YOLOv11 for Psoriasis Classification: Enhancing Feature Extraction for Accurate Severity Differentiation

Psoriasis is a chronic autoimmune skin condition that varies in redness severity from Mild to Very Severe. Accurate classification is essential for early diagnosis, effective treatment, and disease management. However, distinguishing between Absent and Mild cases remains challenging due to their subtle visual similarities, even for experienced dermatologists. Therefore, several Classification techniques for dermatoscopic images have been proposed, but they have been found inadequate or a binary classification This paper introduces an optimized deep-learning model based on the YOLOv11 framework for multi-class psoriasis Readiness severity classification. The standard YOLOv11 model initially demonstrated limitations in classification accuracy. To enhance its performance, the architecture was refined with an advanced feature extraction layer, improving feature differentiation and overall classification effectiveness. The model was trained and validated on a clinically approved dataset comprising 8,792 images across five Redness severity categories. The dataset captures a diverse range of skin conditions, ensuring robust generalization to real-world applications. Evaluation metrics, including accuracy (85.3%), precision (97%), and recall (99%), demonstrate that the modified YOLOv11 model significantly outperforms the original. This lightweight yet efficient approach offers a reliable AI-powered tool for clinical decision-making, large-scale screenings, and deployment in resource-limited settings, ultimately improving diagnostic accuracy and efficiency in psoriasis severity classification.