of default over time. Moreover, by adding an additional one-to-one layer in the neural network, we achieve feature selection and estimation simultaneously by incorporating an L 1 -penalty into the objective function. The minibatch gradient descent algorithm makes it possible to handle massive data. An analysis of a real-world loan data and simulations demonstrate the model's competitive practical performance, which suggests favorable potential applications in peer-to-peer lending platforms.

An Improved Method for Developing Injury Risk Curves Using the Brier Metric Score

Many injury metrics are routinely proposed from measured or derived quantities from biomechanical experiments using post mortem human subjects (PMHS). The existing literature did not provide guidance on deciding between parameters collected in an experiment that would be best to use for the development of human injury probability curves (HIPC). The objective of this study was to use the Brier Metric Score (BMS) to identify the most appropriate metric from an experiment that predicts injury outcomes. The Brier Metric Score assesses how well a metric predicts the outcome for a censored data point (a lower BMS is better). Survival analysis was then conducted with the selected metric and the best distribution was selected using Akaike information criterion (AIC). Confidence intervals (CIs) and the normalized confidence interval width (NCIS) were calculated for the injury probability curve. The testing and validation of the methods described were performed using biomechanics data in the open literature. The methods for the HIPC development procedure detailed herein have been rigorously tested and used in the generation of WIAMan HIPCs and Injury Assessment Reference Curves (IARCs) for the WIAMan ATD, but can also be used in other ATD or PMHS injury risk curve development.