As a consequence of the introduction of mathematical human body models (HBMs) in
consumer information programs, there is an increased need for reliable methods
that can demonstrate and build trust in the capability of HBMs to predict human
response and injury risk in crashes.
Therefore, a framework for validation of strain-based injury prediction is
proposed. The framework comprises stepwise validation with the final step to
validate the utility of risk predictions by means of the area under the curve
(AUC) combined with Brier scores.
SAFER HBM V11.1.0 previously validated at component and body part levels was
selected for the demonstration of the final step of the framework to validate
the capability to predict fracture risk in frontal, oblique, and lateral
loading. For frontal loading, five postmortem human surrogate (PMHS) test series
with 43 PMHS (age range: 19–88 years) were reconstructed. The predicted rib
fracture risk for 2+ and 3+ fractured ribs was compared to the number of
fractured ribs sustained by the PMHS. Using the framework, the SAFER HBM was for
frontal impact analysis found capable of predicting 2+ fractured ribs with an
AUC of 0.90 and 3+ fractured ribs with an AUC of 0.89. For oblique and lateral
impacts, three PMHS test series including 47 PMHS (23 with 0, 22 with 2+, and 20
with 3+ fractures) were reconstructed, and SAFER HBM rib fracture risk
predictions obtained AUC values of 0.84 and 0.87 for 2+ and 3+ fractured ribs,
respectively. In the frontal load case, the Brier score was 0.14 for the Number
of Fractured Ribs (NFR) 2+ model and 0.17 for the NFR 3+ model. In the
lateral/oblique load case, Brier scores were 0.20 and 0.18 for the NFR 2+ and
NFR 3+ models, respectively.
The proposed framework is a suitable method to objectively assess the utility of
HBM injury predictions, demonstrated with the SAFER HBMs capability to predict
rib fracture risk.
