Deep-Learning-Assisted SICM for Enhanced Real-Time Imaging of Nanoscale Biological Dynamics

Area of Science:

• Biophysics
• Microscopy
• Artificial Intelligence

Background:

• Scanning Ion Conductance Microscopy (SICM) offers high-resolution nanoscale imaging of living cells.
• Its slow scan rate limits capturing rapid dynamic biological processes in real time.

Purpose of the Study:

• To enhance the temporal resolution of SICM for real-time dynamic imaging.
• To develop a computational framework for faster SICM data acquisition and image reconstruction.

Main Methods:

• An integrated data acquisition and computational framework was developed.
• A partial convolutional neural network (Partial-CNN) was trained to reconstruct full SICM images from undersampled data by skipping scan lines.
• The method was validated using quantitative metrics.

Main Results:

• The Partial-CNN approach reduced image acquisition time by 30-63% without compromising image quality.
• It demonstrated higher accuracy than conventional deep learning methods in reconstructing fine details and maintaining height map consistency.
• The method successfully increased temporal resolution while retaining image fidelity.

Conclusions:

• The proposed method significantly improves SICM's temporal resolution, enabling real-time dynamic imaging of living cells.
• This advancement enhances smart scanning microscopy applications for time-resolved biological imaging.