Parameter-Free Attention Super-Resolution Network for Accelerated AFM Biological Imaging.

Area of Science:

• Biophysics
• Microscopy
• Artificial Intelligence

Background:

• Atomic force microscopy (AFM) is crucial for high-resolution topographic imaging of biological samples.
• Traditional raster scanning in AFM limits image acquisition speed, hindering high-throughput studies.
• Accelerating AFM imaging is essential for rapid characterization of biological specimens.

Purpose of the Study:

• To develop a super-resolution (SR) model to accelerate AFM imaging.
• To introduce the parameter-free channel-spatial attention network (PCSAN) for rapid, high-fidelity AFM reconstructions.
• To enable high-throughput AFM characterization of viral samples.

Main Methods:

• A paired dataset of high-resolution (HR) and low-resolution (LR) virus images was created for supervised learning.
• The PCSAN model utilizes a parameter-free channel-spatial attention mechanism for feature enhancement and noise reduction.
• Trained PCSAN directly reconstructs HR images from newly acquired LR AFM data.

Main Results:

• PCSAN achieved a peak signal-to-noise ratio (PSNR) of 38.73 dB at a ×2 scaling factor on SARS-CoV-2 and influenza datasets.
• The model reduced AFM imaging time by over 50% while preserving image integrity.
• Inference latency was significantly reduced due to the streamlined PCSAN architecture.

Conclusions:

• The PCSAN model effectively accelerates AFM imaging without compromising image quality.
• This AI-driven approach provides a robust foundation for high-throughput AFM characterization.
• PCSAN offers a significant advancement in rapid topographic imaging of biological specimens.