Missing Wedge Completion via Unsupervised Learning with Coordinate Networks.

Area of Science:

• Structural biology
• Biophysics
• Imaging science

Background:

• Cryo-electron tomography (cryo-ET) provides nanoscale 3D imaging of biological samples.
• The missing wedge artifact, caused by incomplete angular data, degrades reconstruction quality.
• Supervised deep learning methods (CNNs) show promise but require extensive pretraining and can be sensitive to data scarcity.

Approach:

• Introduced a proof-of-concept unsupervised learning method using coordinate networks (CNs).
• CNs optimize network weights directly against input projections, eliminating pretraining requirements.
• Evaluated performance using voxel-based metrics and a novel directional Fourier Shell Correlation (FSC).

Key Points:

• Unsupervised CNs reduce reconstruction runtime by 3-20x compared to supervised methods.
• Demonstrated improved shape completion and mitigation of missing wedge artifacts.
• The approach is less susceptible to inaccuracies arising from limited training data.

Conclusions:

• Unsupervised learning offers a viable and efficient alternative for cryo-ET reconstruction.
• This method enhances image quality and reduces artifacts without extensive pretraining.
• Highlights the benefits and considerations of both supervised and unsupervised deep learning strategies for cryo-ET.