Line-field confocal optical coherence tomography coupled to Artificial Intelligence to identify quantitative biomarkers of facial skin ageing: An exploratory study
Podium 13
Presented by: FRANCK BONNIER
Introduction: Skin ageing results from the accumulation of molecular damage over time, potentially exacerbated by a combination of intrinsic (genetics, cellular metabolism, hormonal and metabolic processes) and extrinsic factors (chronic light exposure, pollution, ionizing radiation, chemicals, toxins). Worldwide, anti-ageing cosmetics have become daily products improving individual’s life through wellbeing and social interaction. While the cosmetics industry has developed a range of products to slow down the degenerative process of the skin, product diversification and personalisation for increased effectiveness requires the visualisation, characterisation, and quantification of histological changes related to ageing. Line-field Confocal Optical Coherence Tomography (LC-OCT) is an emerging in vivo non-invasive 3D imaging technique with high spatial resolution (~1 µm) and depth of analysis (~ 500 μm) providing multiscale information to study age-related modifications in skin morphology and assess the biological age in various populations. (Ogien et al. Biomed. Opt. Express 2020).
Artificial Intelligence (AI) using segmentation algorithms is a powerful method to derive 3D quantitative parameters such as skin layer thicknesses (SC and epidermis) and keratinocyte network information (nuclei size, shape and density). The thickness and density of the superficial dermis can also be assessed (Pedrazzani et al. Skin Res Technol. 2020).
The objective of this study was to identify quantitative biomarkers for a non-invasive assessment of the skin’s ageing state in order to investigate the correlations between biological age, perceived age and clinical parameters from a large cohort of volunteers.
Methods: The study was supervised by the French ethical committee “Comité de protection des personnes sud méditerranée I”. 100 Caucasian female volunteers evenly distributed in 5 age groups 20–30, 31-40, 41-50, 51-60 and 61-70 years old were included. For each volunteer, three 3D LC-OCT stacks of 1.2 mm x 0.5 mm x 0.5 mm (x/y/z) were collected from three facial areas (temple, cheekbone and mandible).
Quantitative measurements of the epidermis at the tissue and cellular levels were obtained by 3D segmentation of the skin surface, the end of the stratum corneum, the dermal-epidermal junction and the keratinocytes using AI algorithms. Dermal metrics such as the thickness of the superficial dermis and the optical attenuation coefficient of the dermis were obtained by analysing the intensity profiles with depth. Biological metrics determined by LC-OCT were then subjected to a quantitative multivariate analysis method, namely Partial Least Squares Regression (PLSR) (Matlab®, Mathworks), to construct predictive models of facial age.
Results and discussions: LC-OCT allowed the thickness of epidermal skin layers to be determined non-invasively at the micrometric level. In the 3 studied areas, the thickness of the stratum corneum tended to increase with age. The living epidermis showed the opposite evolution with a significant decrease in thickness for the mandible (-7 µm) between the age groups 20-30 and 60-70. At the cellular level, the surface density of keratinocytes decreased significantly with age on the temple and mandible. The average volume of the 20% largest nuclei, which are the most advanced nuclei in their maturation process, as well as the standard deviation of the volume of nuclei increases with age on the 3 face sites highlighted disparities in the size of keratinocyte nuclei in elderly subjects. Besides, the sphericity of keratinocytes nuclei decreases with age. Deeper in the skin, the decrease in the thickness of the superficial dermis and the increase in the optical attenuation coefficient of the dermis are also indicators of the ageing process.
Epidermal and dermal metrics are biological variables that can be processed by chemometric methods to correlate with the ageing process. PLSR resulted in a linear regression with R2~0.9.
Conclusions: 3D LC-OCT coupled to AI is a powerful non-invasive tool for in vivo morphological assessment of the skin from the surface to the superficial reticular dermis. Presently, metrics extracted from 3D images are quantitative biomarkers that can be used to construct predictive models to determine biological age based on structural and histological features changing with ageing. While the technique has great potential for exploratory studies to fully characterise the ageing mechanism, including ethnic comparisons, it also opens up new opportunities to test, to visualise and to quantify the efficacy of topical skin cosmetic products.
Artificial Intelligence (AI) using segmentation algorithms is a powerful method to derive 3D quantitative parameters such as skin layer thicknesses (SC and epidermis) and keratinocyte network information (nuclei size, shape and density). The thickness and density of the superficial dermis can also be assessed (Pedrazzani et al. Skin Res Technol. 2020).
The objective of this study was to identify quantitative biomarkers for a non-invasive assessment of the skin’s ageing state in order to investigate the correlations between biological age, perceived age and clinical parameters from a large cohort of volunteers.
Methods: The study was supervised by the French ethical committee “Comité de protection des personnes sud méditerranée I”. 100 Caucasian female volunteers evenly distributed in 5 age groups 20–30, 31-40, 41-50, 51-60 and 61-70 years old were included. For each volunteer, three 3D LC-OCT stacks of 1.2 mm x 0.5 mm x 0.5 mm (x/y/z) were collected from three facial areas (temple, cheekbone and mandible).
Quantitative measurements of the epidermis at the tissue and cellular levels were obtained by 3D segmentation of the skin surface, the end of the stratum corneum, the dermal-epidermal junction and the keratinocytes using AI algorithms. Dermal metrics such as the thickness of the superficial dermis and the optical attenuation coefficient of the dermis were obtained by analysing the intensity profiles with depth. Biological metrics determined by LC-OCT were then subjected to a quantitative multivariate analysis method, namely Partial Least Squares Regression (PLSR) (Matlab®, Mathworks), to construct predictive models of facial age.
Results and discussions: LC-OCT allowed the thickness of epidermal skin layers to be determined non-invasively at the micrometric level. In the 3 studied areas, the thickness of the stratum corneum tended to increase with age. The living epidermis showed the opposite evolution with a significant decrease in thickness for the mandible (-7 µm) between the age groups 20-30 and 60-70. At the cellular level, the surface density of keratinocytes decreased significantly with age on the temple and mandible. The average volume of the 20% largest nuclei, which are the most advanced nuclei in their maturation process, as well as the standard deviation of the volume of nuclei increases with age on the 3 face sites highlighted disparities in the size of keratinocyte nuclei in elderly subjects. Besides, the sphericity of keratinocytes nuclei decreases with age. Deeper in the skin, the decrease in the thickness of the superficial dermis and the increase in the optical attenuation coefficient of the dermis are also indicators of the ageing process.
Epidermal and dermal metrics are biological variables that can be processed by chemometric methods to correlate with the ageing process. PLSR resulted in a linear regression with R2~0.9.
Conclusions: 3D LC-OCT coupled to AI is a powerful non-invasive tool for in vivo morphological assessment of the skin from the surface to the superficial reticular dermis. Presently, metrics extracted from 3D images are quantitative biomarkers that can be used to construct predictive models to determine biological age based on structural and histological features changing with ageing. While the technique has great potential for exploratory studies to fully characterise the ageing mechanism, including ethnic comparisons, it also opens up new opportunities to test, to visualise and to quantify the efficacy of topical skin cosmetic products.