Photographing Alopecia: How Many Pixels Are Needed for Clinical Evaluation?

Area of Science:

• Dermatology
• Computer Vision
• Medical Imaging

Background:

• Determining optimal image resolution is vital for computational efficiency, standardization, and storage in clinical settings.
• Understanding human perception of image details is key to defining diagnostic requirements.

Purpose of the Study:

• To explore the minimum image resolution required for clinical assessment of alopecia.
• To analyze how clinicians detect alopecia characteristics at varying image resolutions.

Main Methods:

• Mimicked computer vision experiments by varying image resolution based on patient head width.
• Recruited 6 clinicians to identify regions of interest in de-identified images at different resolutions.
• Compared clinician accuracy with a neural network's performance in classifying hair density.

Main Results:

• Clinicians detected alopecia presence at very low image resolutions.
• Significantly higher resolution was necessary to identify vellus-like hair.
• Alopecia detection accuracy correlated with resolution, mirroring neural network performance.

Conclusions:

• Low image resolution is sufficient for detecting alopecia presence.
• Higher resolution is critical for identifying subtle hair characteristics like vellus-like hair.
• Human expert performance may set an upper bound for image resolution in alopecia diagnostic algorithms.