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

Whole slide image (WSI) classification is a critical task in computationalpathology, requiring the processing of gigapixel-sized images, which ischallenging for current deep-learning methods. Current state of the art methodsare based on multi-instance learning schemes (MIL), which usually rely onpretrained features to represent the instances. Due to the lack oftask-specific annotated data, these features are either obtained fromwell-established backbones on natural images, or, more recently fromself-supervised models pretrained on histopathology. However, both approachesyield task-agnostic features, resulting in performance loss compared to theappropriate task-related supervision, if available. In this paper, we show thatwhen task-specific annotations are limited, we can inject such supervision intodownstream task training, to reduce the gap between fully task-tuned and taskagnostic features. We propose Prompt-MIL, an MIL framework that integratesprompts into WSI classification. Prompt-MIL adopts a prompt tuning mechanism,where only a small fraction of parameters calibrates the pretrained features toencode task-specific information, rather than the conventional full fine-tuningapproaches. Extensive experiments on three WSI datasets, TCGA-BRCA, TCGA-CRC,and BRIGHT, demonstrate the superiority of Prompt-MIL over conventional MILmethods, achieving a relative improvement of 1.49%-4.03% in accuracy and0.25%-8.97% in AUROC while using fewer than 0.3% additional parameters.Compared to conventional full fine-tuning approaches, we fine-tune less than1.3% of the parameters, yet achieve a relative improvement of 1.29%-13.61% inaccuracy and 3.22%-27.18% in AUROC and reduce GPU memory consumption by 38%-45%while training 21%-27% faster. Our code is available athttps://github.com/cvlab-stonybrook/PromptMIL.

Prompt-MIL: Boosting Multi-Instance Learning Schemes via Task-specific Prompt Tuning