What type of study is this?

Optimizing prompt strategies for SAM: advancing lesion segmentation across diverse medical imaging modalities

View Full Paper

Ask AI

Bookmark

Authors

Retrospective study evaluates SAM prompt strategies across various tumor datasets, indicating improvements in segmentation accuracy.

Increasing the number of prompt points significantly improves segmentation accuracy, with Dice coefficients rising from 0.272 to 0.806.
Performance varies by prompt location, with surface prompts achieving higher Dice coefficients than center-based prompts in ovarian and breast tumors.
A reinforcement learning agent achieved a peak Dice coefficient of 0.595 for ovarian tumors, outperforming alternative strategies.
The RL agent dramatically reduced segmentation time, achieving nearly a 10-fold improvement compared to manual methods.

Wang et al. (2025) studied this question.

View Full Paper

Ask AI

Bookmark