QCell: Recombining and Aligning Cell Queries for Overlapping Instance Segmentation

Hugging Face Daily Papers Papers

Summary

QCell is a query-based model that improves instance segmentation of overlapping cells in microscopy through latent-space recombination and contrastive query alignment, achieving state-of-the-art results on benchmarks.

Instance segmentation of overlapping cells in microscopy remains challenging due to semi-transparent structures that produce weak boundaries and mixed visual evidence in overlap regions. Existing methods address this through local regions of interest or shape priors but lack global reasoning across overlapping objects. We present QCell, a novel query-based model that de-overlaps cell instances in microscopy scenes. Our approach combines (i) an instance recombination module that decomposes and recombines query representations in latent space, enabling the model to reason about complete object structure under overlap, and (ii) a contrastive query alignment objective that combines distinctive instance feature learning and separation of overlapping cell queries. We additionally introduce a new Organoid dataset benchmark for overlapping cell segmentation. We show that QCell outperforms state-of-the-art methods across multiple benchmarks, achieving +2.2 AP and +2.7 AJI on ISBI2014. Code is available at https://github.com/SlavkoPrytula/QCell
Original Article
View Cached Full Text

Cached at: 09/04/26, 11:57 AM

Paper page - QCell: Recombining and Aligning Cell Queries for Overlapping Instance Segmentation

Source: https://huggingface.co/papers/2608.29253

Abstract

QCell is a query-based model that improves instance segmentation of overlapping microscopy cells through latent-space recombination and contrastive query alignment.

Instance segmentation of overlapping cells inmicroscopyremains challenging due to semi-transparent structures that produce weak boundaries and mixed visual evidence in overlap regions. Existing methods address this through local regions of interest or shape priors but lack global reasoning across overlapping objects. We present QCell, a novelquery-based modelthat de-overlaps cell instances inmicroscopyscenes. Our approach combines (i) aninstance recombinationmodule that decomposes and recombines query representations in latent space, enabling the model to reason about complete object structure under overlap, and (ii) acontrastive query alignmentobjective that combines distinctive instance feature learning and separation of overlapping cell queries. We additionally introduce a new Organoid dataset benchmark foroverlapping cell segmentation. We show that QCell outperforms state-of-the-art methods across multiple benchmarks, achieving +2.2 AP and +2.7 AJI on ISBI2014. Code is available at https://github.com/SlavkoPrytula/QCell

View arXiv pageView PDFProject pageGitHubAdd to collection

Get this paper in your agent:

hf papers read 2608\.29253

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper0

No model linking this paper

Cite arxiv.org/abs/2608.29253 in a model README.md to link it from this page.

Datasets citing this paper1

#### YaroslavPrytula/Revvity-25 Updatedabout 5 hours ago • 168 • 5

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2608.29253 in a Space README.md to link it from this page.

Collections including this paper2

Similar Articles

CohortHijack: Robustness of Single Cell Annotation to Companion Cell Removal

arXiv cs.LG

This arXiv paper introduces CohortHijack, a robustness audit that removes non-target cells from single-cell query cohorts to test how annotation tools can be manipulated without altering the target cell's expression profile. It shows that structured removal and search strategies can change refined labels in popular pipelines while preserving the target, identifying query cohort composition as a vulnerability surface.

CellWorld: From Gene-Level Reconstruction to Latent Cell Prediction in Spatial Transcriptomics Foundation Models

arXiv cs.AI

CellWorld introduces a latent-space predictive pretraining approach for spatial transcriptomics foundation models, predicting latent representations of masked cells instead of reconstructing gene measurements. Across held-out datasets, even small variants outperform existing baselines on all benchmarks, showing that scaling and broad biological diversity improve transferability.