Regional Attention-Enhanced Swin Transformer for Clinically Relevant Medical Image Captioning