CC-DenseSTORM: deep learning enables colorimetry camera-based simultaneous two-color single-molecule localization...

Area of Science:

• Nanotechnology
• Biophysics
• Microscopy

Background:

• Colorimetry camera-based single-molecule localization microscopy (CC-STORM) enables nanoscale imaging of multiple targets.
• Existing deep learning methods like CC-DeepSTORM reduce data rejection but struggle with dense emitters, causing artifacts and low detection rates.

Purpose of the Study:

• To develop an improved deep learning algorithm (CC-DenseSTORM) for high-density emitter localization in CC-STORM.
• To address structural artifacts and enhance detection accuracy in dense samples.

Main Methods:

• Implementation of an attention-gated standard-convolution U-Net architecture.
• Development of a dual-channel adaptive classification network for robust dye identification.

Main Results:

• CC-DenseSTORM improved emitter detection rates by 2-fold compared to CC-DeepSTORM at densities up to 5 emitters/µm².
• Maintained a data rejection rate below 30% and achieved <1% crosstalk in experimental validation.
• Successfully enabled simultaneous quantification of CD38 and BCMA densities in multiple myeloma cells.

Conclusions:

• CC-DenseSTORM effectively overcomes limitations of previous methods for dense emitter imaging in CC-STORM.
• The algorithm shows significant potential for advancing dual-target immunotherapy research through precise molecular quantification.