Evaluating the agreement and reliability of a web-based facial analysis tool for rhinoplasty

• Published in: International journal for computer assisted radiology and surgery Vol. 16; no. 8; pp. 1381 - 1391
• Main Authors: Topsakal, Oguzhan, Akbaş, Mustafa İlhan, Smith, Bria Synae, Perez, Michael Francis, Guden, Ege Can, Celikoyar, Mehmet Mazhar
• Format: Journal Article
• Language: English
• Published: Cham Springer International Publishing 01.08.2021; Springer Nature B.V
• Subjects: Agreements; Applications programs; Computer Imaging; Computer Science; Confidence limits; Correlation coefficients; Cost analysis; Evaluation; Health Informatics; Human subjects; Imaging; Medicine; Medicine & Public Health; Modelling; Normality; Original Article; Pattern Recognition and Graphics; Plastic surgery; Radiology; Reliability analysis; Software; Software reliability; Statistical analysis; Statistical methods; Surgery; Three dimensional models; Two dimensional models; Vision; Agreement; Intraclass correlation coefficient; Facial analysis; Surgery planning; 3D model; Rhinoplasty; Facial measurements; Reliability; Bland–Altman
• ISSN: 1861-6410; 1861-6429; 1861-6429
• DOI: 10.1007/s11548-021-02423-z

Purpose Rhinoplasty is one of the most common and challenging plastic surgery procedures. Facial analysis is a crucial step in planning. Utilizing three-dimensional (3D) model of a patient’s face is an emerging way of performing facial analysis. This paper evaluates the agreement and reliability of facial measurements taken using a web app, located at digitized-rhinoplasty.com, that utilizes 3D models of the patient’s face. Methods Eleven measurements were calculated on 16 human subjects. Three methods of measurements were performed: direct measurements on human subjects’ faces, measurements on 2D photographs, and measurements on 3D models of face scans. The Bland–Altman plot is used for testing the agreement between the web app and the well-known Blender 3D modeling software. Intra-rater and inter-rater reliability was calculated and compared for 2D and 3D methods using the intraclass correlation coefficient (ICC) method. The statistical analysis methods were checked for the normality and homoscedasticity assumptions. Results The results indicate that the web app and Blender software show agreement within 95% confidence limits. The web app performs well in intra-rater and inter-rater reliability statistical analysis. The web app’s reliability scores are consistently better than facial analysis software which was found highly reliable in a previous study. We also compare the methods of measurements in terms of time, ease of use, and cost. Conclusion The utilization of 3D computer modeling for facial analysis has its advantages and started to become more common due to recent advances in technology. The web app utilizes 3D face scans for pre-operative planning and post-operative evaluation of facial surgeries. The web app performs well in agreement and inter-/intra-reliability analysis and performs consistently better than software that works utilizing 2D photographs. The web app provides accurate, repeatable, affordable, and fast facial measurements for facial analysis when compared to direct and 2D methods.