Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

An Efficient Inference Frame for SMLM (Single-Molecule Localization Microscopy)

Tingdan Luo

TL;DR

This paper proposes an efficient model deployment framework and introduces a lightweight neural network, DilatedLoc, aimed at enhancing both image reconstruction quality and inference speed, with superior GPU utilization through a novel deployment architecture compatible with various network models.

Abstract

Single-molecule localization microscopy (SMLM) surpasses the diffraction limit, achieving subcellular resolution. Traditional SMLM analysis methods often rely on point spread function (PSF) model fitting, limiting the application of complex PSF models. In recent years, deep learning approaches have significantly improved SMLM algorithms, yielding promising results. However, limitations in inference speed and model size have restricted the widespread adoption of deep learning in practical applications. To address these challenges, this paper proposes an efficient model deployment framework and introduces a lightweight neural network, DilatedLoc, aimed at enhancing both image reconstruction quality and inference speed. Compared to leading network models, DilatedLoc reduces network parameters to under 100 MB and achieves a 50% improvement in inference speed, with superior GPU utilization through a novel deployment architecture compatible with various network models.

An Efficient Inference Frame for SMLM (Single-Molecule Localization Microscopy)

TL;DR

This paper proposes an efficient model deployment framework and introduces a lightweight neural network, DilatedLoc, aimed at enhancing both image reconstruction quality and inference speed, with superior GPU utilization through a novel deployment architecture compatible with various network models.

Abstract

Single-molecule localization microscopy (SMLM) surpasses the diffraction limit, achieving subcellular resolution. Traditional SMLM analysis methods often rely on point spread function (PSF) model fitting, limiting the application of complex PSF models. In recent years, deep learning approaches have significantly improved SMLM algorithms, yielding promising results. However, limitations in inference speed and model size have restricted the widespread adoption of deep learning in practical applications. To address these challenges, this paper proposes an efficient model deployment framework and introduces a lightweight neural network, DilatedLoc, aimed at enhancing both image reconstruction quality and inference speed. Compared to leading network models, DilatedLoc reduces network parameters to under 100 MB and achieves a 50% improvement in inference speed, with superior GPU utilization through a novel deployment architecture compatible with various network models.
Paper Structure (19 sections, 1 equation, 5 figures, 3 tables)

This paper contains 19 sections, 1 equation, 5 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Efficient Inference Frame
  3. Overview
  4. The Structure of the Frame
  5. DilatedLoc
  6. Overview
  7. The Structure of DilatedLoc
  8. Dense Connection Dilated Block
  9. Loss Function
  10. Ablation Experiments
  11. Experiments
  12. Performance of DilatedLoc
  13. Cramer-Rao Lower Bound
  14. SMLM chanage
  15. High resolution on Large Axial PSFs
  16. ...and 4 more sections

Figures (5)

  • Figure 1: Diagram of the reasoning framework. a.CPU, GPU occupancy time for traditional inference vs CPU, GPU occupancy time for high-throughput parallel inference.b
  • Figure 2: The structure of DilatedLoc
  • Figure 3: DilatedLoc, Fd_DeepLoc compare the positioning accuracy of four point spread functions with CRLB. a. Astigmatism point spread function; b. Saddle point spread function; c.3$\mu$m tetrapod point spread function; d.6$\mu$m tetrapod point spread functions
  • Figure 4: DilatedLoc,Fd_DeepLoc, DECODE evaluation results on images generated by four point spread function simulations. The figure shows three metrics, namely recall, root mean square error, 3D efficiency value. a. astigmatism point spread function; b. saddle point spread function; c.3$\mu$m quadruped point spread function; d.6$\mu$m quadruped point spread function
  • Figure 5: Performance of throughput with the increase of the number of GPU devices