Skip to main content
Curved Planar Reconstruction (CPR) is a visualization technique that allows you to “straighten out” curved anatomical structures like blood vessels, airways, or the spine. This makes it easier to view and measure these structures along their entire length without the distortion caused by their natural curvature. RedBrick AI supports CPR views by allowing you to upload centerline data alongside your medical images. This page explains how to format and upload centerline data for CPR visualization.

​
Centerline Data Structure

Centerlines define the path and orientation for CPR reconstruction. Each centerline consists of position coordinates and orientation matrices that describe how to unfold the curved structure.

​
Basic Structure

{
  "name": "Aorta",
  "seriesIndex": 0,
  "centerline": {
    "position": [...],
    "orientation": [...]
  }
}
Properties:
  • name (string) - Identifier for the centerline (e.g., “Aorta”, “Spine”, “Pulmonary Artery”)
  • seriesIndex (number) - Which image series this centerline belongs to (0-based index)
  • centerline (object) - Contains the position and orientation data

​
Position Data

The position array contains 3D coordinates that define the centerline path through the anatomical structure. Format: Flattened array of coordinates [x1, y1, z1, x2, y2, z2, x3, y3, z3, ...]
Coordinates must be in the same world/model space as your image volume. There is no automatic unit conversion, so ensure your centerline coordinates match the coordinate system of your medical images (typically millimeters for medical imaging).
Example: 
{
  "position": [
    45.2, 123.8, 67.1, 46.1, 124.2, 68.3, 47.0, 124.8, 69.5, 47.9, 125.4, 70.7
  ]
}
This example defines 4 points along the centerline path.

​
Orientation Data

The orientation array contains 4x4 transformation matrices that define the local coordinate system at each point along the centerline. Format: Flattened array of 4x4 matrices [m11, m12, m13, m14, m21, m22, ..., m44] for each point
The number of orientation matrices should match the number of position points. If you have N position points (length = 3N), you should have N orientation matrices (length = 16N).
Example: 
{
  "orientation": [
    1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1,

    1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1
  ]
}
This example provides identity matrices for 2 points (2 × 16 = 32 values).

​
Uploading Centerline Data

The RedBrick viewer requires centerline data to be provided as inline JSON objects with position and orientation arrays. You must include the complete centerline data directly in your task JSON.

​
Inline Centerline Data (Required Format)

Include the centerline data directly in your task JSON with the complete position and orientation arrays: 
[
  {
    "name": "CPR Task - Vessel Analysis",
    "items": ["https://example.com/ct_scan.nii.gz"],
    "centerline": [
      {
        "name": "Vessel",
        "seriesIndex": 0,
        "centerline": {
          "position": [45.2, 123.8, 67.1, 46.1, 124.2, 68.3],
          "orientation": [
            1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 1, 0, 0, 0, 0, 1, 0,
            0, 0, 0, 1, 0, 0, 0, 0, 1
          ]
        }
      }
    ]
  }
]
You can include multiple centerlines per task, each with a unique name. This is useful when visualizing multiple anatomical structures in the same scan.
Example with multiple centerlines: 
[
  {
    "name": "CPR Task - Multi-structure Analysis",
    "items": ["https://datasets.redbrickai.com/scan.nii.gz"],
    "centerline": [
      {
        "name": "Aorta",
        "seriesIndex": 0,
        "centerline": {
          "position": [45.2, 123.8, 67.1, 46.1, 124.2, 68.3, 47.0, 124.8, 69.5],
          "orientation": [
            1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 1, 0, 0, 0, 0, 1, 0,
            0, 0, 0, 1, 0, 0, 0, 0, 1, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0,
            0, 1
          ]
        }
      },
      {
        "name": "Spine",
        "seriesIndex": 0,
        "centerline": {
          "position": [12.5, 98.3, 45.7, 13.1, 99.2, 46.8],
          "orientation": [
            1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 1, 0, 0, 0, 0, 1, 0,
            0, 0, 0, 1, 0, 0, 0, 0, 1
          ]
        }
      }
    ]
  }
]
The RedBrick viewer expects the centerline field to contain a parsed JSON object with position and orientation arrays. URL strings in the centerline field are not automatically fetched or resolved by the viewer.

​
Uploading via CLI

Once you’ve prepared your task JSON file with centerline data, upload it using the RedBrick CLI. Step 1: Save your task JSON to a file (e.g., cpr_tasks.json) Step 2: Run the upload command: 
redbrick upload \
  --org-id YOUR_ORG_ID \
  --project-id YOUR_PROJECT_ID \
  --items-list cpr_tasks.json
Make sure you have the RedBrick SDK installed and configured. See Installation and API Keys for setup instructions.

