Automated Skin Cancer Detection and Diagnosis Using Neural Networks and Multi-Agent Systems

Abstract

This paper proposes a multi-agent system based on large-scale language models (LLMs) for automated skin cancer diagnosis from images. The architecture includes segmentation, classification, and report generation steps, integrating specialized agents based on models such as Gemini-2.5-pro and GPT-4o. The approach was evaluated using bootstrap sampling with five replicates of size 80, highlighting the efficiency of the visual agents and the consistency of the diagnoses generated. Preliminary results indicate the potential of integrating computer vision and natural language to support medical diagnosis.