Bhattiprolu, S. | Search bioRxiv

bioRxiv2026

ZEISS arivis Cloud: a cloud-based platform for deep learning model training and scalable bioimage analysis

Modern biological imaging generates large, complex datasets that require scalable and reproducible image analysis methods. Deep learning has demonstrated strong performance on bioimage segmentation tasks, but training custom models has remained inaccessible to many researchers due to requirements for GPU infrastructure, programming expertise, and large annotated training datasets. ZEISS arivis Cloud is a browser-based platform for deep learning model training that addresses these barriers through partial annotation support, AI-assisted labeling with SAM (Segment Anything Model), pretrained model initialization, and automatically configured training pipelines requiring no machine learning expertise. The platform supports two segmentation tasks: semantic segmentation using a U-Net-style architecture with an EfficientNet encoder and PixelShuffle decoder, and instance segmentation based on Mask2Former with a Swin-Tiny backbone. Both pipelines incorporate microscopy-specific adaptations including smooth tiling, multi-channel input support, dataset-specific normalization, and partial-annotation-aware loss functions protected by patents US-20240078681-A1 and US-20250111519-A1. Trained models integrate directly with ZEISS arivis Pro for pipeline-based image analysis, ZEISS arivis Hub for parallel execution across large datasets, and ZEISS ZEN for content-aware guided acquisition. We describe the platform architecture, training methodology, segmentation architectures, reproducibility and versioning mechanisms, and FAIR compliance, and illustrate the complete workflow through two intestinal organoid imaging examples. arivis Cloud is freely accessible to student users; other users access the platform via subscription at https://www.arivis.cloud/.

bioinformatics↗

bioRxiv2026

Parametric Physics-Based Synthesis of 3D Fluorescence Organoid Images with Exact Ground Truth for Deep Learning Pipeline Development

1Three-dimensional organoid cultures have emerged as powerful models for studying human tissue biology, disease mechanisms, and drug responses. Fluorescence confocal microscopy of organoids generates complex volumetric image data that is increasingly analyzed using deep learning pipelines for cell segmentation, morphometry, and phenotyping. However, training and benchmarking such pipelines requires large annotated datasets, the manual curation of which is prohibitively expensive and time-consuming. Here we present a parametric, physics-based computational framework for generating synthetic 3D fluorescence organoid images with exact ground-truth cell body and nucleus label masks. The framework models cell placement using force-directed sphere packing with optional hollow lumen exclusion for cyst-forming organoids, cell morphology using power-diagram (Laguerre) tessellation with apical-basal elongation and surface flattening for polarized epithelial cells, membrane curvature using low-frequency coordinate displacement, nuclear shape using irregular ellipsoid deformation with smooth radial eccentricity direction blending, and optical effects using depth-dependent point-spread function broadening, a physically motivated staining diffusion gradient with residual interior plateau, z-attenuation, haze, shot noise, and channel crosstalk. The necrotic core model uses a three-phenotype nuclear population, pyknotic, ghost, and karyorrhectic, reflecting the histological diversity of real necrotic zones. Five condition-specific presets are provided, calibrated to published biological measurements and covering PDAC osmotic stress, HMECyst normal and cyst-forming organoids, and a large primary PDAC organoid with a necrotic core. Unlike generative adversarial network approaches, our method requires no training data, produces exact ground truth by construction, and allows systematic and interpretable control over every morphological and optical parameter. The framework is released as open-source Python software with a PyQt5 graphical interface and produces OME-TIFF output compatible with arivis Pro, FIJI, and napari, as well as most other microscopy image analysis software.

bioinformatics↗