Empirical Validation of SNR-Dependent Deep Learning Architectures for Label-Free Super-Resolution in Optical Microscopy
6 affiliations
- 1Newcastle University in Singapore, 1 Punggol Coast Road, Block E1, Level 2, Singapore 828608, Singapore
- 2Faculty of Science, Agriculture and Engineering (SAgE), Newcastle University, Newcastle upon Tyne NE1 7RU, UK
- 3Department of Electrical, Computer, and Biomedical Engineering, Toronto Metropolitan University, Toronto, ON M5B 2K3, Canada
- 4Cavendish Laboratory, University of Cambridge, JJ Thomson Avenue, Cambridge CB3 0US, UK
- 5Engineering Cluster, Singapore Institute of Technology, 10 Dover Drive, Singapore 138683, Singapore
- 6School of Computer Science, Northumbria University, Newcastle upon Tyne NE1 8ST, UK
Abstract
Optical microscopy is widely used across scientific, industrial, and clinical settings, yet its lateral resolution is fundamentally limited by diffraction (down to ~200 nm in 400 nm visible light). While super-resolution techniques can overcome this barrier, most existing approaches rely on specialised hardware or fluorescent labelling, restricting their accessibility in routine microscopy workflows. As such, past research studies (including our own) have focused on developing deep neural network (DNN) architectures capable of super-resolving optical micrographs. In the present context, two of our previously developed DNN architectures, termed O-Net and Θ-Net, are systematically evaluated using a novel custom-fabricated optical test target containing nanoscale features independently calibrated by atomic force microscopy (AFM). The results indicate that both architectures surpass the diffraction-limited resolution of the optical system employing halogen illumination (peak λem = 619.4 nm; λem¯ = 612.2 nm), resolving features down to 170 nm in PCM and 200 nm in DIC (at least). Additionally, the models exemplify a complementary SNR-dependent behaviour—O-Net achieves superior performance under high-SNR conditions, while Θ-Net demonstrates enhanced robustness when SNR is reduced. These findings have important implications in computational super-resolution microscopy, highlighting the potential applicability of the assayed model architectures under differing image acquisition protocols.
Related
- Correlative super-resolution microscopy reveals preferential localization of quantum color center formation on carbon nanotubes
- A Molecular Generation–Validation Workflow Combining Reversible Junction Tree Reinforcement Learning and Semi-empirical Quantum Chemistry: A Case Study on Molecular Solar Thermal Fuels
- Quantifying Validation-Induced Optimism in Machine Learning for Food Spectroscopy: An Empirical Multi-Dataset Benchmark
- Polymer Network Evolution Induced by Dual-Crosslinking Architectures in UV-Curable Waterborne Polyurethanes
- A Scoping Review of Diffusion-Based De Novo Molecular Generation: Architectures, Evaluation, and Open Challenges for Computational Drug Discovery
BibTeX
@article{kaderuppan2026empirical,
title = {{Empirical Validation of SNR-Dependent Deep Learning Architectures for Label-Free Super-Resolution in Optical Microscopy}},
author = {Shiraz S. Kaderuppan and Jonathan D. Mar and Andrew C. Irvine and Anurag Sharma and Muhammad Ramadan Saifuddin and Wai Leong Eugene Wong and Wai Lok Woo},
journal = {Nanomaterials},
year = {2026},
month = oct,
doi = {10.3390/nano16201334},
url = {https://www.mdpi.com/2079-4991/16/20/1334},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Empirical Validation of SNR-Dependent Deep Learning Architectures for Label-Free Super-Resolution in Optical Microscopy
AU - Shiraz S. Kaderuppan
AU - Jonathan D. Mar
AU - Andrew C. Irvine
AU - Anurag Sharma
AU - Muhammad Ramadan Saifuddin
AU - Wai Leong Eugene Wong
AU - Wai Lok Woo
PY - 2026
DA - 2026/10/01
JO - Nanomaterials
DO - 10.3390/nano16201334
UR - https://www.mdpi.com/2079-4991/16/20/1334
AB - Optical microscopy is widely used across scientific, industrial, and clinical settings, yet its lateral resolution is fundamentally limited by diffraction (down to ~200 nm in 400 nm visible light). While super-resolution techniques can overcome this barrier, most existing approaches rely on specialised hardware or fluorescent labelling, restricting their accessibility in routine microscopy workflows. As such, past research studies (including our own) have focused on developing deep neural network (DNN) architectures capable of super-resolving optical micrographs. In the present context, two of our previously developed DNN architectures, termed O-Net and Θ-Net, are systematically evaluated using a novel custom-fabricated optical test target containing nanoscale features independently calibrated by atomic force microscopy (AFM). The results indicate that both architectures surpass the diffraction-limited resolution of the optical system employing halogen illumination (peak λem = 619.4 nm; λem¯ = 612.2 nm), resolving features down to 170 nm in PCM and 200 nm in DIC (at least). Additionally, the models exemplify a complementary SNR-dependent behaviour—O-Net achieves superior performance under high-SNR conditions, while Θ-Net demonstrates enhanced robustness when SNR is reduced. These findings have important implications in computational super-resolution microscopy, highlighting the potential applicability of the assayed model architectures under differing image acquisition protocols.
SN - 2079-4991
ER -