Differences in the annotation between facial images and videos for training an artificial intelligence for skin type determination
These videos have been matched automatically. Contact us if they are not relevant.
These videos have been matched automatically. Contact us if they are not relevant.
View abstract on PubMed
Summary
This summary is machine-generated.Videos may offer a more sensitive assessment of dynamic skin features like wrinkles compared to static images. Subjectivity in skin analysis highlights the need for consistent rater training and cross-validation in AI development.
Area Of Science
- Dermatology
- Artificial Intelligence
- Medical Imaging
Background
- The Grand-AID project aims to develop an AI-powered digital tool for personalized skin analysis and care routines.
- Training this AI necessitates accurate annotation of skin features from facial imagery.
- A key research question is whether video analysis is superior to static images for assessing dynamic skin parameters.
Purpose Of The Study
- To compare the effectiveness of video versus static image analysis for annotating eight distinct skin features.
- To evaluate inter-rater reliability in assessing these features across different modalities.
Main Methods
- 25 healthy volunteers provided standardized image sequences and video recordings with facial expressions.
- Four dermatologically expert raters annotated eight skin features using semi-quantitative and linear scales.
- A cross-over design was employed to compare image and video modalities and assess rater differences.
Main Results
- Most assessed skin parameters showed higher scores in video analysis compared to static images, with some differences being statistically significant.
- Significant variability was observed among the expert raters' assessments.
Conclusions
- Significant differences exist between image and video analysis for skin feature assessment.
- Subjectivity inherent in skin analysis underscores the importance of rigorous rater training and cross-validation for AI model development.
These videos have been matched automatically. Contact us if they are not relevant.
These videos have been matched automatically. Contact us if they are not relevant.