​
Important Considerations

​
Series Index

The seriesIndex property indicates which image series in your task the centerline belongs to.
Always explicitly set seriesIndex in your task JSON. While the viewer may default missing values to 0 during study setup, uploads via API/SDK require this field to be specified.
  • Use 0 for the first (or only) series in your task
  • Use 1, 2, etc. for subsequent series if you have multiple series per task
  • Multiple centerlines can reference the same seriesIndex
Example with multiple series: 
{
  "name": "Multi-series Task",
  "items": [
    "https://example.com/ct_baseline.nii.gz",
    "https://example.com/ct_followup.nii.gz"
  ],
  "centerline": [
    {
      "name": "Baseline Aorta",
      "seriesIndex": 0,
      "centerline": {
        "position": [45.2, 123.8, 67.1, 46.1, 124.2, 68.3],
        "orientation": [
          1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 1, 0, 0, 0, 0, 1, 0,
          0, 0, 0, 1, 0, 0, 0, 0, 1
        ]
      }
    },
    {
      "name": "Followup Aorta",
      "seriesIndex": 1,
      "centerline": {
        "position": [44.8, 122.5, 66.9, 45.7, 123.1, 68.1],
        "orientation": [
          1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 1, 0, 0, 0, 0, 1, 0,
          0, 0, 0, 1, 0, 0, 0, 0, 1
        ]
      }
    }
  ]
}

​
Multiple Centerlines

You can include multiple centerlines for a single series, and the viewer will make them all available for CPR visualization:
  • Each centerline must have a unique name for identification
  • Common examples include: “Aorta”, “Pulmonary Artery”, “Spine”, “Airways”, “LAD”, “RCA”
  • Users can select which centerline to visualize in the CPR view

​
Data Requirements

To ensure proper CPR rendering, your centerline data should follow these guidelines:

​
Position Array

The position array length should be divisible by 3 to represent complete 3D points: Valid: [1, 2, 3, 4, 5, 6] (2 complete points) Invalid: [1, 2, 3, 4, 5] (incomplete point - missing z coordinate)
Malformed position arrays may result in incorrect CPR rendering or missing visualizations.

​
Orientation Array

The orientation array length should be divisible by 16 to represent complete 4x4 matrices: Valid: [1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1] (1 complete matrix) Invalid: [1, 0, 0, 0, 0, 1, 0, 0] (incomplete matrix - only 8 values)

​
Matching Counts

The number of orientation matrices should match the number of position points for proper CPR reconstruction.
If position length = 3N (N points), orientation length should = 16N (N matrices). Mismatches may cause rendering issues.
Examples:
  • 5 position points → position length = 15, orientation length should = 80
  • 10 position points → position length = 30, orientation length should = 160

​
Numeric Values

  • All values should be valid numbers (avoid NaN or Infinity)
  • Position coordinates must match the world/model space of your image volume
  • Orientation matrices should be valid 4x4 transformation matrices

​
Troubleshooting

​
Common Issues and Solutions

​
CPR view not rendering

If the CPR view doesn’t appear or renders incorrectly, check the following: Centerline format:
  • Ensure the centerline field contains a JSON object with position and orientation arrays, not a URL string
  • Verify arrays are properly formatted (not empty or malformed)
// ❌ Won't render - URL string instead of object
{
  "name": "Vessel",
  "seriesIndex": 0,
  "centerline": "https://example.com/centerline.json"
}

// ✅ Correct - inline object with arrays
{
  "name": "Vessel",
  "seriesIndex": 0,
  "centerline": {
    "position": [1, 2, 3, 4, 5, 6],
    "orientation": [
      1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1,
      1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1
    ]
  }
}

​
Incomplete or malformed position data

Issue: CPR view renders incorrectly or shows artifacts Solution: Verify position array has complete 3D coordinates: 
// ❌ Invalid - incomplete point (5 values, missing z for 2nd point)
"position": [1, 2, 3, 4, 5]

// ✅ Valid - complete points (6 values = 2 points)
"position": [1, 2, 3, 4, 5, 6]

​
Mismatched array lengths

Issue: CPR reconstruction appears distorted or incomplete Solution: Ensure you have one 4x4 orientation matrix (16 values) for each position point: 
// ❌ Invalid - 2 position points but only 1 orientation matrix
{
  "position": [1, 2, 3, 4, 5, 6],
  "orientation": [1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1]
}

// ✅ Valid - 2 position points with 2 orientation matrices
{
  "position": [1, 2, 3, 4, 5, 6],
  "orientation": [
    1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1,
    1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1
  ]
}

​
Series index issues

Issue: CPR view doesn’t display for the correct image series Solution: Verify seriesIndex corresponds to an existing image series: 
// ❌ Invalid - only 1 series (index 0), but referencing index 1
{
  "items": ["scan.nii.gz"],
  "centerline": [{
    "seriesIndex": 1,
    "centerline": { ... }
  }]
}

// ✅ Valid - seriesIndex matches available series
{
  "items": ["scan.nii.gz"],
  "centerline": [{
    "seriesIndex": 0,
    "centerline": { ... }
  }]
}

​
Coordinate space mismatch

Issue: CPR view shows the wrong anatomical region or appears offset Solution: Ensure centerline coordinates are in the same world/model space as your image volume. Check that:
  • Units match (typically millimeters for medical imaging)
  • Origin and coordinate system match your image data
  • There’s no mismatch between RAS/LPS coordinate conventions

​
Need Help?

If you encounter issues not covered here or need assistance with CPR view setup, please reach out to support@redbrickai.com.
⌘I