Predicting respirator size and fit from 2D images

Individuals rely heavily on the efficacy of respirators when there is potential for chemical, biological, radiological, or nuclear threats (CBRN). The US Department of Defense (DoD) requires personnel to undergo time-consuming fit tests for full face respirators, which are a critical component of the personal protective equipment (PPE) ensemble. The quality of respirator fit directly contributes to its effectiveness. Leveraging the ubiquity and capabilities of mobile devices, machine learning, and advances in computerised 3D modelling, we seek to make this process simpler, faster, and more accurate. This paper introduces the mask analysis and size quantification (MASQ) framework: an extensible mobile-based semi-automated system that: 1) combines 2D images of a subject's head, an existing 3D headform model generation approach, and an analytic model to recommend a size; 2) establishes a platform for future research of the shape and anthropometric features of the human face and respirator sizing and fit effectiveness.