BACKGROUND: The protective effectiveness of an N95 respirator depends on the filtration efficiency of the material from which the N95 respirator is made of, as well as the wearers' facial fit. The facial fit of an N95 respirator mainly depends on the degree of matching between the wearers' facial dimension characteristics and the N95 respirator. Quantitative fit testing objectively evaluates the fit of N95 respirators; however, it is not easy to promote because of the limitations of testing conditions. The aim of this study is to establish a fit prediction model of N95 respirator based on facial images. METHODS: Facial images and fit factor (FF) value of 5 N95 respirators were gathered from 299 medical staffs in 10 hospitals in Beijing. Face geometry measurement was based on 3D face modelling, and the American TSI-8038 Porta Count Pro+ was used to conduct quantitative fit test. Multiple linear regression analysis was employed to identify facial dimensional features that significantly influenced the fit of N95 respirators. Through matching training of facial image and FF values, a fit prediction model has been established, enabling rapid recommendation of N95 respirators meeting the fit standard via facial image recognition. RESULTS: A fit prediction model for N95 respirators based on facial images has been developed, which enables the rapid recommendation of N95 respirators with acceptable FF value for healthcare personnel. The model demonstrated an accuracy of 55.93%, a precision of 98.43%, a recall of 51.65%, and an F1 score of 0.68. CONCLUSIONS It is feasible to utilize computer-based facial recognition technology to rapidly recommend N95 respirators for medical personnel. Given the high level of accuracy achieved, the model demonstrates significant potential for practical application.
Humans ; Face/anatomy & histology* ; N95 Respirators/standards* ; Male ; Adult ; Female ; Middle Aged ; Beijing