Skip to content

Slice Analysis

The Slice Module is optimized for conventional wide-field microscope data where you have individual coronal rather than a continuous 3D volume.

Workflow Overview

The analysis is driven by the main script: ls_analyze_slice_volume.

The pipeline consists of the following stages: 1. Volume Generation: Load CZI files and create a centered 3D volume. 2. Slice Reordering: Manually reorder and flip slices as needed. 3. Volume Alignment: Initial rigid alignment to the Allen Atlas. 4. Cutting Angle Determination: Visually match the atlas cutting plane to your tissue. 5. Control Point Matching: Manually refine registration by placing landmarks. 6. Registration Refinement: Elastix-based affine and B-spline deformation per slice. 7. Registered Volume Generation: Apply transforms to all channels. 8. Cell Detection: Detect cells in the registered volume. 9. Cell Registration to Atlas: Transform cell coordinates to atlas space.

Configuration Parameters

Parameters are stored in a sliceinfo struct defined at the top of ls_analyze_slice_volume.m.

Parameter Description Example
sliceinfo.mousename Mouse identifier string 'M001'
sliceinfo.channames Channel names {'DAPI', 'Cy3'}
sliceinfo.pxsize Original pixel size (µm) 0.65
sliceinfo.px_process Processing resolution (µm) 1.25
sliceinfo.px_register Registration resolution (µm) 20
sliceinfo.px_atlas Atlas resolution (µm, typically 10) 10
sliceinfo.slicethickness Physical spacing between slices (µm) 40
sliceinfo.Nslices Total number of slices 50
sliceinfo.celldiam Expected cell diameter (µm) 14
sliceinfo.thresuse SNR thresholds [detection, expansion] [0.75, 0.4]
sliceinfo.debug Enable detection debug plots false
sliceinfo.use_gpu Use GPU for processing false
sliceinfo.atlasaplims Atlas AP axis limits [min, max] [200, 400]

1. Volume Generation

generateSliceVolume(sliceinfo, regchan) reads your CZI files and produces a centered 3D volume:

  • Reads multi-scene CZI files using BioformatsImage.
  • Applies a median filter to remove salt-and-pepper noise.
  • Resamples to the processing resolution (px_process).
  • Crops or pads each slice to a uniform size based on detected brain boundaries.
  • Optionally denoises the DAPI channel.
  • Saves a downsampled RGB preview (volume_for_ordering.tiff) for use in the next step.

2. Slice Reordering

Because slices may be loaded out of order or incorrectly oriented, SliceOrderEditor provides a GUI to review and correct the stack before proceeding.

Controls

Key / Action Function
← / → Navigate to previous / next slice
f Flip current slice horizontally
o Toggle slice as excluded (removes from final stack)
Enter Move current slice to a new position (prompts for index)
s Save ordering decisions
Escape Close the GUI

Output: volume_for_ordering_processing_decisions.txt — a table recording the new order, flip state, and exclusion flags for each slice.

3. Volume Alignment

alignSliceVolume(slicevol, sliceinfo) performs an initial rigid alignment of the slice stack to the Allen Brain Atlas:

  • Extracts a point cloud from each slice using difference-from-background (DFF) images.
  • Loads the Allen template and annotations at the registration resolution.
  • Runs a rigid → affine registration (alignAtlasToSample, refineSampleFromAtlas).

Output: regopts.mat — registration parameters including the rigid transform matrix (tformrigid_allen_to_samp_20um) and axis permutation.

4. Cutting Angle Determination

determineCuttingAngleGUI(opts) lets you visually match the 3D atlas cutting plane to the angle at which your tissue was sectioned.

Controls

Key / Action Function
← / → Navigate slices
Shift + ← / → / ↑ / ↓ Rotate the 3D atlas view (0.3° per press)
Scroll Wheel Move the atlas slice plane perpendicular to the view
Spacebar Toggle atlas boundary overlay
1 / 2 / 3 / 0 Toggle individual channels (0 = all)
Return Save the atlas position for the current slice
c Clear the saved position for the current slice

Output: cutting_angle_data.mat — camera direction vectors and atlas-space points defining the cutting plane for each slice.

5. Control Point Matching

matchControlPointsInSlices(opts) opens a GUI to manually refine the 2D registration of each slice by placing corresponding anatomical landmarks.

Interface

  • Left panel: Your histology slice.
  • Right panel: The corresponding atlas plane.

