Implementation of artificial intelligence algorithms for melanoma screening in a primary care setting

Area of Science:

• Dermatology
• Oncology
• Medical Informatics

Background:

• Skin cancer is the most common cancer in Caucasians, posing a significant healthcare burden.
• Early skin tumor detection is crucial for reducing healthcare costs and improving patient outcomes.
• Artificial intelligence (AI) shows promise in assisting dermatologists with skin cancer diagnosis.

Purpose of the Study:

• To develop an efficient computer-aided diagnosis (CAD) system for primary care physicians (PCPs) to enable early skin cancer detection.
• To create a smartphone application for acquiring patient data and classifying skin lesions using AI.
• To provide PCPs with a tool for risk stratification and timely referral of suspected skin cancers.

Main Methods:

• Development of a CAD system integrated with a smartphone application for data acquisition (images, demographics, clinical history).
• Implementation of AI algorithms for classifying clinical and dermoscopic images of skin lesions.
• Generation of reports including lesion images, heat maps, probability of malignancy, diagnosis, and management suggestions.

Main Results:

• The dermoscopy AI model achieved 89.3% accuracy for melanoma detection.
• The clinical image model demonstrated 84.7% accuracy for skin cancer diagnosis.
• Both AI models exhibited high sensitivity, specificity, and an area under the curve (AUC) above 0.9.

Conclusions:

• The developed CAD system effectively screens skin cancers and guides lesion management by PCPs.
• This AI tool is particularly beneficial in areas with limited access to dermatologists.
• The system facilitates risk stratification and improves timely access to specialist care for high-risk patients.