Abstract
Background Histological alterations such as nuclear abnormalities are sensitive biomarkers associated with diseases, tissue injury and environmental insults. While visual inspection and human interpretation of histology images are useful for initial characterization, such low-throughput procedures suffer from inherent limitations in terms of reliability, objectivity and reproducibility. Artificial intelligence and digital morphometry offer unprecedented opportunities to quickly and accurately assess nuclear morphotypes in relation to tissue damage including skin injury. Methods In this work, we designed NoxiScore, a pipeline providing an integrated, deep learning-based software solution for fully automated and quantitative analysis of nucleus-related features in histological sections of human skin biopsies. We used this pipeline to evaluate the efficacy and safety of three dermato-cosmetic products massively sold at the time of the study in the Montpellier area (South of France): a sunscreen containing UV filters, a mosquito repellent (with synthetic active ingredient IR3535) and a product combining a natural insect repellent plus a sunscreen. Hematoxylin and eosin or hematoxylin-eosin saffron staining was performed to assess skin structure before morphometric parameter computation. Results We report the identification of a specific nuclear feature based on variation in texture information that can be used to assess skin tissue damage after oxidative stress or UV exposure. Our data show that application of the commercial sun cream provided efficient protection against UV effects in our ex vivo skin model, whereas application of the mosquito repellent as a single product exerted no protective or toxic effect. Notably, we found that concurrent application of the insect repellent with the sunscreen significantly decreased the UVB protective effect of the sunscreen. Last, histometric analysis of human skin biopsies from multiple donors indicates that the sunscreen-insect repellent combo displayed variable levels of protection against UV irradiation. Conclusions To our knowledge, our study is the first to evaluate the potential toxicity of combining real-life sunscreen and insect repellent products using ex vivo human skin samples, which most closely imitate the cutaneous physiology. The NoxiScore wet-plus-dry methodology has the potential to provide information about the pharmaco-toxicological profile of topically applied formulations and may also be useful for diagnostic purposes and evaluation of the skin exposome including pesticide exposure, air pollution and water contaminants. Graphical Abstract