Controls

Key / Action Function
Click (Left panel) Place a control point on the histology image
Click (Right panel) Place the corresponding control point on the atlas
← / → Navigate to previous / next slice
Spacebar Toggle atlas overlay
g Toggle grid overlay
c Clear control points for the current slice
1 / 2 / 3 / 0 Toggle channels
Enter Jump to a specific slice number
s Save control points

Tips: * At least 3 matched point pairs are required per slice for an affine fit. * Spread points across the full extent of the slice — don't cluster them in one region. * The MSE displayed updates live; lower is better.

Output: atlas2histology_tform.mat — per-slice affine transforms and control point arrays.

6. Registration Refinement

registerSlicesToAtlas(opts) applies a two-stage registration per slice:

  1. Affine stage: Uses your control points (if ≥ 5 are available) or falls back to image-based affine fitting.
  2. B-spline stage: Runs an Elastix deformable registration to capture local deformations not covered by the affine.

Both forward (atlas → sample) and reverse (sample → atlas) transforms are saved, as the reverse is needed for mapping cell coordinates.

Outputs: * transform_params.mat — complete per-slice registration parameters. * elastix_forward/NNN_slice_bspline_atlas_to_samp_20um.txt * elastix_reverse/NNN_slice_bspline_samp_to_atlas_20um.txt

7. Registered Volume Generation

generateRegisteredSliceVolume(sliceinfo, transformparams) maps all channels into atlas space:

  • For each slice, applies the inverse B-spline deformation followed by the inverse affine transform.
  • Places each transformed slice at its corresponding atlas Z position.
  • Fills gaps between slices via nearest-neighbor interpolation.
  • Applies the global rigid transform to place the volume in Allen atlas coordinates.

Output: volume_registered — the full 3D volume in atlas space for all channels.

8. Cell Detection

extractCellsFromSliceVolume(opts, ichan) detects cells slice by slice using a 2D SNR-based approach.

Algorithm

  1. Background subtraction: Computes a local background using a median filter and generates a difference-from-background (DFF) image.
  2. Bandpass filtering: Enhances objects matching the expected cell size while suppressing noise and background.
  3. Local maxima detection: Identifies candidate cells as local maxima above thresuse(1).
  4. Morphological filtering: Removes candidates that are too elongated (circularity < 0.7), too small, or too dim (below thresuse(2)).

Key Parameters

  • celldiam: Expected cell diameter in µm. This directly controls the bandpass filter kernel size — it is the most important detection parameter.
  • thresuse: Two-element SNR threshold vector.
  • thresuse(1) — initial detection threshold. Lower values detect dimmer cells but increase false positives.
  • thresuse(2) — expansion threshold used for cell boundary growth and minimum intensity filtering.

Output: cell_locations_sample.mat — an N×5 array with columns [x, y, slice_index, diameter, mean_intensity].

If debug = true, PNG overlays are saved to cell_detections/ showing the DFF image and detected cell masks per slice.

9. Cell Registration to Atlas

slicePointsToAtlas(inputpts, transformparams) transforms detected cell coordinates from slice space into Allen Atlas space via a chain of transforms:

  1. B-spline deformation field (elastix transformix)
  2. Inverse affine per slice
  3. Global inverse rigid transform

Output: cell_locations_atlas.mat — an N×3+ array with columns [ML, AP, DV] in atlas voxels, plus the original diameter and intensity properties.

Output File Summary

<save_folder>/
├── volume_centered                    # Full-resolution centered slice volume
├── volume_for_ordering.tiff           # RGB preview for the ordering GUI
├── volume_for_ordering_processing_decisions.txt
├── volume_ordered.tiff                # Reordered/flipped volume
├── sample_register_20um.tif           # Volume at registration resolution
├── volume_for_inspection.tiff         # Downsampled RGB for inspection
├── regopts.mat                        # Initial alignment parameters
├── sliceinfo.mat                      # Slice metadata
├── cutting_angle_data.mat             # Cutting plane angles/vectors
├── atlas2histology_tform.mat          # Control point registration data
├── transform_params.mat               # Final per-slice registration parameters
├── elastix_forward/                   # Forward B-spline transforms
│   └── NNN_slice_bspline_atlas_to_samp_20um.txt
├── elastix_reverse/                   # Reverse B-spline transforms (for cell mapping)
│   └── NNN_slice_bspline_samp_to_atlas_20um.txt
├── volume_registered                  # Final atlas-space volume (all channels)
├── cell_locations_sample.mat          # Detected cells in sample space
├── cell_locations_atlas.mat           # Detected cells in atlas space
└── cell_detections/                   # Debug visualizations (if debug=true)
    └── slice_NNN_detections.png