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

Image-text contrastive learning has proven effective for pretraining medicalimage models. When targeting localized downstream tasks like semanticsegmentation or object detection, additional local contrastive losses thatalign image regions with sentences have shown promising results. We study howlocal contrastive losses are related to global (per-sample) contrastive lossesand which effects they have on localized medical downstream tasks. Based on atheoretical comparison, we propose to remove some components of local lossesand replace others by a novel distribution prior which enforces uniformity ofrepresentations within each sample. We empirically study this approach on chestX-ray tasks and find it to be very effective, outperforming methods withoutlocal losses on 12 of 18 tasks.

The Role of Local Alignment and Uniformity in Image-Text Contrastive Learning on Medical Images