Coordinate systems

Introduction

One of the issues while dealing with medical images and applications are the differences between the coordinate systems. There are three coordinate systems commonly used in imaging applications: a difference can be made between the world, anatomical and the image coordinate system.

The following figure illustrates the three spaces and their corresponding axes:

Each coordinate system serves one purpose and represents its data in a particular way.

Anatomy image based on an image shared by the My MS organization.

Note that Chand John of Stanford created a detailed presentation about the way coordinates are handled in Slicer.

World coordinate system

The world coordinate system is typically a Cartesian coordinate system in which a model (e.g. a MRI scanner or a patient) is positioned. Every model has its own coordinate system but there is only one world coordinate system to define the position and orientation of each model.

Anatomical coordinate system

The most important model coordinate system for medical imaging techniques is the anatomical space (also called patient coordinate system). This space consists of three planes to describe the standard anatomical position of a human:

• the axial plane is parallel to the ground and separates the head (Superior) from the feet (Inferior).

• the coronal plane is perpendicular to the ground and separates the front (Anterior) from the back (Posterior).

• the sagittal plane is perpendicular to the ground and separates the Left from the Right.

From these planes it follows that all axes have their notation in a positive direction (e.g. the negative Superior axis is represented by the Inferior axis).

The anatomical coordinate system is a continuous three-dimensional space in which an image has been sampled. In neuroimaging, it is common to define this space with respect to the human whose brain is being scanned. Hence, the 3D basis is defined along the anatomical axes of anterior-posterior, inferior-superior, and left-right.

However different medical applications use different definitions of this 3D basis. Most common are the following bases:

• LPS (Left, Posterior, Superior) is used in DICOM images

\[\begin{split} LPS = \begin{Bmatrix} \text{from right towards left} \\ \text{from anterior towards posterior} \\ \text{from inferior towards superior} \end{Bmatrix} \end{split}\]

• RAS (Right, Anterior, Superior) is similar to LPS with the first two axes flipped

\[\begin{split} RAS = \begin{Bmatrix} \text{from left towards right} \\ \text{from posterior towards anterior} \\ \text{from inferior towards superior} \end{Bmatrix} \end{split}\]

Thus, the only difference between the two conventions is that the sign of the first two coordinates is inverted.

Both bases are equally useful and logical. It is just necessary to know to which basis an image is referenced.

Image coordinate system

The image coordinate system describes how an image was acquired with respect to the anatomy. Medical scanners create regular, rectangular arrays of points and cells which start at the upper left corner. The \(i\) axis increases to the right, the \(j\) axis to the bottom and the \(k\) axis backwards.

In addition to the intensity value of each voxel \((i j k)\) the origin and spacing of the anatomical coordinates are stored too.

• The origin represents the position of the first voxel (0, 0, 0) in the anatomical coordinate system, e.g. (100, 50, -25) mm.

• The spacing specifies the distance between voxels along each axis, e.g. (1.5, 0.5, 0.5) mm.

The following 2D example shows the meaning of origin and spacing:

Using the origin and spacing, the corresponding position of each (image coordinate) voxel in anatomical coordinates can be calculated.

Image transformation

The transformation from an image space vector \((i j k)'\) to an anatomical space vector \(\vec{x}\) is an affine transformation, consists of a linear transformation \(\mathbf{A}\) followed by a translation \(\vec{t}\).

\[ \vec{x} = A \begin{pmatrix} i & j & k \end{pmatrix}' + \vec{t} \]

The transformation matrix \(\mathbf{A}\) is a \(3 \times 3\) matrix and carries all information about space directions and axis scaling.

\(\vec{t}\) is a \(3 \times 1\) vector and contains information about the geometric position of the first voxel.

\[\begin{split} \begin{pmatrix} x_1 \\ x_2 \\ x_3 \end{pmatrix} = \begin{pmatrix} A_{11} & A_{12} & A_{13} \\ A_{21} & A_{22} & A_{23} \\ A_{31} & A_{32} & A_{33} \end{pmatrix} \begin{pmatrix} i \\ j \\ k \end{pmatrix} + \begin{pmatrix} t_1 \\ t_2 \\ t_3 \end{pmatrix} \end{split}\]

The last equation shows that the linear transformation is performed by a matrix multiplication and the translation by a vector addition. To represent both, the transformation and the translation, by a matrix multiplication an augmented matrix must be used. This technique requires that the matrix \(\mathbf{A}\) is augmented with an extra row of zeros at the bottom, an extra column-the translation vector-to the right, and a \(1\) in the lower right corner. Additionally, all vectors have to be written as homogeneous coordinates, which means that a \(1\) is augmented at the end.

\[\begin{split} \begin{pmatrix} x_1 \\ x_2 \\ x_3 \\ 1 \end{pmatrix} = \begin{pmatrix} A_{11} & A_{12} & A_{13} & t_1 \\ A_{21} & A_{22} & A_{23} & t_2 \\ A_{31} & A_{32} & A_{33} & t_3 \\ 0 & 0 & 0 & 1 \end{pmatrix} \begin{pmatrix} i \\ j \\ k \\ 1 \end{pmatrix} \end{split}\]

Depending on the used anatomical space (LPS or RAS) the \(4 \times 4\) matrix is called IJKtoLPS- or IJKtoRAS-matrix, because it represents the transformation from IJK to LPS or RAS.

2D example or calculating an IJtoLS-matrix

The following figure shows the anatomical space with an L(P)S basis on the left and the corresponding image coordinates on the right.

