Super-resolved Robust Principal Component Analysis (SRPCA)
Table of Contents
Overview
We present the codes for the SRPCA (Super-Resolved Robust Principal Component Analysis) paper.
SRPCA is intended to be used in conjunction with the code associated with the OPUS-PALA article.
SRPCA is a deterministic method that enhances ultrasound localization microscopy (ULM) by improving the isolation and localization of overlapping microbubbles (MBs). It combines robust principal component analysis with complex-valued deconvolution, using the alternating direction method of multipliers optimization to address the joint problem of IQ deconvolution and RPCA with a measured system point spread function. This integration replaces the traditional ULM steps of tissue filtering, MB detection, and super-localization with a single super-resolution inverse problem. SRPCA effectively separates MB signals from background noise and enhances the resolution of the MB signals.
In the attached codes, SRPCA was applied to the rat brain dataset from the supplementary PALA dataset (https://doi.org/10.5281/zenodo.7883227).
Requirements
- MATLAB: R2020a or newer (Previous versions might work)
-
Required Toolboxes:
- Communications
- Bioinformatics
- Image Processing
- Curve Fitting
- Signal Processing
- Statistics and Machine Learning
- Parallel Computing
- Computer Vision Toolbox
- RAM: 64GB or more recommended for SRPCA
Installation and Usage
Installation
- First, install the OPUS-PALA software.
- Download the supplementary rat brain PALA dataset (https://doi.org/10.5281/zenodo.7883227).
- Decompress the zip files into a single folder.
- Place the provided
PALA_InVivoULM_Supplementary.m
andPALA_InVivoULM_SRPCA.m
scripts in thePALA_scripts
folder. - Place the provided
SRPCA.m
andULM_localization2DSR.m
functions in thePALA_addons
folder. - Modify file paths and locations in
PALA_InVivoULM_Supplementary.m
andPALA_InVivoULM_SRPCA.m
to match the folder where you decompressed the rat brain data.
Usage
- Run
PALA_InVivoULM_Supplementary.m
to compute the result of the ULM process (with radial symmetry super-localization) as presented in the PALA article, referred to as RS ULM in the SRPCA paper. - Run
PALA_InVivoULM_SRPCA.m
to obtain the SRPCA outcome.
SRPCA Paper Results
We provide the final rendering results of the rat brain dataset (Fig. 5 & 6):
-
SRPCA, SR=4, Full dataset (250 blocks):
SRPCAexample_matouts_250blocks.mat
-
SRPCA, SR=4, First 40 blocks:
SRPCAexample_matouts_40blocks.mat
-
RS ULM, SR=10, Full dataset (250 blocks):
example_matouts_250blocks.mat
-
RS ULM, SR=10, First 40 blocks:
example_matouts_40blocks.mat
Copyright / Warranty
This code is for academic purposes only.
Copyright © 2024 Vassili Pustovalov, Duong Hung Pham, Denis Kouamé, IRIT Laboratory
Contact: vassili.pustovalov@irit.fr
MATLAB is a registered trademark of The MathWorks.
All PALA functions are usable with agreement from their owner.
The authors claim no responsibility for this software and code.