Landmarks notes

[weiglszonja]

widefieldLandmarks.dorsalCortex.json

Collection: alf/widefield
Description: User selected landmarks and transform for widefield imaging to allow images to be transformed and aligned to Allen Atlas

IBL notebooks:

• https://github.com/jcouto/wfield/blob/master/notebooks/match_dorsal_cortex_to_the_allen.ipynb
• https://github.com/jcouto/wfield/blob/master/notebooks/tutorial_example_warping_stack_ROIs.ipynb

Example NWB schema

groups:

  - neurodata_type_def: Landmarks
    neurodata_type_inc: DynamicTable
    doc: "2D spatial landmarks for brain regions or image landmarks."
    datasets:
      - name: x
        neurodata_type_inc: VectorData
        dtype: float
        doc: "X coordinate of each landmark."

      - name: y
        neurodata_type_inc: VectorData
        dtype: float
        doc: "Y coordinate of each landmark."

      - name: name
        neurodata_type_inc: VectorData
        dtype: text
        doc: "Name of the landmark."

      - name: color
        neurodata_type_inc: VectorData
        dtype: text
        doc: "Color of the landmark (hex code)."


  - neurodata_type_def: LandmarksMetaData
    neurodata_type_inc: LabMetaData
    doc: "Metadata storing multiple landmark sets and their spatial transform."

    attributes:
      - name: bregma_offset
        dtype: float
        shape: [2]
        doc: "Offset from bregma in reference coordinates."

      - name: resolution
        dtype: float
        doc: "Pixel resolution."

    datasets:
      - name: transform
        dtype: float
        shape: [3, 3]
        doc: "Similarity transform matrix between coordinate systems."

    groups:
      - name: landmarks
        neurodata_type_inc: Landmarks
        doc: "Landmarks in reference coordinate system."

      - name: landmarks_im
        neurodata_type_inc: Landmarks
        doc: "Landmarks in image coordinate system."

      - name: landmarks_match
        neurodata_type_inc: Landmarks
        doc: "Matched landmarks used for transform fitting."

The schema diagram would look like:

LandmarksMetaData (LabMetaData)
│
├── attributes
│     • bregma_offset: float[2]
│     • resolution: float
│     • transform: float[3,3]
│
├── landmarks (Landmarks: DynamicTable)
│      ├── x : float
│      ├── y : float
│      ├── name : text
│      └── color : text
│
├── landmarks_im (Landmarks: DynamicTable)
│      ├── x : float
│      ├── y : float
│      ├── name : text
│      └── color : text
│
└── landmarks_match (Landmarks: DynamicTable)
       ├── x : float
       ├── y : float
       ├── name : text
       └── color : text

Example NWB output

/processing/ophys/LandmarksMetaData
    bregma_offset: [320, 270]
    resolution: 0.0194
    transform: 3x3 float matrix

    /landmarks            (DynamicTable)
    /landmarks_im         (DynamicTable)
    /landmarks_match      (DynamicTable)

[weiglszonja]

Summarising the discussion:

Create an extension with a generic name ndx-rigid-transformation:

• mention that it applies to the ibl transformation
• make it compatible for specific for ibl having both in the same extension
• It will have two objects: image transformation matrix, and the landmarks

NWB Extension for image transformation object

• for rigid transformation
• storing the rigid transformation object: translation, scale and rotation -> this is a 3x3 matrix
• this transformation object can be linking to an imaging plane or image
• since this can be useful for other cases this could be an extension on its own

groups:

• neurodata_type_def: Landmarks
  neurodata_type_inc: DynamicTable
  doc: "2D spatial landmarks for brain regions or image landmarks."
  datasets:

  • name: x
    neurodata_type_inc: VectorData
    dtype: float
    doc: "X coordinate of each landmark."

  • name: y
    neurodata_type_inc: VectorData
    dtype: float
    doc: "Y coordinate of each landmark."

  • name: registered_x
    neurodata_type_inc: VectorData
    dtype: text
    doc: "Name of the landmark."

  • name: registered_y
    neurodata_type_inc: VectorData
    dtype: text
    doc: "Name of the landmark."

  • name: rigid_transformation
    neurodata_type_inc: VectorData from the ndx rigid ransformation

  optional columns

  • name: name

  • name: color
    neurodata_type_inc: VectorData
    dtype: text
    doc: "Color of the landmark (hex code)."

[alessandratrapani]

Initial proposal --> https://github.com/catalystneuro/ndx-spatial-transformation