Annotation type definitions
This section covers definitions of all the annotation object types. The objects are the same for importing annotations, and exporting annotations.
Label Definition
2D Image
3D Volume
Video
CLASSIFICATION (non objectType)
SEGMENTATION
LENGTH
ANGLE
POINT
POLYLINE
POLYGON
ELLIPSE
BBOX
CUBOID
class redbrick.types.task.Point2D[source]
2D pixel point.
The Point2D coordinates are normalized in relation to the dimensions of the image, meaning they are scaled down to a range between 0 and 1. To convert these normalized points back to their original scale on the image, you “un-normalize” them by multiplying by the image’s dimensions. This operation typically results in a floating-point number.
RedBrick AI enhances the precision of annotations on its canvas by supporting a resolution higher than that of the underlying image. This capability facilitates sub-pixel annotation accuracy. However, if sub-pixel precision is not necessary for your application, you can simplify the process by rounding the un-normalized coordinates to the nearest whole number. This adjustment converts the precise floating-point values back to standard pixel coordinates, making them easier to work with for general purposes.
Important
The origin point (0,0) from which these coordinates are measured is at the top left of the image. However, for mammography and DBT the origin is the top right of the image.
xNorm: float
X co-ordinate normalized by the width of the image.
yNorm: float
Y co-ordinate normalized by the height of the image.
class redbrick.types.task.VoxelPoint[source]
Represents a three-dimensional point in image-space, where i, j, and k are columns, rows, and k is the slice number.
i: int
j: int
k: int
class redbrick.types.task.WorldPoint[source]
Represents a three-dimensional point in physical space/world co-ordinates. The world co-ordinates are calculated using redbrick.types.task.VoxelPoint and the DICOM Image Plane Module.
x: float
y: float
z: float
class redbrick.types.task.MeasurementStats[source]
Measurement statistics for annotations.
average: float
Average pixel/voxel intensity within the annotation. In CT, this is the average HU value.
area: float
Area contained within the annotation in mm^2.
volume: float
Volume contained within the annotation in mm^3
minimum: float
Minimum intensity value within the annotation.
maximum: float
Maximum intensity value within the annotation.
class redbrick.types.task.VideoMetaData[source]
Contains annotation information along the third axis. Frames for video, and slices for 3D volumes.
Warning
redbrick.types.task.VideoMetaData has a misleading name. It contains information for both videos, and 3D volumes.
Hint
Watch this video for a detailed explaination of all the attributes of this object.
seriesItemIndex: int
The index of the file in series “items” list that this annotation is present on.
seriesFrameIndex: int
Frame index of the annotation for nifti and dicom multipart files.
frameIndex: int
The frame number (for video) or slice index (for 3D volumes) the annotation is present on.
trackId: str
Each distinct object has a unique trackId. Two annotations on different frameIndex’s with the same trackId’s represent the same distinct object.
keyFrame: bool
If True, this annotation is user-defined. If False, this annotation is interpolated.
endTrack: bool
If True, this annotation is the last annotation of a specific track defined by trackId.
redbrick.types.task.Attributes
Attributes for Taxonomy objects. Attributes can be boolean, textfields, select’s or multi-select’s.
alias of Dict[str, Union[str, bool, List[str]]]
class redbrick.types.task.CommonLabelProps[source]
Full version of redbrick.types.task.Series.segmentMap.
category: Union[int, str, List[str]]
Taxonomy object category.
attributes: Dict[str, Union[str, bool, List[str]]]
Associated taxonomy object attributes.
mask: Union[str, List[str]]
Filepath to segmentation file for this annotation.
overlappingGroups: List[int]
Overlapping instances that this label is part of (when without_masks is set).
group: str
Linked label group id.
readOnly: bool
Read only status
class redbrick.types.task.InstanceClassification[source]
Instance level classifications for frame-by-frame (video) or slice-by-slice (volume) classifications..
fileIndex: int
For video this is the frameIndex, for DICOM volumes this is the sliceIndex.
fileName: str
The file name represented by frameIndex.
group: str
Linked label group id.
readOnly: bool
Read only status
class redbrick.types.task.Classification[source]
Study or series classification.
attributes: Dict[str, Union[str, bool, List[str]]]
group: str
Linked label group id.
readOnly: bool
Read only status
class redbrick.types.task.Polyline[source]
Open polylines, not supported in 3D images.
points: List[Point2D]
category: Union[int, str, List[str]]
attributes: Dict[str, Union[str, bool, List[str]]]
video: VideoMetaData
group: str
Linked label group id.
readOnly: bool
Read only status
class redbrick.types.task.Polygon[source]
Closed polygons, not supported in 3D images.
points: List[Point2D]
category: Union[int, str, List[str]]
attributes: Dict[str, Union[str, bool, List[str]]]
stats: MeasurementStats
video: VideoMetaData
group: str
Linked label group id.
readOnly: bool
Read only status
class redbrick.types.task.Cuboid[source]
3D bounding boxes for 3D images.
point1: VoxelPoint
Top left diagonal corner.
point2: VoxelPoint
Bottom right diagonal corner.
absolutePoint1: WorldPoint
absolutePoint2: WorldPoint
category: Union[int, str, List[str]]
attributes: Dict[str, Union[str, bool, List[str]]]
stats: MeasurementStats
group: str
Linked label group id.
readOnly: bool
Read only status
class redbrick.types.task.BoundingBox[source]
2D bounding box for 2D images, or slice by slice annotation in 3D images.
