Rethinking BCC diagnosis: Automating concept-specific detection of BCC in dermatoscopic images

Area of Science:

• Dermatology and Artificial Intelligence
• Medical Image Analysis
• Oncology

Background:

• Basal cell carcinoma (BCC) diagnosis is challenging due to dermoscopy subjectivity.
• Current AI systems offer limited interpretability for clinical decision-making.

Purpose of the Study:

• To develop an interpretable AI system for enhanced BCC diagnosis.
• To improve diagnostic accuracy and clinical utility of AI in dermatology.

Main Methods:

• A refined BCC dataset from HAM10000 was created with clinician-annotated features.
• ResNet50 and Mask R-CNN architectures were integrated for enhanced performance.
• Statistical evaluations validated the clinical diagnosis evaluation scheme.

Main Results:

• The RFSD-BCC system demonstrated significant improvements in sensitivity, specificity, and accuracy.
• Achieved an area under the precision-recall curve of 0.84, closely matching physician diagnoses.
• Showcased increases in sensitivity (7%), specificity (11%), accuracy (10%), and ICC (6%).

Conclusions:

• The RFSD-BCC system enhances BCC diagnosis through feature combination models.
• Provides interpretable diagnoses, bridging AI with clinical practice.
• Significantly improves clinician accuracy and patient understanding of BCC.