The paper presents a novel approach, ChatCAD, that integrates large language models (LLMs) with computer-aided diagnosis (CAD) networks to enhance medical image analysis. The framework leverages LLMs' medical knowledge and reasoning to improve the quality of CAD outputs, such as diagnosis, lesion segmentation, and report generation. By converting CAD network outputs into natural language, LLMs can summarize and refine these outputs, leading to more accurate and comprehensive reports. The study demonstrates that ChatCAD significantly improves diagnostic performance, with ChatGPT achieving a 16.42% increase in F1-score compared to state-of-the-art models. Additionally, the approach enhances patient communication by providing interactive explanations and advice, making it a promising tool for clinical decision-making and online healthcare services. The paper also discusses the limitations and future directions, including the need for better datasets, data privacy considerations, and further research on LLMs' logical reasoning capabilities.The paper presents a novel approach, ChatCAD, that integrates large language models (LLMs) with computer-aided diagnosis (CAD) networks to enhance medical image analysis. The framework leverages LLMs' medical knowledge and reasoning to improve the quality of CAD outputs, such as diagnosis, lesion segmentation, and report generation. By converting CAD network outputs into natural language, LLMs can summarize and refine these outputs, leading to more accurate and comprehensive reports. The study demonstrates that ChatCAD significantly improves diagnostic performance, with ChatGPT achieving a 16.42% increase in F1-score compared to state-of-the-art models. Additionally, the approach enhances patient communication by providing interactive explanations and advice, making it a promising tool for clinical decision-making and online healthcare services. The paper also discusses the limitations and future directions, including the need for better datasets, data privacy considerations, and further research on LLMs' logical reasoning capabilities.