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

Fluorescence microscopy allows for a detailed inspection of cells, cellularnetworks, and anatomical landmarks by staining with a variety ofcarefully-selected markers visualized as color channels. Quantitativecharacterization of structures in acquired images often relies on automaticimage analysis methods. Despite the success of deep learning methods in othervision applications, their potential for fluorescence image analysis remainsunderexploited. One reason lies in the considerable workload required to trainaccurate models, which are normally specific for a given combination ofmarkers, and therefore applicable to a very restricted number of experimentalsettings. We herein propose Marker Sampling and Excite, a neural networkapproach with a modality sampling strategy and a novel attention module thattogether enable (i) flexible training with heterogeneous datasets withcombinations of markers and (ii) successful utility of learned models onarbitrary subsets of markers prospectively. We show that our single neuralnetwork solution performs comparably to an upper bound scenario where anensemble of many networks is na\"ively trained for each possible markercombination separately. In addition, we demonstrate the feasibility of thisframework in high-throughput biological analysis by revising a recentquantitative characterization of bone marrow vasculature in 3D confocalmicroscopy datasets and further confirm the validity of our approach on anadditional, significantly different dataset of microvessels in fetal livertissues. Not only can our work substantially ameliorate the use of deeplearning in fluorescence microscopy analysis, but it can also be utilized inother fields with incomplete data acquisitions and missing modalities.

Modality Attention and Sampling Enables Deep Learning with Heterogeneous Marker Combinations in Fluorescence Microscopy