Automatic Focus Assessment on Dermoscopic Images Acquired with Smartphones.

Area of Science:

• Dermatology
• Medical Imaging
• Machine Learning

Background:

• Mobile devices, particularly smartphones, are increasingly used for acquiring dermoscopic skin lesion images.
• The growing demand for self-care and telemedicine necessitates robust methods for evaluating image quality in skin lesion monitoring.

Purpose of the Study:

• To develop an automated system for assessing the focus and quality of dermoscopic images captured by mobile devices.
• To guide users during image acquisition, ensuring reliable data for skin lesion monitoring.

Main Methods:

• A feature-based machine learning approach was employed to develop the automated assessment system.
• The system incorporates preview image validation, including artifact detection and focus validation, followed by acquired image quality assessment.
• Two datasets of dermoscopic skin lesions and artifacts, collected via mobile devices, were used for system development and testing.

Main Results:

• The best model for automatic preview assessment achieved an overall accuracy of 77.9%.
• Focus assessment of the acquired images reached a global accuracy of 86.2%.
• Implementation in an Android application demonstrated promising results and real-world viability.

Conclusions:

• The developed system effectively assesses the quality of mobile-acquired dermoscopic images.
• This automated solution supports telemedicine and self-care by ensuring reliable image data for skin lesion monitoring.