YOLOv7-based automated detection platform for scalp lesions

Area of Science:

• Dermatology and Computational Imaging
• Artificial Intelligence in Healthcare

Background:

• Scalp health, including conditions like dandruff and erythema, is a growing area of dermatological research.
• Automated detection of scalp lesions is crucial for timely diagnosis and treatment.

Purpose of the Study:

• To develop an automated, web-based system for detecting scalp lesions such as dandruff and erythema using the YOLOv7 model.
• To evaluate the performance of YOLOv7 against established deep learning models for scalp lesion identification.

Main Methods:

• A dataset of 2200 clinical scalp images was utilized.
• Images were preprocessed using Roboflow, and the YOLOv7 model was trained and evaluated.
• Performance was compared against YOLOv5, YOLOF, and a single-shot detector, with integration into a Flask API web application.

Main Results:

• YOLOv7 achieved a mean average precision of 98.6%, with precision and recall rates of 98.6% and 97.2%, respectively.
• The YOLOv7 model outperformed baseline models in both training and testing phases on unseen data.

Conclusions:

• The study validates YOLOv7's effectiveness for diagnosing scalp lesions.
• This research highlights the successful integration of advanced AI models into practical dermatological healthcare solutions.