English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Deep learning models trained with large amounts of data have become a recentand effective approach to predictive problem solving -- these have become knownas "foundation models" as they can be used as fundamental tools for otherapplications. While the paramount examples of image classification (earlier)and large language models (more recently) led the way, the Segment AnythingModel (SAM) was recently proposed and stands as the first foundation model forimage segmentation, trained on over 10 million images and with recourse to over1 billion masks. However, the question remains -- what are the limits of thisfoundation? Given that magnetic resonance imaging (MRI) stands as an importantmethod of diagnosis, we sought to understand whether SAM could be used for afew tasks of zero-shot segmentation using MRI data. Particularly, we wanted toknow if selecting masks from the pool of SAM predictions could lead to goodsegmentations.  Here, we provide a critical assessment of the performance of SAM on magneticresonance imaging data. We show that, while acceptable in a very limited set ofcases, the overall trend implies that these models are insufficient for MRIsegmentation across the whole volume, but can provide good segmentations in afew, specific slices. More importantly, we note that while foundation modelstrained on natural images are set to become key aspects of predictivemodelling, they may prove ineffective when used on other imaging modalities.

Testing the Segment Anything Model on radiology data