An Automatic AI-Based Algorithm That Grades the Scalp Surface Exfoliating Process From Video Imaging. Application to Dandruff Severity and Its Validation on Subjects of Different Ages and Ethnicities.

Summary

A handheld multi-illumination imaging device coupled to an AI algorithm was trained on 3,600 images (234 subjects) and validated on 460 images (192 subjects) to automatically grade dandruff severity. Device outputs showed significant correlation with expert ratings and low mean average error, supporting objective, rapid scalp assessment across ages and ethnicities.

Key Findings

• Handheld device with three illumination modes and AI achieved significant correlation with dermatologist 6-point atlas ratings.
• Validation on 192 subjects (460 images) across ages and ethnicities demonstrated generalizability.
• Mean Average Error metrics indicated low discrepancy between AI and expert assessments.

Clinical Implications

Dermatologists can use AI-assisted grading to monitor treatment response and support diagnosis of scalp disorders; industry can employ it for standardized efficacy endpoints in trials.

Limitations

• Abstract lacks exact performance statistics (e.g., r value, MAE magnitude)
• Clinical utility beyond dandruff (other scalp disorders) requires dedicated validation

Future Directions

Report full performance metrics, assess longitudinal responsiveness to therapy, and expand to other scalp and skin conditions with device calibration across phototypes.