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 Tools annual

Global: Zendy Free Plan

Global: Zendy Tools monthly

Global: Zendy Plus monthly

Recently, spatiotemporal graphs have emerged as a concise and elegant mannerof representing video clips in an object-centric fashion, and have shown to beuseful for downstream tasks such as action recognition. In this work, weinvestigate the use of latent spatiotemporal graphs to represent a surgicalvideo in terms of the constituent anatomical structures and tools and theirevolving properties over time. To build the graphs, we first predict frame-wisegraphs using a pre-trained model, then add temporal edges between nodes basedon spatial coherence and visual and semantic similarity. Unlike previousapproaches, we incorporate long-term temporal edges in our graphs to bettermodel the evolution of the surgical scene and increase robustness to temporaryocclusions. We also introduce a novel graph-editing module that incorporatesprior knowledge and temporal coherence to correct errors in the graph, enablingimproved downstream task performance. Using our graph representations, weevaluate two downstream tasks, critical view of safety prediction and surgicalphase recognition, obtaining strong results that demonstrate the quality andflexibility of the learned representations. Code is available atgithub.com/CAMMA-public/SurgLatentGraph.

Encoding Surgical Videos as Latent Spatiotemporal Graphs for Object and Anatomy-Driven Reasoning