An Empirical Evaluation of Low-Rank Adapted Vision–Language Models for Radiology Image Captioning | Synapse