Quantitative phase gradient metrology using diffraction phase microscopy and deep learning.

Area of Science:

• Biophysics
• Optical Microscopy
• Computational Imaging

Background:

• Quantitative phase microscopy (QPM) is crucial for label-free imaging of biological cells.
• Accurate measurement of the phase gradient is essential for reconstructing cell morphology.
• Existing methods often require complex post-processing steps like phase unwrapping and numerical differentiation.

Purpose of the Study:

• To develop a direct phase gradient estimation method using deep learning.
• To eliminate the need for phase unwrapping and numerical differentiation in QPM.
• To demonstrate the method's applicability to biological cell imaging.

Main Methods:

• A deep learning model was designed for direct phase gradient estimation.
• Numerical simulations were performed to assess robustness under noisy conditions.
• The method was validated using a diffraction phase microscopy setup for imaging biological cells.

Main Results:

• The deep learning approach successfully estimated phase gradients directly.
• The method demonstrated robustness even under severe noise conditions.
• The technique was effective for imaging diverse biological cell types.

Conclusions:

• Deep learning offers a powerful alternative for phase gradient estimation in QPM.
• The proposed method simplifies the analysis of QPM data for cell morphology.
• This technique enhances the utility of QPM for biological applications.