Evaluating the Baseline Capability of Large Language Models (LLMs) in Interpreting Imaging for Neurosurgical Applications