High-resolution imaging system for integration into intelligent noncontact total body scanner.

Area of Science:

• Medical Imaging
• Dermatology
• Computer Vision

Background:

• Rising melanoma incidence necessitates advanced early detection methods.
• Current total body scanners require high-quality input data for accurate computer-aided diagnosis.

Purpose of the Study:

• To develop a high-resolution optical skin imaging module and software for processing raw data into dermoscopic images.
• To enhance diagnostic performance of total body scanners using focus stacking for hyperfocus images.

Main Methods:

• Employed focus stacking with an electrically tunable liquid lens for in vivo image acquisition.
• Developed algorithms for image alignment, focus measurement, fusion, and deep learning-based super-resolution.
• Trained an artificial intelligence (AI) classification model for lesion analysis.

Main Results:

• Achieved noncontact dermoscopic images with complete focus across varied skin topographies.
• Generated hyperfocus images with $28\mu m$ resolution at 50 frames per second.
• Observed some resolution variance compared to contact dermoscopy standards.

Conclusions:

• Focus stacking improves noncontact dermoscopy image quality for early melanoma detection.
• Advanced imaging and AI techniques integrated into total body scanners enable clinical early detection.
• Careful consideration of super-resolution techniques is needed to maintain data authenticity.