The Height Map Segmentation IOD uses an approach similar to the Segmentation IOD for planar segmentation without a Frame of Reference, which specifies segmentation in the imaging plane of a referenced image (the "derivation image") using that image’s pixel spacing. The Height Map Segmentation specifies a single row of "pixels" (height data) aligned to each referenced image plane and pixel matrix. The segmented surface position is represented by the number of (fractional) rows from the top of the pixel matrix of the referenced image frame (in accordance with the DICOM convention of locating a position in an image by rows and columns offset from the top left corner). Since each referenced image frame has a single row of Height Map Segmentation data, a referenced multi-frame volume therefore has a set of Height Map Segmentation rows. If the referenced muti-frame image frames are regularly spaced, the Height Map Segmentation rows may be represented as a 2D plane orthogonal to the referenced image planes. See the description in Section C.8.20.5 in PS3.3 and especially the following figures therein:

As with the Segmentation IOD, the Height Map Segmentation IOD allows a SOP Instance to describe multiple segments, i.e., layer surfaces. Each segment may be associated with one or more frames in the Height Map Segmentation SOP Instance.

Since a segmented surface might not extend across the entire referenced derivation image volume, typical DICOM pixel padding mechanisms are used. A Height Map Segmentation pixel value in the pixel padding range indicates the absence of the surface at the corresponding derivation image location.

Note that Height Map Segmentation does not use the second method defined in the Segmentation IOD for volumetric segmentation within a Frame of Reference, which allows segmentation in the real-world space defined by a Frame of Reference, with segmentation frame position, orientation, and matrix pixel spacing independent of the referenced image characteristics. Such an approach requires support for 3D volumetric reorientation and reconstruction, and is unnecessary for the primary height map use case.