Visual-Language Model Fine-Tuning via LoRA for Structed Medical Reports Generating for Lung X-Ray Skans

This research introduces a novel approach for automating the generation of structured clinical reports from chest radiographs by fine-tuning a pre-trained vision-language model (VLM). We leverage the Low-Rank Adaptation (LoRA) method to efficiently adapt the VLM to the specific requirements of radiological report generation. This adaptation allows for the precise control of the output format, ensuring the consistent inclusion of key structured data fields: image tags, patient sex, patient age, and a comprehensive image description that incorporates relevant diagnostic findings. The methodology encompasses a detailed explanation of the VLM architecture, a description of the curated chest X-ray dataset employed for fine-tuning, and a thorough exposition of the training procedure. Crucially, we implement 4-bit quantization during the LoRA fine-tuning process to minimize computational resource requirements and facilitate broader accessibility. We evaluate the performance of the fine-tuned model using established natural language generation metrics. Quantitative results demonstrate substantial improvements in the quality and coherence of the generated reports compared to baseline models. These findings highlight the potential of our approach to significantly enhance the efficiency and consistency of clinical report generation, reducing the manual burden on radiologists and potentially improving the standardization of reporting practices. The presented methodology offers a promising pathway towards practical, automated clinical report generation from medical imaging data.