Hint
See the following diagram to understand the coordinate system.
pointTopLeft: Point2D
Coordinates of the top left of the bounding box.
wNorm: float
Width of the bounding box, normalized by image width.
hNorm: float
Height of the bounding box, normalized by the image height.
category: Union[int, str, List[str]]
attributes: Dict[str, Union[str, bool, List[str]]]
stats: MeasurementStats
video: VideoMetaData
group: str
Linked label group id.
readOnly: bool
Read only status
class redbrick.types.task.Ellipse[source]
Ellipse annotation.
Hint
See this ellipse diagram to understand the coordinate system. Please read the redbrick.types.task.Point2D documentation to understand how normalization works & the origin of the coordinate system.
Warning
For DICOM images of a certain type, ellipse annotations might be flipped i.e., rotating the ellipse clockwise would result in counter-clockwise rotation. If you encounter these cases, reach out to our support for instructions on how to handle this support@redbrickai.com.
pointCenter: Point2D
The normalized center of the ellipse.
xRadiusNorm: float
The x axis of the ellipse, normalized with by the image width. Adjusting for rotationRad, the x-axis of the ellipse aligns with the x-axis of the image.
yRadiusNorm: float
The y axis of the ellipse, normalized with by the image width. Adjusting for rotationRad, the y-axis of the ellipse aligns with the y-axis of the image.
rotationRad: float
The rotation of the ellipse measured clockwise as the angle between the y-axis of the ellipse and y-axis of the image.
category: Union[int, str, List[str]]
attributes: Dict[str, Union[str, bool, List[str]]]
stats: MeasurementStats
video: VideoMetaData
group: str
Linked label group id.
readOnly: bool
Read only status
class redbrick.types.task.MeasureAngle[source]
Angle measurement label.
An angle measurement is defined by three points, where vertex is the middle point between point1 and point2. The angle between the two vectors <point1, vertex> and <point2, vertex> defines the angle measurement.
Hint
See this angle diagram to understand the coordinate system.
type: Literal['angle']
point1: VoxelPoint
vertex: VoxelPoint
point2: VoxelPoint
absolutePoint1: WorldPoint
absoluteVertex: WorldPoint
absolutePoint2: WorldPoint
normal: List[float]
Measurements can be made on oblique planes. normal defines the normal unit vector to the slice on which the annotation was made. For annotations made on non-oblique planes, the normal will be [0,0,1]. The measurement is fully defined even without normal, however, for completeness see this angle diagram for it’s definition.
angle: float
Measurement angle in degrees.
category: Union[int, str, List[str]]
attributes: Dict[str, Union[str, bool, List[str]]]
group: str
Linked label group id.
readOnly: bool
Read only status
class redbrick.types.task.MeasureLength[source]
Length measurement label.
A length measurement is defined by two points, and the length measurement is the distance between the two points.
type: Literal['length']
point1: VoxelPoint
point2: VoxelPoint
absolutePoint1: WorldPoint
absolutePoint2: WorldPoint
normal: List[float]
Measurements can be made on oblique planes. normal defines the normal unit vector to the slice on which the annotation was made. For annotations made on non-oblique planes, the normal will be [0,0,1]. The measurement is fully defined even without normal, however, for completeness see this length diagram for it’s definition.
length: float
The value of the measurement in millimeters.
category: Union[int, str, List[str]]
attributes: Dict[str, Union[str, bool, List[str]]]
group: str
Linked label group id.
readOnly: bool
Read only status
class redbrick.types.task.Landmark3D[source]
3D landmark for 3D data.
point: VoxelPoint
category: Union[int, str, List[str]]
attributes: Dict[str, Union[str, bool, List[str]]]
group: str
Linked label group id.
readOnly: bool
Read only status
class redbrick.types.task.Landmark[source]
2D landmark for 2D data.
point: Point2D
category: Union[int, str, List[str]]
attributes: Dict[str, Union[str, bool, List[str]]]
video: VideoMetaData
group: str
Linked label group id.
readOnly: bool
Read only status
NextTaxonomy formatPreviousFormats
Copyright © 2023, RedBrick AI
Made with Sphinx and @pradyunsg’s Furo
On this page
Was this page helpful?