Phase-edge imaging using q-plate shifts for faster and simpler microscopy

Technical Summary: Phase-Edge Imaging Using q-Plate Shifts for Faster and Simpler Microscopy

Problem Statement
Microscopy techniques for transparent specimens often struggle to distinguish between phase objects (which alter the phase of light) and amplitude objects (which alter intensity). While methods like Zernike phase contrast and Differential Interference Contrast (DIC) exist, they have limitations: Zernike lacks direct edge enhancement, and DIC detects edges only in a single direction. Spiral phase contrast microscopy, utilizing vectorial vortex filters like q-plates, offers all-directional edge enhancement for both phase and amplitude objects. However, standard vectorial filtering enhances both types of edges equally, making them difficult to distinguish. Previous work demonstrated that isolating phase-object edges requires combining images from four distinct off-axis shifts of a q-plate. While effective, this four-shift approach limits imaging speed and increases the risk of errors due to the complexity of multiple mechanical shifts.

Methodology
The authors propose a simplified method to isolate phase edges using a $4f$ imaging system equipped with an off-axis q-plate ($q=1/2$). The core innovation lies in reducing the required measurements from four shifts to just two.

• Optical Setup: The system uses horizontally polarized light to illuminate a sample containing both amplitude ($A$) and phase ($B$) components. The light passes through a Fourier lens, interacts with a q-plate shifted along the $x$-axis by $\pm k_0$, and undergoes an inverse Fourier transform.
• Measurement Protocol: Instead of four shifts, the method employs two shifts ($+k_0$ and $-k_0$ along the $x$-direction). For each shift, the output field is projected through a linear polarizer oriented at either $45^\circ$ or $135^\circ$.
• Signal Processing: The intensities recorded by the CMOS camera for the four combinations ($I^+_{45}, I^-_{45}, I^+_{135}, I^-_{135}$) are processed by calculating the differences $\Delta_{45} = I^+_{45} - I^-_{45}$ and $\Delta_{135} = I^+_{135} - I^-_{135}$. The final isolated phase edge image is derived from the combination $I_{PO} = (\Delta_{45})^2 + (\Delta_{135})^2$.
• Theoretical Basis: The authors derive that this specific combination yields an intensity proportional to $8k_0^2 A(\mathbf{r})^4 |\nabla B(\mathbf{r})|^2$. Crucially, this formulation recovers both gradient components ($\partial B/\partial x$ and $\partial B/\partial y$) without requiring $y$-shifts, while mathematically eliminating contributions from amplitude-object edges.

Key Contributions

1. Reduction in Measurement Complexity: The method halves the number of required measurements (from four to two) compared to the prior state-of-the-art, potentially doubling the image acquisition speed.
2. Theoretical Derivation: The paper provides a rigorous theoretical analysis demonstrating that the two-shift method with $45^\circ/135^\circ$ polarizers successfully isolates the phase gradient squared while suppressing amplitude edges.
3. Experimental Validation: The approach was validated on two distinct phase-amplitude object samples using a custom-built microscope setup.

Experimental Results
The study tested the method on two samples:

• Sample 1: A USAF 1951 phase test target surrounded by a circular amplitude aperture (non-overlapping).
  • Amplitude Suppression: The method achieved an amplitude-edge reduction of up to 97.6% at a q-plate displacement of $125\,\mu\text{m}$.
  • Phase Recovery: The correlation coefficient (CC) between the isolated off-axis phase edges and the on-axis reference reached 0.78. The authors note this is lower than the 0.93 reported in the four-shift method, attributing the difference to non-optimal system alignment rather than a fundamental flaw in the technique.
• Sample 2: A phase test target overlapping with a white hair (amplitude object) and surrounded by an aperture.
  • Amplitude Suppression: Amplitude-edge reduction reached 93.9% (at $100\,\mu\text{m}$ displacement).
  • Phase Recovery: The CC peaked at 0.75. The authors observed that while the isolated phase edges became sharper with larger displacements, the correlation with the reference decreased because the reference image itself suffered from diffraction artifacts between the overlapping hair and phase target.
  • Limitation: In overlapping regions, the phase edge was only partially recovered. The authors clarify that full recovery in such cases would require additional post-processing, such as inverse filtering, to remove the $A(\mathbf{r})^4$ factor inherent in the intensity equation.

Significance and Claims
The paper claims that this two-shift method represents a practical step forward for biological imaging applications where the rapid and simple isolation of phase edges is critical. By reducing the measurement count, the method addresses speed and complexity bottlenecks associated with previous multi-shift techniques. The authors maintain a modest stance, acknowledging that while the method effectively reduces amplitude contributions, perfect separation in complex, overlapping scenarios requires further processing (inverse filtering or deep learning) which is identified as future work. The research aims to make phase-edge isolation more accessible and efficient for distinguishing transparent biological structures from amplitude-based artifacts.