Body mass index estimation using photogrammetry and whole-body videography

Summary: Abstract Body: Current proof-of-concept methods for using total body photography to assess body mass index (BMI) utilize expensive commercial whole-body scanners that are inaccessible to patients. This study aims to validate a novel approach using smartphone videos and publicly available software libraries for photogrammetry. Our approach leverages Apple’s RealityKit (RK) software package and publicly available python libraries to parse a video recorded from an iPhone 16 Pro (with 4K resolution at 60 frames per second). Videos were taken of two participants in a spiral fashion beginning at the head and ending at the feet. Their heights and weights were recorded at that time. The videos averaged 600.4 megabytes in size and 89 seconds in length (with an average of 5340 frames). 1000 frames were extracted at equal intervals, per video, from which three-dimensional (3D) models were reconstructed using RK on a MacBook Pro (Model: A2485). Model volumes were calculated using Riemann summation of two-dimension axial slices. The average percent error of the slice-based BMI was 2.37% (SD=0.57%) with a mean difference of -0.6 units with 95% limit of agreement (LoA) of ±0.12 using Bland-Altman analysis, compared to commercial systems which had a mean difference of −0.1 units with 95% LoA of ±2.1 units. Literature shows that increases in BMI and height are associated with increased Breslow thickness and melanoma risk, respectively. Our approach provides a low-cost alternative to commercial whole-body scans, increasing accessibility to novel technology without compromising precision. The minimum achievable cost of our proposed system is 0.01% of the cost of previously studied commercial systems ($429 vs $435,000), and our entire process can be completed on any iPhone that supports iOS 17.0 onwards. Our approach can also be applied to determine body surface area involvement of inflammatory skin diseases, helping to more objectively monitor disease progression or improvements on therapy, in clinical and research settings. The same calculations will need to be performed on a larger, representative populations. Christopher Guirguis¹, Joe Tung² 1. Georgetown University School of Medicine, Washington, DC, United States. 2. UPMC, Pittsburgh, PA, United States. Bioinformatics, Computational Biology, and Imaging