My research primarily focuses on eye-tracking enhanced medical image analysis, as well as exploring the integration of large language models (LLMs) into computer-aided diagnosis (CAD). During my Master's program, I have dedicated my efforts to aligning CAD models and human perception. This endeavor encompasses two critical aspects:
Vision-Level Alignment: Utilizing eye-tracking data, I aim to deepen the understanding of healthcare professionals' visual attention and replicate it within our models. This approach seeks to align the visual processing capabilities of CAD systems with those of human experts.
Language-Level Alignment: I employ the reasoning capabilities of LLMs to transform complex, difficult-to-interpret diagnostic results from CAD models into comprehensive and explainable reports. This not only aids in making diagnoses more accessible but also enhances the communicative efficiency between medical professionals and AI systems.