Design customization of respiratory mask based on 3D face anthropometric data

Abstract

This study presents a machine learning based method for design customization of a 3D respiratory mask. A parametric model of a 3D human face was constructed from an anthropometric database consisting of 495 facial models. An AdaBoost.R algorithm was applied to identify a set of measurable parameters most related to the facial geometry. The correlation between parameters was estimated using principal component analysis and linear regression. With those parameter values as input, the parametric model generates 3D meshes of a human face that serve as a design reference for the construction of a customized respiratory mask of a good fit. We conducted a series of experiments with 10-fold cross-validation to validate the effectiveness of the proposed method.