Evaluating the Performance of Large Language Models in Generating Impressions for Radiology Reports | Synapse