[Image-based computer diagnosis of melanoma]

Area of Science:

• Dermatology and Medical Imaging
• Artificial Intelligence in Healthcare

Background:

• Automated diagnostic systems are increasingly used for melanoma diagnosis.
• Traditional systems involve preprocessing, segmentation, feature extraction, and classification.
• Transfer learning algorithms are emerging as a more efficient approach.

Purpose of the Study:

• To review the established steps in automated melanoma diagnosis.
• To highlight the advantages of transfer learning in computer-assisted melanoma diagnosis.
• To discuss the role and limitations of smartphone applications in melanoma screening.

Main Methods:

• Review of traditional automated diagnostic system steps.
• Analysis of transfer learning algorithms for melanoma diagnosis.
• Evaluation of current smartphone applications for melanoma screening.

Main Results:

• Transfer learning can potentially make segmentation and feature extraction steps obsolete.
• Transfer learning algorithms show improved results in melanoma diagnosis.
• Smartphone applications for melanoma screening lack rigorous clinical validation.

Conclusions:

• Transfer learning offers a promising advancement in automated melanoma diagnosis.
• Caution is advised regarding the use of smartphone applications by laypersons due to unverified diagnostic accuracy.