Wavelet-Based Particle Localization for Real-Time Super-Resolution Microscopy

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current super-resolution optical microscopy techniques face challenges in real-time data processing and localization of particles in three-dimensional space, leading to time-consuming post-processing and inability to view results during acquisition, which hampers experimental adjustments and efficiency.

Innovation Solution

The use of a processor with a GPU configuration performing wavelet decomposition and segmentation of images, allowing real-time or near real-time localization of particles in 3D space with high precision, utilizing asymmetric point-spread functions and adjusting imaging parameters during acquisition.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If traditional Gaussian fitting is used for particle localization, then measurement precision is achieved, but processing time is excessively long

Engineering Contradiction:
Improveparticle localization accuracyVSAvoidprocessing time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent segments the image processing task into multiple stages: wavelet decomposition to identify candidate regions, thresholding to isolate potential particles, and centroid calculation for localization. This segmentation avoids the computationally intensive iterative Gaussian fitting while maintaining nanometer-scale localization accuracy through a pipeline that processes images in parallel on GPUs.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent replaces the mechanical iterative optimization process of Gaussian fitting with a direct computational approach using wavelet transforms and centroid calculations. This substitution eliminates the need for repeated trial-and-error fitting iterations, achieving comparable precision through mathematical transformation and direct computation instead of iterative mechanical adjustment.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

2Measurement precision

If sequential post-processing is used for super-resolution imaging, then measurement precision is maintained, but productivity is severely reduced

Engineering Contradiction:
Improvesuper-resolution image qualityVSAvoidimage acquisition throughput
Core Design Contradiction:
Measurement precisionVSProductivity

Solution Approach 1:

The patent implements continuous processing by executing the wavelet decomposition, thresholding, and centroid calculation pipeline without interruption during image acquisition. The GPU-based parallel processing ensures that each frame is processed as it is captured, maintaining uninterrupted super-resolution image reconstruction and enabling real-time visualization throughout the acquisition sequence.

Inventive Principle:
Principle #20Continuity of useful action

Solution Approach 2:

The patent performs preliminary wavelet decomposition and region identification before final centroid calculation and image reconstruction. By pre-processing images to identify candidate particle regions and prepare data structures in advance, the system reduces the computational burden during final reconstruction, enabling faster overall processing while maintaining image quality.

Inventive Principle:
Principle #10Preliminary action

3Productivity

If high frame rate acquisition is used, then productivity is improved, but processing complexity increases

Engineering Contradiction:
Improveacquisition frame rateVSAvoidprocessing system complexity
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The patent replaces complex iterative processing algorithms with direct wavelet transform and centroid calculation methods that are inherently more suitable for parallel execution. This substitution simplifies the processing pipeline while enabling high frame rates by reducing the computational complexity of each processing step and making the system more amenable to GPU acceleration.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The patent transitions from sequential CPU-based processing to parallel GPU-based processing, adding a dimensional aspect to the computational architecture. This dimensional change in processing architecture enables simultaneous execution of multiple processing operations, dramatically increasing throughput and handling the complexity of high frame rate acquisition without proportionally increasing processing time.

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

Data Source

PatentEP2966492B1Method and apparatus for single-particle localization using wavelet analysis
Publication Date: 2020.10.21 CENT NAT DE LA RECH SCI (C N R S)
  • EP2966492B1 patent drawingFigure 1
  • EP2966492B1 patent drawingFigure 2A
  • EP2966492B1 patent drawingFigure 2B

AI summary

Accurate localization of isolated particles is important in single particle based super-resolution microscopy. It allows the imaging of biological samples with nanometer-scale resolution using a simple fluorescence microscopy setup. Nevertheless, conventional techniques for localizing single particles can take minutes to hours of computation time because they require up to a million localizations to form an image. In contrast, the present particle localization techniques use wavelet-based image decomposition and image segmentation to achieve nanometer-scale resolution in two dimensions within seconds to minutes. This two-dimensional localization can be augmented with localization in a third dimension based on a fit to the imaging system's point-spread function (PSF), which may be asymmetric along the optical axis. For an astigmatic imaging system, the PSF is an ellipse whose eccentricity and orientation varies along the optical axis. When implemented with a mix of CPU/GPU processing, the present techniques are fast enough to localize single particles while imaging (in real-time).