Authors: Zekun Jiang, Dongjie Cheng, Ziyuan Qin, Jun Gao, Qicheng Lao, Kang Li, Le Zhang

Published on: February 24, 2024

Impact Score: 8.0

Arxiv code: Arxiv:2402.15759

Summary

• What is new: Develops TV-SAM, a novel algorithm for zero-shot segmentation in medical images without manual annotations.
• Why this is important: Need for efficient zero-shot segmentation in medical imaging without manual input.
• What the research proposes: Introducing TV-SAM, which uses text and visual prompts generated by integrating GPT-4, GLIP, and SAM models.
• Results: TV-SAM effectively segments unseen targets across various modalities, outperforming previous models and closely matching gold standard performance.

Technical Details

Technological frameworks used: Text-Visual-Prompt SAM (TV-SAM)

Models used: GPT-4, Vision Language Model GLIP, Segment Anything Model (SAM)

Data used: Seven public datasets across eight imaging modalities

Potential Impact

Healthcare industry, specifically medical imaging companies and diagnostic software providers

Want to implement this idea in a business?

We have generated a startup concept here: MediSegmentAI.