On cross-correlations, averages and noise in electron microscopy

Area of Science:

• Structural biology
• Biophysics
• Microscopy imaging

Background:

• Biological samples are sensitive to radiation, necessitating low-dose imaging.
• Low-dose imaging results in high noise and low signal-to-noise ratios (SNR < 1).
• Averaging techniques are crucial for overcoming noise limitations in image analysis.

Observation:

• Averaging 2D images or 3D subvolumes enhances structural resolution.
• Cross-correlation methods are widely used for aligning and classifying images and volumes in structural biology.
• High noise levels can introduce bias in alignment and classification, particularly in multi-reference and projection-based alignments.

Findings:

• The study analyzes the direct influence of image signal-to-noise ratio on cross-correlation coefficient values.
• A novel method is presented to compensate for the effects of noise on cross-correlation analysis.
• This compensation improves the accuracy of alignment and classification in high-noise imaging scenarios.

Implications:

• The findings enable more accurate 3D structure determination from noisy biological images.
• Improved alignment and classification methods can lead to higher resolution structural models.
• This research enhances the utility of low-dose imaging for sensitive biological sample analysis.