ABSTRACT The smartphone has become dermatology's primary clinical camera, yet the photo it saves is assembled by software rather than being a faithful unedited photograph: several frames are merged, grain is removed, and texture is altered before the image file exists, so no unedited original survives. This processing is hidden, inconsistent between shots, recorded only in undocumented vendor‐specific form, and cannot be fully disabled in the camera app. Smartphone photographs still serve to locate a lesion and support a referral, but the processing pipeline is ill‐suited and potentially misleading for serial monitoring, research, and training of artificial intelligence models. We propose a per‐image record of capture and processing, practical measures clinicians can take now, and smartphone camera processing pipelines built for clinical fidelity rather than aesthetic ends.
Harrison A. Edwards, Philip L. Tong. Clinical Photography's Computational Blind Spot. // Australasian Journal of Dermatology. – 2026. – DOI: 10.1111/ajd.70225. – URL: https://onlinelibrary.wiley.com/doi/10.1111/ajd.70225?af=R
Harrison A. Edwards & Philip L. Tong (2026). Clinical Photography's Computational Blind Spot. Australasian Journal of Dermatology. https://doi.org/10.1111/ajd.70225
Harrison A. Edwards and Philip L. Tong. “Clinical Photography's Computational Blind Spot.” Australasian Journal of Dermatology, 2026. doi:10.1111/ajd.70225.
@article{Edwards2026,
title = {Clinical Photography's Computational Blind Spot},
author = {Edwards, Harrison A. and Tong, Philip L.},
journal = {Australasian Journal of Dermatology},
year = {2026},
doi = {10.1111/ajd.70225},
url = {https://onlinelibrary.wiley.com/doi/10.1111/ajd.70225?af=R}
}