Multi-Height Extraction of Clinical Parameters Improves Classification of Craniosynostosis

Abstract Introduction: 3D surface scan-based diagnosis of craniosynostosis is a promising radiation-free alternative to traditional using computed tomography. The cranial index (CI) and the vault asymmetry (CVAI) are well-established clinical parameters that widely used. However, they also have benefit being easily adaptable for automatic without need extensive preprocessing. Methods: We propose multi-height-based classification approach uses CI CVAI in different height layers compare it initial only one layer. use ten-fold cross-validation test seven classifiers. dataset 504 patients consists three types control group consisting healthy non-synostotic subjects. Results: improved all k-nearest neighbors classifier scored best with mean accuracy 89% F1-score 0.75. Conclusion: Taking into account beneficial classification. Based on accepted used parameters, this might be step towards an easy-to-understand transparent both physicians patients.