Statistical deconvolution for superresolution fluorescence microscopy.

Area of Science:

• Biophysics
• Optical Microscopy
• Computational Biology

Background:

• Superresolution microscopy achieves ~10 nm resolution by sequentially activating fluorophores.
• Current methods require sparse fluorophore distribution, discarding data from overlapping molecules.
• This data loss limits imaging speed and the study of dynamic biological processes.

Purpose of the Study:

• To develop a computational method that utilizes all fluorescence data, including overlapping signals, to improve superresolution imaging speed.
• To enable the study of fast, dynamic biological processes by increasing the number of molecules processed per frame.

Main Methods:

• Introduced deconvolution-STORM (deconSTORM), a computational method using iterative image deconvolution.
• DeconSTORM estimates the sample by approximating maximum likelihood under a statistical model.
• The model incorporates Poisson noise, sparse fluorophore distribution, and temporal correlations.

Main Results:

• Validated deconSTORM with simulated data, showing accurate superresolution image estimation even at high fluorophore densities.
• Demonstrated deconSTORM's effectiveness on experimental cellular structure data.
• Achieved an approximately fivefold or greater increase in imaging speed by allowing higher fluorophore densities per frame.

Conclusions:

• Deconvolution-STORM (deconSTORM) overcomes limitations of traditional single- or multiemitter localization methods.
• This approach significantly accelerates superresolution imaging by exploiting all fluorescence data.
• DeconSTORM facilitates the investigation of fast biological dynamics previously inaccessible to superresolution microscopy.