Brain section alignment with fiducials

Background As part of the data ingestion pipeline onto brainsharer.org, sections must be aligned. The general pipeline steps allow for automated alignment post-masking. Unfortunately, some sections may still be out of alignment and require fiducials to properly align specific sections to the rest of the image stack.

After the pipeline has completed (Neuroglancer-compatible image stack created and visible from brainsharer.org), the user should identify which sections are out of alignment, create fiducials and re-rerun alignment. This document describes these steps.

Process and example

  • Create the first aligned image stack in the pipeline:

    python src/pipeline/scripts/create_pipeline.py –task align –animal {ANIMAL_ID}

  • Create the first aligned precomputed data stack in the pipeline:

    python src/pipeline/scripts/create_pipeline.py –task neuroglancer –animal {ANIMAL_ID}

  • In a browser, open up Neuroglancer and create a new view.

  • Add the first layer with the ‘+’ icon. This initial layer will have a source such as: https://imageserv.dk.ucsd.edu/data/{ANIMAL_ID}/neuroglancer_data/C1T_aligned.

  • This will be the layer where you will add the fiducials. This layer has already been aligned, but most brains need fiducials to make the alignment more accurate.

  • Open up the upper left quadrant. You may need to adjust the brightness and contrast to see the image better. See the ‘Adjusting colors’ page for more information.

  • Identify which section(s) are not properly aligned through Neuroglancer interface. The z value is the section number.

  • Navigate to non-aligned section (in figure below this is section 75).

../_images/realignment.1.png

  • Create a new annotation layer (just click CTRL and  the “+” icon next to the “C1T_aligned” layer. That creates a new annotation layer.

  • Click the ‘Annotations’ tab on the top far right, then  click the ‘Annotate cloud’ icon on the far right (in red box)

../_images/realignment.2.png

  • Pick 2 or 3 points on aligned section and save with a Fiducial label.

Adding fiducial label

  • Go to the lower right panel and type ‘Fiducial’ in the text box that says ‘Add annotation labels’. A search result will pop up.

  • Select the ‘Fiducial’ selection

Save current state (the fiducial points must be exported!)

  • Click the ‘new’ (lower-case ‘n’) icon in lower-right corner of screen (see figure below)

  • Note: These are NOT same new/save buttons used for saving entire state! Those have upper-case letters.

  • N.B. If you add more points, make sure to click the ‘save’ button on the bottom far right.

  • Repeat cloud annotations for all non-aligned sections (same number of points)

Re-run alignment

  1. Login into one of the linux workstations

  2. Navigate to /data/pipeline/ (contains all scripts)

    cd /data/pipeline

  3. Active the python virtual enviroment

    source /usr/local/share/pipeline/bin/activate

  4. Run realign step [for channel 1, downsampled]

    python src/pipeline/scripts/create_pipeline.py --task realign --animal {ANIMAL_ID}

  5. Run neuroglancer step [for channel 1, downsampled]

    python src/pipeline/scripts/create_pipeline.py --task neuroglancer --animal {ANIMAL_ID}

  6. The newly aligned data will be available as a new precomputed data stack. Open up a new layer and enter this as a precomputed data source:

    https://imageserv.dk.ucsd.edu/data/{ANIMAL_ID}/neuroglancer_data/C1T_realigned

  7. Check the alignment of the sections with fiducials. If the alignment is still not correct, add more/better fiducials and repeat the process.

Note

The neuroglancer process will use the realigned data stack if it exists, otherwise, it will use the original aligned data stack.