AI-Assisted Hyperspectral Interferometry and Single-Cell Dispersion Imaging

Interferometry techniques are essential for extracting phase information from optical systems enabling precise measurements of dispersion and highly sensitive detection of perturbations. While phase sensing offers enhanced sensitivity compared to conventional spectroscopy methods, this sensitivity often makes systems more vulnerable to external factors such as vibrations, introducing instability and noise. In this work, we demonstrate a broadband and AI-enhanced interferometry method, denoted general polarization common-path interferometry (GPCPI) that relaxes the polarization constraints of traditional common-path interferometry. The polarization decoupling feature enables simultaneous amplitude and phase measurements supplemented with deep neural autoencoders to detect phase anomalies in the spectrum through the analysis of second order derivative mapping of the phase profile, enhancing the accuracy of broadband phase measurements. The approach enables an order of magnitude improvement in phase stability compared to state-of-the-art interferometry techniques, leading to higher accuracy in phase sensing. Plasmonic metasurface phase sensing and hyperspectral single-cell dispersion imaging demonstrate the capability and sensitivity of the method over conventional spectroscopy. Our adopted version of deep learning model, ConvNeXt V2, enables real-time tracking of phase variation with minimized noise. Interference fringes affected by the cell-cultured samples reveal the fingerprints of the normal (CCD-32Sk) vs cancerous (COLO-829) skin cells, enabling cell classification and disease diagnosis at single-cell level through hyperspectral dispersion imaging. The proposed technique offers a reliable, compact, and stable solution for broadband phase measurements and single-cell dispersion imaging for applications in metrology, molecular diagnostics, drug discovery, and quantum sensing.