Assessing the contribution of orthodontic profiles in predicting facial soft tissue thickness for forensic facial approximation.

BACKGROUND

Facial soft tissue thickness (FSTT) is essential for forensic facial approximation. Although its correlations with age, sex, and body mass index (BMI) are well documented, the potential correlations between FSTT and various orthodontic profiles-such as cephalic index (CI), skeletal class (SC), Tweed and Northwestern analyses-remain unexplored collectively. This study examined these correlations and their impact on FSTT prediction accuracy.

METHODS

We analyzed 103 postmortem computed tomography datasets from Japanese cadavers aged 18-86 years. Moderate-to-high multicollinearity was identified among orthodontic profile variables (SC, Tweed, and Northwestern) and addressed using principal component analysis (PCA), yielding two principal components (PC1 and PC2). Predictive formulas were constructed incorporating age, sex, BMI, CI, PC1, and PC2. To evaluate model performance, we conducted two comparative approaches: (1) comparing root mean squared error (RMSE) and mean absolute error (MAE) from the PCA-based regression model with those derived from holdout dataset's BMI-based mean estimates, and (2) with primary dataset's baseline regression model including only age, sex, and BMI, across all landmarks.

RESULTS AND DISCUSSION

PCA reduced multicollinearity, retaining 77% of total data variability. Based on the two comparative approaches, the PCA-based regression model demonstrated marginal improvements in predictive accuracy, as indicated by slightly lower RMSE and MAE across most landmarks. It indicates a limited yet consistent benefit of using orthodontic profiles for enhancing model accuracy beyond basic demographic predictors.

CONCLUSION

The inclusion of orthodontic profiles demonstrated modest improvements in predictive accuracy and may enhance the interpretive value of FSTT predictive models in forensic contexts.

CLINICAL TRIAL NUMBER

Not applicable.