Developing “Skintelix”: An Intelligent Virtual Assistant for Supporting Pharmacists’ DecisionMaking in the Preliminary Assessment and Management of Common Dermatological Conditions

Abstract

This project introduces "Skintelix," a novel, multimodal artificial intelligence (AI) assistant designed to augment pharmacists' decision-making in the preliminary assessment and management of common skin conditions. Leveraging advanced generative AI models, specifically GPT-4V, the system integrates textual symptom analysis with clinical image processing to provide accurate differential diagnoses and evidence-based treatment recommendations. The development of Skintelix followed a three-tiered architectural model, comprising a user-centric conversational interface, a robust AI core engine, and a scalable pharmacy integration layer. The model was trained on a multi-source dataset, combining established global dermatology image repositories (such as HAM10000, DDI, and SCIN) with real-world clinical data from community pharmacies and authoritative medical references. This hybrid approach was deliberately employed to mitigate common algorithmic biases and enhance clinical relevance. The system's ability to process both images and text, its adoption of a conversational format, and its focus on locally relevant clinical contexts represent significant contributions to the field. Skintelix offers a comprehensive theoretical framework to enhance diagnostic accuracy, minimise unnecessary specialist referrals, and enable pharmacists to serve as a more effective first-line defence in dermatological care. While a full-scale clinical validation is a crucial next step, the preliminary design demonstrates high feasibility and considerable promise for sustainable deployment in challenging environments.