Generative super-resolution AI accelerates nanoscale analysis of cells

• DOI: 10.1088/2632-2153/adc3e9
• Publication Date: 2025-03-21T23:17:42Z
• Authors: Simone Lossano; Simone Capaccioli; Francesca Cella Z...; Eleonora Da Pozzo; Francesca Del Debbio; Maria Evelina Fan...; Francesca Lizzi; Raffaella Magrassi; Benedetta Noferi; Dario Pisignano; Camilla Scapicchio; Alessandra Retico

Abstract Super-Resolution Microscopy (SRM) surpasses Abbe's diffraction limit, thus enabling nanoscale observation of cells. However, SRM techniques, such as Stochastic Optical Reconstruction (STORM), suffer from long acquisition times which can significantly impact imaging throughput. To address this issue, we adapted the Enhanced Generative Adversarial Network (ESRGAN) natural to microscopy images. Our goal is generate super-resolution images widefield in shorter times. We implemented for microtubules cells obtain STORM-like Different models were trained by using transfer learning and progressive fine-tuning. The generated images, evaluated PSNR, SSIM expert human validation, prove that deep approach suitable microscopy, allowing 4x-higher throughput compared unsupported techniques.