Everything everywhere... with just one prompt

Following the path set in one of our previous papers, we are pleased to announce the publication of the article “Prompt Once, Segment Everything: Leveraging SAM 2 Potential for Infinite Medical Image Segmentation with a Single Prompt”. The starting question behind these two papers is similar: given the enormous resources needed to develop Artificial Intelligence models specialized in a task, is it feasible to use foundation models for the same tasks without performing costly optimizations on them? To answer this, we focus on a task of vital importance with applications that can save lives: medical image segmentation.

In this article we present a novel approach that leverages SAM 2’s video segmentation capabilities to reduce the number of prompts needed to segment a complete volume of medical images. The study first compares the performance of SAM and SAM 2 in medical image segmentation. Then, our novel approach is presented. The results show that SAM 2 achieves an average improvement of 1.76% in Jaccard Index and 1.49% in Dice Score compared to SAM.

Our next goal is to provide a methodology that allows defining the limit of what we can achieve when using these foundation models.

Surely we’re missing something: after all, we can’t be everywhere at once. 😉 Would you like to share some of your ideas with us? Come see us and let’s talk.

Prompt Once, Segment Everything is one of the outcomes of musicgenia, a project funded by Grant CPP2021-008491 from MICIU/AEI/10.13039/50100011033 and by the European Union through NextGenerationEU/PRTR.