DARE3d-v2

DARE 3D v2 This repository contains the 3D datasets and v2 models and associated files related to the paper: Romain Karpinski†, Alice Gros◊, Marc Karnat*, Qazi Saaheelur Rahaman*◊, Jules Vanaret◊, Mehdi Saadaoui◊, Sham Tlili◊, Jean-François Rupprecht** Aix Marseille Univ, Université de Toulon, CNRS, CPT (UMR 7332), Turing Centre for Living Systems, Marseille, France† LORIA, CNRS, Nancy, France◊ Aix Marseille Univ, CNRS, IBDM (UMR 7288), Turing Centre for Living Systems, Marseille, France Contact: [e-mail], [e-mail], [e-mail] Description We propose a two-stage deep learning method to characterize cell divisions in 3D time-lapse microscopy sequences. 1. Stage 1 – Division Detection: A 3D semantic segmentation network (3D U-Net) identifies candidate division events in volumetric image sequences using temporal context from consecutive frames. 2. Stage 2 – Regression of Division Geometry: A 3D convolutional regression model estimates the orientation and separation of daughter cells for each detected event. The method was applied to challenging 3D live-imaging datasets, including avian neural tube explants labeled with **SiR-actin** (membrane signal) and mouse gastruloids expressing **H2B–GFP** (nuclear signal). Ground-truth annotations consist of the two daughter-cell centers represented as 3D spheres centered on the estimated daughter positions. The workflow is intended for datasets in which full segmentation and cell tracking are difficult or unreliable. Repository contents - **DARE3dv2_Zenodo_040926.zip:** 3D+time image sequences and annotations used for training, validation, and testing. Citation If you use this dataset or associated files, please cite: Romain Karpinski, Alice Gros, Marc Karnat, Qazi Saaheelur Rahaman, Jules Vanaret, Mehdi Saadaoui, Sham Tlili, and Jean-François Rupprecht, “DARE: Division Axis and Region Estimation from time-lapse image sequences in 2D and 3D.” License MIT licence