Object ID: 00000018WIA303AA970GYZ
Topic ID: id_40023053 Version: 1.3
Date: Mar 30, 2022 1:31:21 PM

Measurement accuracy

The accuracy of on-view measurements depends on various factors, and, in particular, on the size of the ROI being measured. A typical MR acquisition has a DFOV from about 25 to about 48 cm. In the examples below, 25 mm is used.

Inter-slice distances can vary from less than 1 up to 10 mm. For best results with 3D imaging, the optimum inter-slice distance is the one that results in isotropic voxels (with the same dimensions in all three axes). However, considerations such as type of exam and patient irradiation dose levels may lead to the choice of a larger inter-slice distance. The choice of the optimum inter-slice distance for a given case is outside the scope of this manual.

Note: Regardless of the zoom factor being used to view images, ROI measurements and statistics are calculated based on the pixels from the original, unzoomed image data as they arrived on the workstation.

CAUTION

Measurements are more reliable when done on 2D views. Always check on the 2D reformatted views where exactly the points have been deposited.

CAUTION

Distance, angle, and area measurements are valid only if all trace segments are longer than the inter-slice distance.

Measurement resolution

The software calculates and displays measurements with a resolution of one decimal place (such as 0.1 mm, 0.1 degree, or 0.1 mm²). However, the real measurement accuracy is generally considerably less for a number of different reasons.

Geometrical accuracy

Geometric accuracy is limited by display resolution (pixel size). When four views are displayed, each view equals 512x512 pixels. With a DFOV of 25 cm, a pixel is equivalent to 0.5x0.5 mm, so you cannot place a measurement point with a precision better than this. As a result:

For a distance measurement, the geometrical accuracy is equal to the displayed length +/- image pixel size.
For an angle measurement, the geometrical accuracy is equal to the displayed angle value +/-10 degrees for an angle measured between segments which are five times larger than the image pixel size. Accuracy improves as the length of the segments increases.
For an area measurement, the geometrical accuracy is equal to the displayed area value +/- the circumference of the region of interest multiplied by (image pixel size)2 / 2. Note the region-of-interest measurements and statistics are based on the pixels INSIDE the graphic defining the region.

The geometrical accuracy defines a lower bound on the overall accuracy that can be obtained. Further limiting factors are image set resolution, acquisition accuracy, display settings, and partial volume effects.

Image set resolution

The image set resolution is determined by the size of the FOV, the matrix, and the inter-slice distance.

In the acquisition plane, for a 25 cm FOV, the smallest detail in an image acquired with a 512x512 matrix will be about 0.5x0.5 mm. With a 256x256 matrix, the smallest detail will be 1x1 mm. In the acquisition plane, the measurement accuracy cannot be better than the size of the smallest element.

In the same way, the accuracy in a direction perpendicular to the acquisition plane cannot be better than the inter-slice distance.

CAUTION

Distance, angle, and area measurements are valid only if all trace segments are longer than the inter-slice distance.

Acquisition accuracy

Any errors in the original image set resulting from the acquisition process (calibration, slice interpolation) will be added to the same extent to the measurement error.

As an example, the spatial accuracy of MR images can vary, depending on the patient, the pulse sequence, and the MR system itself. Metallic implants or air-bone interfaces may lead to susceptibility artifacts and spatial distortions greater than those observed when calibrating the system with a Quality Assurance phantom, even on a perfectly tuned MR system.

Display settings

Since anatomical features are rarely of a uniform density, the apparent dimension of an anatomical feature can change when you modify the display settings (window width and level), thereby adding another factor of uncertainty to an on-view measurement.

3D object measurements

You can measure voxel value, distance, angle, area, and total volume on the views. When dealing in 3D, the rules are a bit more complex. For instance, to measure a distance, you still need to place two points to define a line segment. But these two points can be placed at entirely different views in the 3D volume.

At all times, the views will only show the projection of the measurement (distance, angle, area) onto the plane of the views. The displayed measurement value, however, can be either the true three-dimensional measurement (3D mode) or the measurement of the projection (2D mode).

Note: Rather than move through the image set to place measurement points, you may find it easier to set up an oblique view that contains all the points of the feature you want to measure and perform the measurement on this view. While, in theory, you can place measurement points on 3D views, you should NOT use this technique, because on such views you do not have ANY indication of how deep the point is located inside the 3D volume, without continuously correlating the position of the 3D cursor on the baseline views.