Unification of frequentist inference and machine learning for pterygomaxillary morphometrics

Abstract

Background:The base of the skull, particularly the pterygomaxillary region, has a sophisticated topography, the morphometry of which interests pathologists, maxillofacial and plastic surgeons. The aim of the study was to conduct pterygomaxillary morphometrics and test relevant hypotheses on sexual and laterality-based dimorphism, and causality relationships. Materials and methods:We handled 60 dry skulls of adult Asian males (36.7%) and females (63.3%). We calculated the prime distance D [prime] for the imaginary line from the maxillary tuberosity to the midpoint of the pterygoid process between the upper and the lower part of the pterygomaxillary fissure, as well as the parasagittal D [x-y inclin.] and coronal inclination of D [x-z inclin.] of the same line. We also took other morphometrics concerning the reference point, the maxillary tuberosity.Results: Significant sexual as well as laterality-based dimorphism and bivariate correlations existed. The univariate models could not detect any significant effect of the predictors. On the contrary, summative multivariate tests in congruence with neural networks, detected a significant effect of laterality on D [x-y inclin.] (p-value = 0.066, partial eta squared = 0.030), and the interaction of laterality and sex on D [x-z inclin.] (pvalue = 0.050, partial eta squared = 0.034). K-means clustering generated three clusters highlighting the significant classifier effect of D [prime] and its three-dimensional inclination.Conclusions: Although the predictors in our analytics had weak-to-moderate effect size underlining the existence of unknown explanatory factors, it provided novel results on the spatial inclination of the pterygoid process, and reconciled machine learning with non-Bayesian models, the application of which belongs to the realm of oral-maxillofacial surgery.

Surface Web Merits for SARS-CoV-2 Pandemic in Iraq

“…Besides, the multivariate tests have inherent limitations of their own as per the renowned British statistician George Box's aphorism, "All models are wrong, but some are useful" [28]. Hence, optimizing statistical models and implementing them for realtime temporal analyses is valuable for future research [29][30][31]. There are also implicit constraints of the statistical packages, including IBM SPSS and Microsoft Excel, when loading a specific type or a count of variables into a data model, including the multivariate analysis of variance, supervised neural networks, and cluster analysis.…”

The Anatomy, Features and Sex Correlations (Dimorphism) of Tubero–Palato–Pterygoid Region among Adult Population—Single Center Study Based on 3D Printed Models

“…Oliveira et al, (2013) clearly state that gender and skeletal pattern, and craniofacial pattern, may influence the anatomy of the pterygopalatine region [17]. Imam et al (2020) state that all of the skull's dimorphic parameters (e.g., the distance between maxillary tuberosity and spinous foramen) have a big effect size and are in favor of males [18]. Thus, there is a great deal of research on the nature of morphological variation in structures located in the vicinity of the tubero-palato-pterygoid region.…”

Evaluation of Anatomy Course Teaching and Learning Outcomes for Iraqi Pharmacy Students: Internet-based Learning versus Blended Learning During the Pandemic

“…Our study has some limitations, including implementing a cross-sectional study design, the relatively small sample size, and the inherent limitations of statistical analysis [30], [31]. Cross-sectional surveys frequently suffer from nonresponse bias; nonetheless, we excluded missing data from our data analytics.…”