This research introduces CLISC, a novel approach combining CLIP and SAM foundation models to accurately segment brain tumors without human annotations.

• Leverages vision-language models to identify tumors in medical images autonomously
• Achieves superior performance by enhancing Class Activation Maps (CAM) techniques
• Eliminates costly and time-consuming manual annotation requirements
• Demonstrates potential for wider application across medical imaging diagnostics

Why It Matters: This breakthrough reduces diagnostic bottlenecks in healthcare by enabling automatic tumor identification without expert annotations, potentially accelerating treatment decisions and improving patient outcomes.

CLISC: Bridging clip and sam by enhanced cam for unsupervised brain tumor segmentation