FLAMEm

FLuctuation-based high-order super-resolution Acoustic MicroscopE

FLAME reconstruction with MATLAB

This repository is for our developed FLuctuation-based high-order super-resolution Acoustic MicroscopE (FLAME), and it will be in continued development. It is distributed as accompanying software for publication: Weisong Zhao et al. High-throughput 3D super-resolution ultrasound imaging, Nature, submitted (2025). Please cite FLAME in your publications, if it helps your research.

FLAME reconstruction

Instruction

Load the target mat file using load and change the variable name to input. The FLAME reconstruction requires some parameters.

Necessary Parameters

Important parameters that must be set according to actual needs.

SVD_option Enable SVD filtering. {default: 0}

MB_option Enable MB (multi-band) direction filtering. {default: 0}

pixel Pixel size of input data (µm). {default: 60}

fidelity Sparsity reconstruction fidelity (controls data fidelity term weight). {default: 200}

sparsity Sparsity reconstruction strength (controls sparsity term weight). {default: 10}

FWHM2 Full-width half-maximum (FWHM) of post-deconvolution kernel (µm). {default: 240}

iter2Number of post-deconvolution iterations. {default: 15}

Expert parameters

Some adjustable parameters that can optimize the reconstruction results.

Stab_option Remove unstable frames (e.g., due to breathing/heartbeat). {default: 1}

cutoff1 Low threshold for SVD filtering (range: 0–1). {default: 0.25}

cutoff2 High threshold for SVD filtering (range: 0–1). {default: 0.8}

BF_option1 Enable additional background filtering. Note: Significantly reduces speed. {default: 0}

finter1 First upsampling factor. Tips: Improves quality but reduces speed/increases memory.

​ Increase only with proportional reduction in fidelity/sparsity. {default: 2}

FWHM1 FWHM of pre-deconvolution kernel (µm). {default: 180}

iter1 Number of pre-deconvolution iterations. {default: 10}

hawk_option Enable HAWK processing. Note: Improves quality but increases memory usage. {default: 0}

order Autocorrelation order.

​ Tips: Higher values improve resolution but reduce image continuity/linearity. {default: 6}

finter2 Second upsampling factor. {default: 2}

fidelity_z Z-axis fidelity weight. Use 1 for isotropic data. {default: 1}

BF_option2 Secondary background filtering. Note: Significantly reduces speed. {default: 0}

Here are 4 examples:

[output_CEUS, output_deconv_n, output_deconv_p] = FLAME(input,'pixel',60,'FWHM2',330);
[output_CEUS, output_deconv_n, output_deconv_p] = FLAME(input,'MB_option',1,'fidelity',10,'sparsity',1);
[output_CEUS, output_deconv_n, output_deconv_p] = FLAME(input,'SVD_option',1,'MB_option',1,'cutoff1',0.1,'cutoff2',0.9);
[output_CEUS, output_deconv_n, output_deconv_p] = FLAME(input,'iter1',5,'iter2',30);

Fusion

Generate better quality intensity and flow velocity images using 4 ultra fast SR frames.

for k = 1:floor(size(data,4)/120)
[intensity_n, intensity_p, speed] = fusion(SR_volume_n(:,:,:,(k-1)*4+1:(k-1)*4+4),SR_volume_p(:,:,:,(k-1)*4+1:(k-1)*4+4));
end

Rolling fusion can also be chosen to obtain fusion results with higher temporal resolution.

for k = 1:floor(size(data,4)/30)-3
[intensity_n, intensity_p, speed] = fusion(SR_volume_n(:,:,:,k:k+3),SR_volume_p(:,:,:,k:k+3));
end

Visualization

Use FLAME's specially designed color encoding to render the final result.

rendering(intensity_n, intensity_p, speed, output_CEUS,'MB_option',1);

You can also export a mat file containing the results and render it using other software.

Declaration

This repository contains the MATLAB source code for FLAME.

Open source FLAMEm

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.