From Human Attention to Diagnosis: Semantic Patch-Level Integration of Vision-Language Models in Medical Imaging | Read Paper on Bytez