AI-Assisted Hyperspectral Interferometry and Single-Cell Dispersion Imaging

Area of Science:

• Optical Metrology
• Spectroscopy
• Artificial Intelligence in Optics

Background:

• Interferometry is crucial for optical phase measurements but is sensitive to environmental noise.
• Conventional methods struggle with stability and accuracy in broadband phase sensing.
• AI-enhanced techniques are emerging to overcome limitations in optical sensing.

Purpose of the Study:

• To introduce a broadband, AI-enhanced interferometry method for improved phase stability and accuracy.
• To demonstrate simultaneous amplitude and phase measurements with enhanced sensitivity.
• To apply the technique for hyperspectral single-cell dispersion imaging and disease diagnosis.

Main Methods:

• Developed general polarization common-path interferometry (GPCPI) to relax polarization constraints.
• Utilized deep neural autoencoders for phase anomaly detection via second-order derivative mapping.
• Employed ConvNeXt V2 deep learning model for real-time phase variation tracking with noise reduction.

Main Results:

• Achieved an order of magnitude improvement in phase stability compared to existing techniques.
• Demonstrated high accuracy in plasmonic metasurface phase sensing and hyperspectral single-cell imaging.
• Successfully classified normal vs. cancerous skin cells using interference fringe analysis.

Conclusions:

• GPCPI offers a reliable, compact, and stable solution for broadband phase measurements.
• The AI-enhanced method enables sensitive single-cell dispersion imaging for diagnostics.
• This technique has broad applications in metrology, diagnostics, drug discovery, and quantum sensing.