Automated quantification of the skin aging process using in-vivo confocal microscopy

Abstract : This research work deals with the application of Reflectance Confocal Microscopy to the study of in-vivo skin structures, with a view to understand and characterize markers associated with aging. However, skin aging has been studied using this image modality but so far only by experienced dermatologists. The objective of this thesis is to develop new methods to quantify automatically the phenomenon of skin aging using in vivo reflectance confocal microscopy. In our work we first address the quantification of the epidermal state. Then, we characterize the Dermal-Epidermal Junction (DEJ). Finally, we validate the proposed methods through both clinical and cosmetic product efficacy studies. The epidermal layer appears on RCM images as a honeycomb pattern. Its regularity decreases with age. We propose an algorithm composed of two steps: 1) the image is segmented into individual cells, 2) each cell is classified as regular or irregular by machine learning based on spatial features. Then, we propose two measures to quantify the regularity of the honeycomb pattern on each image stack: 1) the percentage of regular cells, 2) the average size of the regular regions. The aggregated scores defined by the classification results show significant difference among groups of different ages and photo-exposition sites. The DEJ is a complex, surface-like, 3D structure separating the epidermis from the dermis. We provide a method for segmenting the 3D confocal images into three regions with reduced uncertainty: Epidermis, DEJ and Dermis. The proposed approach relies on a 3D Conditional Random Field to model the skin biological properties and impose regularization constraints. Classical methods for analyzing the DEJ shape rely on the characterization of its peaks and valleys. The inclusion relation between the level-lines of the DEJ forms a tree that can be analysed to extract relevant attributes. The choice of attributes is driven by the dermatologists expertise, they have identified the loss of circularity in the dermal papillae with aging. We show that, with aging, the DEJ is composed of more irregular objects, which is consistent with the dermatologists expertise. We carry out a clinical validation study involving 160 subjects from 4 different ethnic background. Clinical annotations are performed by experienced dermatologists. Using our proposed measurements, we are able to retrieve the statistical differences between the different annotated sets. When predicting the clinical annotations, we obtain, for the epidermis, the following scores: 80% accuracy, 81% sensibility, 81% specificity and for the DEJ 83% accuracy, 76% sensibility, 81% specificity. We finally perform a cosmetic product efficacy study. After only 2 weeks of product application, early signs of the product’s anti-aging action are detected thanks to the proposed methods.
Complete list of metadatas

Cited literature [132 references]  Display  Hide  Download
Contributor : Julie Robic <>
Submitted on : Friday, October 12, 2018 - 9:40:30 AM
Last modification on : Tuesday, March 19, 2019 - 11:43:24 PM
Long-term archiving on : Sunday, January 13, 2019 - 12:37:01 PM


Files produced by the author(s)


  • HAL Id : tel-01884978, version 1



Julie Robic. Automated quantification of the skin aging process using in-vivo confocal microscopy. Computer Science [cs]. Université Paris Est, 2018. English. ⟨tel-01884978⟩



Record views


Files downloads