User:Millerjv/MultiparametricTimeSeriesAndLongitudinalConcepts

From NAMIC Wiki
< User:Millerjv
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search

Concepts of multi parametric, timseries, and longitudinal images/volumes

In ITK, the image class is templated over pixel type (short, float, vector, RGB) and dimension (2,3,4,...). This design provides an explicit distinction between the independent variables (specified by dimension) and the dependent variables (specified by the pixel type). While this design can meet many needs, there are a few concepts that are not captured by this design. The intent, here, is to collect the concepts and formalize a vernacular to describe these concepts, independent of the implementation design.


Image, Volume
generic terms to refer to images of various dimensions and pixel types.
VectorVolume
generic term to refer to volumes where multiple values are stored at a pixel. The values at a pixel may form a vector space or may just be multiple physical properties measured at each pixel.

SignalVolume
volume where each pixel represents a timecourse or a signal. Implicit to this concept is the volume is physically stationary over time and the sampling rate of the physical property is much higher than the sampling rate individual SignalVolumes. From a memory indexing view, time is the fastest moving index. Example: a 3D volume where each pixel is encodes signal intensity over time.

VolumeStack
a collection of volumes where each volume measures the same physical property. VolumeStacks may or may not span the same physical space but each volume in the stack should be the same size and shape (number of pixels in each dimension).
VolumeTimeStack
a collection of volumes where each volume measure the same physical property over the same physical location. Like a VolumeStack, each volume should be the same size and shape.
VolumeCollection
a collection of volumes that may represent different physical properties, span different physical spaces, have different shapes, and be associated with different time points.


Note: when the concepts align with how the images are acquired or frequently visualized, then the same concepts can be used for processing offline sequences as well as online sequences (ultrasound, fluoro, etc.)