Guardado en:
Detalles Bibliográficos
Autores principales: Soltaninezhad, Mohammad, Rouzbahani, Yashar, Contreras, Jhonatan, Chippalkatti, Rohan, Abankwa, Daniel Kwaku, Eggeling, Christian, Bocklitz, Thomas
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:https://arxiv.org/abs/2510.15579
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866909852712828928
author Soltaninezhad, Mohammad
Rouzbahani, Yashar
Contreras, Jhonatan
Chippalkatti, Rohan
Abankwa, Daniel Kwaku
Eggeling, Christian
Bocklitz, Thomas
author_facet Soltaninezhad, Mohammad
Rouzbahani, Yashar
Contreras, Jhonatan
Chippalkatti, Rohan
Abankwa, Daniel Kwaku
Eggeling, Christian
Bocklitz, Thomas
contents Lightweight deep learning models offer substantial reductions in computational cost and environmental impact, making them crucial for scientific applications. We present a lightweight CycleGAN for modality transfer in fluorescence microscopy (confocal to super-resolution STED/deconvolved STED), addressing the common challenge of unpaired datasets. By replacing the traditional channel-doubling strategy in the U-Net-based generator with a fixed channel approach, we drastically reduce trainable parameters from 41.8 million to approximately nine thousand, achieving superior performance with faster training and lower memory usage. We also introduce the GAN as a diagnostic tool for experimental and labeling quality. When trained on high-quality images, the GAN learns the characteristics of optimal imaging; deviations between its generated outputs and new experimental images can reveal issues such as photobleaching, artifacts, or inaccurate labeling. This establishes the model as a practical tool for validating experimental accuracy and image fidelity in microscopy workflows.
format Preprint
id arxiv_https___arxiv_org_abs_2510_15579
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Lightweight CycleGAN Models for Cross-Modality Image Transformation and Experimental Quality Assessment in Fluorescence Microscopy
Soltaninezhad, Mohammad
Rouzbahani, Yashar
Contreras, Jhonatan
Chippalkatti, Rohan
Abankwa, Daniel Kwaku
Eggeling, Christian
Bocklitz, Thomas
Computer Vision and Pattern Recognition
Artificial Intelligence
Lightweight deep learning models offer substantial reductions in computational cost and environmental impact, making them crucial for scientific applications. We present a lightweight CycleGAN for modality transfer in fluorescence microscopy (confocal to super-resolution STED/deconvolved STED), addressing the common challenge of unpaired datasets. By replacing the traditional channel-doubling strategy in the U-Net-based generator with a fixed channel approach, we drastically reduce trainable parameters from 41.8 million to approximately nine thousand, achieving superior performance with faster training and lower memory usage. We also introduce the GAN as a diagnostic tool for experimental and labeling quality. When trained on high-quality images, the GAN learns the characteristics of optimal imaging; deviations between its generated outputs and new experimental images can reveal issues such as photobleaching, artifacts, or inaccurate labeling. This establishes the model as a practical tool for validating experimental accuracy and image fidelity in microscopy workflows.
title Lightweight CycleGAN Models for Cross-Modality Image Transformation and Experimental Quality Assessment in Fluorescence Microscopy
topic Computer Vision and Pattern Recognition
Artificial Intelligence
url https://arxiv.org/abs/2510.15579