Repository navigation
Method Overview
FRC (2D) and FSC (3D) identify the spatial-frequency range supported by correlated information in two observations. Fluorescence imaging often makes a truly matched second acquisition difficult: sample dynamics, drift, bleaching, or changes during z-stack collection can bias paired measurements. Effective resolution also depends on specimen and reconstruction, not only the nominal objective or PSF specification.
- Partition each non-overlapping 2 × 2 block into two opposite-diagonal pairs.
- Average the two pixels within each pair, producing two structurally similar, complementary half-sampled images.
- Apply identical Hanning windows and calculate their FRC.
- Compensate for the deterministic frequency-dependent correlation bias introduced by diagonal resampling.
- Estimate the resolution from the chosen correlation cutoff, usually the 1/7 hard threshold.
Diagonal resampling preserves useful noise independence under the method's spatial-redundancy and PSF-symmetry assumptions, but halves the effective lateral sampling frequency. The manuscript describes a stricter sampling requirement for direct sFRC on raw data: approximately four times the highest spatial frequency of interest. Curve extrapolation can help in mildly undersampled cases, but cannot recover information that was never measured.
The complementary XY sampling is repeated separately on each z-plane. The paired volumes are then processed with consistent physical frequency coordinates, including appropriate resampling and padding, to calculate directional FSC. A global spherical-shell FSC alone cannot distinguish lateral from axial frequency support.
Direct measurements near the axial Fourier direction can be contaminated by paraxial artifacts associated with unequal voxel spacing or dimensions, padding, and inconsistencies between adjacent optical sections. DRI:
- establishes a lateral frequency cutoff from the data;
- measures directional FSC cutoffs at selected non-paraxial angles;
- fits a low-dimensional smooth directional-support model while keeping the lateral cutoff fixed;
- infers the axial cutoff from the off-axis directional measurements.
Lateral and axial resolution are obtained as the reciprocals of their corresponding cutoff frequencies. The fitting angles are measured from the XY plane toward the z-axis; angle units depend on the software implementation.
Important: E_DRI measures internal fitting consistency, not ground-truth error, a confidence interval, or proof that the inferred axial resolution is accurate.
Rolling FRC evaluates local windows and places each estimate at the window centre. Background masking excludes regions unsuitable for analysis. The resulting map captures heterogeneity that a global sFRC number conceals. A zero-valued map pixel represents background or an unavailable local estimate—not 0 nm resolution. Map statistics may be calculated at the native rolling-grid scale before interpolation to the image grid.
When evaluating computationally reconstructed super-resolution images, simply splitting the final reconstructed image can create shared reconstruction artifacts. The manuscript instead describes generating the complementary observations before reconstruction, restoring their sampling grids as appropriate, reconstructing both independently, and then calculating FRC/FSC on those independent outputs. This workflow is methodologically distinct from direct sFRC of an acquired raw image; it is not a one-switch operation in all three code repositories.
- A valid cutoff requires adequate SNR and spatial sampling. FRC/FSC quantify statistical consistency, not absolute morphological correctness.
- Shared periodic artifacts and reconstruction-induced correlations can produce deceptively optimistic estimates.
- The simple DRI directional-support model may be less suitable for highly distorted or non-smooth anisotropy.
- The manuscript also discusses srFSC, a rolling 3D FSC extension. These three software entry-point guides cover sFRC, sFSC/DRI and srFRC; they do not assert that srFSC is exposed as a separate public command in the current repositories.
This project accompanies the manuscript Single-acquisition estimation of anisotropic 3D resolution in super-resolution fluorescence microscopy. Bibliographic details and DOI should be added after publication. See the repository README for project-specific availability and the shared example data.
srFRCj · Repository · Issues · Documentation
Manuscript: Single-acquisition estimation of anisotropic 3D resolution in super-resolution fluorescence microscopy — publication details pending.
Start here
Analysis guides
Reference