RIEDm

luminescent Reaction enabled super-resolution Imaging via Entropy-weighted correlation combined with Deconvolution.

v0.8.0

This repository is for RIED reconstruction, and it will be in continued development. It is distributed as accompanying software for publication: Luminescent-reaction-enabled super-resolution imaging, Nature (2026) . Please cite RIED in your publications if it helps your research.

📖 Introduction | 🧬 RIED reconstruction workflow | 🔧 Installation | 🌐 Parameters | 🚀 Example demonstration | 📦 Version | 🔗 Resources

📖 Introduction

The RIED is a new conceptual and methodological framework for super-resolution luminescence imaging. It establishes that fluctuation information inherent in reaction-driven luminescence—including electrochemiluminescence (ECL), chemiluminescence (CL), and bioluminescence (BL)—can be harnessed to achieve super-resolution reconstruction, enabling zero-background, ultra-long-term super-resolution imaging with sub-100 nm resolution. Within this framework, we developed a specific computational reconstruction algorithm, which implements entropy-weighted correlation with dual-step deconvolution to extract super-resolved information from continuously collected short-exposure frames.

🧬 RIED reconstruction workflow

Main step	Function
Pre-processing	Gaussian pre-filtering to suppress high-frequency sampling noise; Fourier interpolation to provide a finer grid and ensures sufficient pixel support for subsequent resolution enhancement; Pre-deconvolution to further suppress sampling noise and enhance fluctuation.
Entropy-weighted correlation	Identify emitters and extract super-resolved information
Post sparse deconvolution	Maximize resolution enhancement without introducing artefacts

🔧 Installation

• Tested platforms

Matlab 2022b, with Wavelet Toolbox, Image Processing Toolbox, Parallel Computing Toolbox (Win 10, 128 GB RAM, NVIDIA RTX 4090 24 GB, CUDA 11.6)

• Quick start

% Add RIEDm to your path
addpath(genpath('RIEDm'));

% Basic reconstruction (GPU accelarated)
imgstack = imreadstack('bl.tif'); 
RIEDrecon = RIEDm(imgstack);

% Reconstruction based on pure CPU
imgstack = imreadstack('bl.tif','gpu',0); 
RIEDrecon = RIEDm(imgstack);

🌐 Parameters

• Basic parameters

Parameter	Description	Default	When to adjust
pixel	Pixel size (nm)	160	Match the microscope acquisition
NA	Numerical aperture	1.45	Match the objective lens
wavelength	Emission wavelength (nm)	620	Match the luminescence peak

• Advanced parameters

Parameter	Description	Recommended Range	Default	When to adjust
gauss	Gaussian pre-filter kernel size	0.5-2	0.7	Higher for low photon budget
iter1	Pre-deconvolution iterations	2-10	4	Increase to further enhance fluctuation and resolution, reduce if introduces artefacts
subfactor	Subtraction factor for cumulant	0-1	0.5	Increase if high fluctuation
wavelet	Weight for wavelet defocus signal estimation	0; 1-5	0	Set to '0' for ECL, where defocus signal is negligible. For CL and BL, lower values remove defocus signal more aggressively
fidelity	Fidelity weight for sparse deconvolution	10-100	50	Lower for high continuity constraint
sparsity	Sparsity weight for sparse deconvolution	0.1-5	0.5	Higher for further resolution improvement, Lower if real structure signals are filtered out
finter2	Factor of fourier interpolation after entropy-weighted correlation	1-2	1	'1' is enough for most cases. Increase if higher resolution demand.
iter2	Sparse deconvolution iterations	2-15	5	Increase to further enhance fluctuation, reduce if introduces artefacts
gamma	Factor of intensity correction	0.5-1	1	reduce to further intensity correction
gpu	Enable GPU acceleration	1	1	Enable GPU acceleration to speed up processing. '0' for pure CPU computation.

🚀 Example demonstration

• Reconstruction for ECL

(microtubules, 50 frames)

% Involve RIED
addpath(genpath('./RIED_core'));
addpath(genpath('./Utils'));

% RIED recon
imgstack = imreadstack('RIEDm_data/ecl2.tif');
RIEDrecon = RIEDm(imgstack,'pixel',160,'NA',1.45,'wavelength',620,'iter1',3,'subfactor',0.6,'fidelity',50,'sparsity',9,'iter2',2);

% Visualization
visualize(imgstack, RIEDrecon, 480, 1, [1, 99.98], 'Green');

• Reconstruction for BL

(mitochondria, 200 frames)

% Involve RIED
addpath(genpath('./RIED_core'));
addpath(genpath('./Utils'));

% RIED recon
imgstack = imreadstack('RIEDm_data/bl.tif');
RIEDrecon = RIEDm(imgstack,'pixel',160,'NA',1.42,'wavelength',517,'iter1',5,'subfactor',0,'fidelity',150,'sparsity',20,'iter2',5);

% Visualization
visualize(imgstack, RIEDrecon, 480, 1, [1, 99.98], 'Yellowhot');

📦 Version

Version	Changes
v0.8.0	RIED reconstruction core for ECL, CL, and BL
v0.2.0	RIED reconstruction core for ECL
v0.1.0	initial version

🔗 Resources

• Preprint: Reaction-enabled, highly sensitive super-resolution imaging, bioRxiv (2026).

• Publication: Luminescent-reaction-enabled super-resolution imaging, Nature (2026).

Open source RIED

This software and corresponding methods can only be used for non-commercial use, and they are under Open Data Commons Open Database License v1.0.