The origin (the coordinates of the first pixel in anatomical space) is (50, 300) mm and the spacing (the distance between two pixels) is (50, 50) mm.

As this is a 2D example \(\mathbf{A}\) is a \(2 \times 2\) matrix and \(\vec{t}\) a \(2 \times 1\) vector. Therefore, the equation of the affine transformation is:

\[\begin{split} \begin{pmatrix} L \\ S \\ 1 \end{pmatrix} = \begin{pmatrix} A_{11} & A_{12} & t_1 \\ A_{21} & A_{22} & t_2 \\ 0 & 0 & 1 \end{pmatrix} \begin{pmatrix} i \\ j \\ 1 \end{pmatrix} \end{split}\]

By multiplying the IJtoLS-matrix and the vector of the right side, the following product will be obtained:

The last equation and the matrix product show that a total of 6 unknown variables \((A_{11}, A_{12}, A_{21}, A_{22}, t_1, t_2)\) have to be determined. The knowledge of origin and spacing however allows the following relations between image and anatomical space:

\[\begin{split} \begin{pmatrix} L \\ S \end{pmatrix} \equiv \begin{pmatrix} i \\ j \end{pmatrix} \qquad \begin{pmatrix} 50 \\ 300 \end{pmatrix} \equiv \begin{pmatrix} 0 \\ 0 \end{pmatrix} \qquad \begin{pmatrix} 100 \\ 300 \end{pmatrix} \equiv \begin{pmatrix} 1 \\ 0 \end{pmatrix} \qquad \begin{pmatrix} 50 \\ 250 \end{pmatrix} \equiv \begin{pmatrix} 0 \\ 1 \end{pmatrix} \qquad \dots \end{split}\]

Thus, at least six equations can be derived:

\[\begin{split} \begin{align*} 50 &= A_{11} \cdot 0 + A_{12} \cdot 0 + t_1 \cdot 1 \\ 300 &= A_{21} \cdot 0 + A_{22} \cdot 0 + t_2 \cdot 1 \\ 100 &= A_{11} \cdot 1 + A_{12} \cdot 0 + t_1 \cdot 1 \\ 300 &= A_{21} \cdot 1 + A_{22} \cdot 0 + t_2 \cdot 1 \\ 50 &= A_{11} \cdot 0 + A_{12} \cdot 1 + t_1 \cdot 1 \\ 250 &= A_{21} \cdot 0 + A_{22} \cdot 1 + t_2 \cdot 1 \end{align*} \end{split}\]

As mentioned above, the translation \(\vec{t}\) contains the information about the geometric position of the first pixel and is therefore equivalent to the origin. This result is also confirmed by the first equations.

The solution of the other equations leads to the following IJtoLS-matrix:

\[\begin{split} IJtoLS = \begin{pmatrix} 50 & 0 & 50 \\ 0 & -50 & 300 \\ 0 & 0 & 1 \end{pmatrix} \end{split}\]

In the event that a R(A)S basis was used, just the left and anterior axis of the anatomical space are flipped, and the image coordinate system appears in the same way as in the L(P)S case.

For this 2D example the IJtoRS-matrix would be:

\[\begin{split} IJtoRS = \begin{pmatrix} -50 & 0 & 250 \\ 0 & -50 & 300 \\ 0 & 0 & 1 \end{pmatrix} \end{split}\]

This matrix looks very similar to the IJtoLS-matrix with 2 differences:

• The translation \(\vec{t}\) has changed because of another origin.

• The right axis is flipped, so the first column of the IJtoRS-matrix has just an inverted sign.

Coordinate system convention in Slicer

DICOM and most medical imaging software use the LPS coordinate system for storing all data. The choice of origin is arbitrary because only relative differences have meaning, so there is no universal standard, but it is often set to some geometric center of the imaging system, or it is chosen to be near the center of an object of interest.

Both LPS and RAS were in wide use in the early 2000s when development of Slicer was started and Slicer developers chose to use the RAS coordinate system. Historically scans by GE equipment used RAS while Siemens and others used LPS. Since several GE researchers were early contributors to Slicer, RAS was adopted for the internal representation.

Slicer still uses RAS coordinate system for storing coordinate values internally for all data types, but for compatibility with other software, it assumes that all data in files are stored in LPS coordinate system (unless the coordinate system in the file is explicitly stated to be RAS). To achieve this, whenever Slicer reads or writes a file, it may need to flip the sign of the first two coordinate axes to convert the data to RAS coordinate system.

Relations to other software/conventions

ITK

ITK uses the LPS convention.

Using MATLAB to map Slicer RAS coordinates (e.g. fiducials) to voxel space of a NIfTI Image

To extract the “voxel to world” transformation matrix from a NIFTI file’s header (entry: qto_xyz:1-4) in MATLAB:

d = inv(M) * [ R A S 1 ]'

where M is the matrix and R A S are coordinates in Slicer, then d gives a vector of voxel coordinates.

(Solution courtesy of András Jakab, University of Debrecen)

References

• https://people.cs.uchicago.edu/~glk/unlinked/nrrd-iomf.pdf

• http://www.grahamwideman.com/gw/brain/orientation/orientterms.htm

• https://nifti.nimh.nih.gov/nifti-1/documentation/faq

• https://teem.sourceforge.net/nrrd/format.html

• DICOM 2013 PS3.3 Image Position and Image Orientation