Digital photographic grading tool for online visual skin evaluation on images
Podium 41
Presented by: Juliette Rengot
INTRODUCTION
To assess skincare product efficacy, it is critical to objectively follow skin evolution and quantify the product effect. Experts usually study cutaneous attributes by examining the skin in vivo. To make their evaluation reliable, they need visual grading tools to refer to and a strong training background. Powerful paper references[1] exist, especially for wrinkle evaluations.
However, there is still room for progress, in particular for digital image grading. Indeed, online image grading has several advantages. One same rater can manage multicentric studies, or several raters can remotely grade the same subjects. Views and cropped regions of interest (ROI) from different timepoints can be precisely compared. Besides, it allows a higher level of training and standardization, complementary to full-face in vivo grading. Even so, no online tool already reaches a consensus among experts.
We propose a complete online grading system for reliable skin evaluation from digital images. We describe the creation of a new set of digital scales for pores, eye bags, vascular and pigmented dark circles grading, in European female population. We chose these clinical signs because they are not yet perfectly described in existing references (not available for all ethnic groups, too small pictures…). We also present an innovative online image evaluation system[2] which enables several raters to easily grade or compare clinical images using live-display of the newly created digital image scales.
METHODOLOGY
We acquired photographic data in Sofia (Bulgaria), Lyon and Saint-Étienne (France). ColorFace®[3] was used to collect a large and representative set of standardized multi-modalities full-face photographs in high resolution.
Afterwards, the data were preprocessed with color registration, and targeted cropping of the ROIs (cheeks in front view and parallel polarization for pores, eyes in profile views and cross polarization for eye bags and eyes in front view and cross polarization for dark circles).
Evaluation criteria were defined: pore visibility, eye bag visibility, dark circle color intensity and dark circle surfaces. Images were sorted based on a quality control (“rejected”, “slightly visible criterion”, “highly visible criterion” classes…) to end with a set of 60 images per clinical sign, well distributed in terms of criterion evolution range.
Digital images were compared by pairs, by three experts, using the newly developed online tool. It ends in rankings, illustrating smooth evolutions. Using combined image analysis and visual assessment, a set of final images was selected to build grading scales.
RESULTS
Four innovating photographic scales were developed. They illustrate a linear and representative evolution of the different clinical signs. For dark circles, a bi-directional scale was created in order to represent both surface and color intensity evolutions, which was a completely new approach.
The photographic scales were implemented in the online tool and can now be used as visual reference for clinical image grading on the same exact screen. The experts select the reference that they feel is the closest to the image to assess, by examining at both images side by side. The algorithm assigns the corresponding score to the graded image. This process enables user friendly digital image grading in a standardized and objective way for the selected criteria.
CONCLUSIONS AND DISCUSSIONS
In addition to in vivo full-face evaluation, our tool opens a complementary standardized way to grade specific facial areas, in reference to standardized photographic scales of these exact same ROIs, side by side.
We consider enlarging this promising approach to other clinical signs. A specific module will also be developed for clinical rater training, to help them improve precision and repeatability.
The next steps are user experience testing and operational testing in the context of clinical studies for product performance assessment.
[1] Bazin, R. and Doublet, E., 2007. Skin aging atlas. Med'com.
[2] Voegeli, R., Schoop, R., Prestat-Marquis, E., Rawlings, A.V., Shackelford, T.K. and Fink, B., 2021. Cross-cultural perception of female facial appearance: A multi-ethnic and multi-centre study. Plos one.
[3] Campiche, R., Trevisan, S., Séroul, P., Rawlings, A.V., Adnet, C., Imfeld, D. and Voegeli, R., 2019. Appearance of aging signs in differently pigmented facial skin by a novel imaging system. Journal of cosmetic dermatology.
To assess skincare product efficacy, it is critical to objectively follow skin evolution and quantify the product effect. Experts usually study cutaneous attributes by examining the skin in vivo. To make their evaluation reliable, they need visual grading tools to refer to and a strong training background. Powerful paper references[1] exist, especially for wrinkle evaluations.
However, there is still room for progress, in particular for digital image grading. Indeed, online image grading has several advantages. One same rater can manage multicentric studies, or several raters can remotely grade the same subjects. Views and cropped regions of interest (ROI) from different timepoints can be precisely compared. Besides, it allows a higher level of training and standardization, complementary to full-face in vivo grading. Even so, no online tool already reaches a consensus among experts.
We propose a complete online grading system for reliable skin evaluation from digital images. We describe the creation of a new set of digital scales for pores, eye bags, vascular and pigmented dark circles grading, in European female population. We chose these clinical signs because they are not yet perfectly described in existing references (not available for all ethnic groups, too small pictures…). We also present an innovative online image evaluation system[2] which enables several raters to easily grade or compare clinical images using live-display of the newly created digital image scales.
METHODOLOGY
We acquired photographic data in Sofia (Bulgaria), Lyon and Saint-Étienne (France). ColorFace®[3] was used to collect a large and representative set of standardized multi-modalities full-face photographs in high resolution.
Afterwards, the data were preprocessed with color registration, and targeted cropping of the ROIs (cheeks in front view and parallel polarization for pores, eyes in profile views and cross polarization for eye bags and eyes in front view and cross polarization for dark circles).
Evaluation criteria were defined: pore visibility, eye bag visibility, dark circle color intensity and dark circle surfaces. Images were sorted based on a quality control (“rejected”, “slightly visible criterion”, “highly visible criterion” classes…) to end with a set of 60 images per clinical sign, well distributed in terms of criterion evolution range.
Digital images were compared by pairs, by three experts, using the newly developed online tool. It ends in rankings, illustrating smooth evolutions. Using combined image analysis and visual assessment, a set of final images was selected to build grading scales.
RESULTS
Four innovating photographic scales were developed. They illustrate a linear and representative evolution of the different clinical signs. For dark circles, a bi-directional scale was created in order to represent both surface and color intensity evolutions, which was a completely new approach.
The photographic scales were implemented in the online tool and can now be used as visual reference for clinical image grading on the same exact screen. The experts select the reference that they feel is the closest to the image to assess, by examining at both images side by side. The algorithm assigns the corresponding score to the graded image. This process enables user friendly digital image grading in a standardized and objective way for the selected criteria.
CONCLUSIONS AND DISCUSSIONS
In addition to in vivo full-face evaluation, our tool opens a complementary standardized way to grade specific facial areas, in reference to standardized photographic scales of these exact same ROIs, side by side.
We consider enlarging this promising approach to other clinical signs. A specific module will also be developed for clinical rater training, to help them improve precision and repeatability.
The next steps are user experience testing and operational testing in the context of clinical studies for product performance assessment.
[1] Bazin, R. and Doublet, E., 2007. Skin aging atlas. Med'com.
[2] Voegeli, R., Schoop, R., Prestat-Marquis, E., Rawlings, A.V., Shackelford, T.K. and Fink, B., 2021. Cross-cultural perception of female facial appearance: A multi-ethnic and multi-centre study. Plos one.
[3] Campiche, R., Trevisan, S., Séroul, P., Rawlings, A.V., Adnet, C., Imfeld, D. and Voegeli, R., 2019. Appearance of aging signs in differently pigmented facial skin by a novel imaging system. Journal of cosmetic dermatology.