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 Tools annual

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 monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

Few-shot learning is an established topic in natural images for years, butfew work is attended to histology images, which is of high clinical value sincewell-labeled datasets and rare abnormal samples are expensive to collect. Here,we facilitate the study of few-shot learning in histology images by setting upthree cross-domain tasks that simulate real clinics problems. To enablelabel-efficient learning and better generalizability, we propose to incorporatecontrastive learning (CL) with latent augmentation (LA) to build a few-shotsystem. CL learns useful representations without manual labels, while LAtransfers semantic variations of the base dataset in an unsupervised way. Thesetwo components fully exploit unlabeled training data and can scale gracefullyto other label-hungry problems. In experiments, we find i) models learned by CLgeneralize better than supervised learning for histology images in unseenclasses, and ii) LA brings consistent gains over baselines. Prior studies ofself-supervised learning mainly focus on ImageNet-like images, which onlypresent a dominant object in their centers. Recent attention has been paid toimages with multi-objects and multi-textures. Histology images are a naturalchoice for such a study. We show the superiority of CL over supervised learningin terms of generalization for such data and provide our empiricalunderstanding for this observation. The findings in this work could contributeto understanding how the model generalizes in the context of bothrepresentation learning and histological image analysis. Code is available.

Towards better understanding and better generalization of few-shot classification in histology images with contrastive learning