Spatial Calibration

Stradx offers two methods for spatial calibration. For the quickest and most accurate calibration, you'll need to manufacture or purchase a mechanical alignment instrument, as described in this technical report. The Stradx features which support this technique are described here.

Most users, however, will use Stradx's alternative spatial calibration technique, for which you need only a flat-bottomed waterbath (or a Cambridge phantom for more accuracy). This technique is still very accurate and takes only a few minutes to perform. Details of how it works can be found in this technical report. To perform spatial calibration this way, you need to do the following things:

• Fill a flat-bottomed water bath with water.
• Mount your position sensor rigidly on the ultrasound probe. An optical or Bird sensor may be mounted directly on the body of the probe, but a Fastrak or Patriot should be fixed about 10cm away using some sort of rigid plastic strut.
• If you have a Cambridge phantom available, mount the probe in the phantom, ensuring that the brass strip in the phantom is at all times in the centre of the ultrasound beam. If you don't have a Cambridge phantom, ensure that the bottom of the water bath is flat and roughen it uniformly using small circular motions with very fine emery cloth or sand paper. This will improve the reflection of ultrasound at oblique angles.
• Use the `Edit setup' controls or the `Preview' window to select the area of the ultrasound scanner screen that you want to record. When you are happy with the image cropping you should save it using the `Save setup' menu item.
• Configure the controls of the ultrasound machine so that the line feature appears clearly in the preview window.
• Using the `Edit setup' menu, choose a slow acquisition rate. 5 frames a second, motion-gated or stable-gated acquisition are the best options.
• Record a sequence of images as detailed below. You will probably need to record between 400 and 800 frames to complete the sequence. All probe movements should be as slow and steady as possible.
• Open the automatic calibration window and, if necessary, use the `Define scan area' button to outline the region of the image occupied by the scan (the cyan polygon in the example below). The left mouse button is used to specify each point round the edge of the region, and the right button is used to indicate the final point.
• Scroll through the B-scans and check that the line detector (the green line in the example below) latches on to the real image of the brass beam (or the bottom of the water bath if you are not using a Cambridge phantom). If the line image is weak, it is possible that the detector will not latch on to the line, but in this case check that the line is rejected and not accepted for subsequent processing. The automatic accept/reject decision is displayed near the bottom of the window. As you scan through the frames, the positions of accepted green lines are recorded, and a count of how many frames have been processed is maintained at the bottom of the window.
• If the window incorrectly accepts a green line which has in fact been badly fitted, press the space bar to force it to reject the line. Conversely, if the window incorrectly rejects a green line which has in fact been properly fitted, press the space bar to force it to accept the line. Pressing the space bar a second time forces the window to re-detect the line automatically, overriding the manual accept/reject decision. If you find yourself making many manual corrections, try adjusting the sliders in the calibration controls window to improve the line detection. When you've finished adjusting these sliders, go back to the first B-scan and press `Reset' to clear the window's memory of previously detected lines. Then start scrolling through the B-scans again.
• Press the `Solve' button to run the calibration solution algorithm. This algorithm uses the recorded positions of those green lines which were accepted as you scrolled through the B-scans. If there are any B-scans which you did not visit, the solution algorithm runs the automatic line detector on these frames and uses any lines which are subsequently accepted. If you don't trust the automatic detector's accept/reject decisions, make sure you scroll through all the B-scans by hand before pressing the `Solve' button.
• Use the `Verify' toggle button to check the calibration. When in `Verify' mode, the calibration window shows the intersection of the calibrated plane with the current B-scan (instead of the results of the line detection algorithm). If the calibration has gone well, you should observe the plane passing through the image of the brass beam (or the bottom of the water bath if you are not using a Cambridge phantom) in all B-scans.
• Save your results in a calibration file.

Sequence of Probe Movements for Automatic Calibration

There are some movements of the probe which it is vital to perform in order to sufficiently constrain the calibration process. A suggested procedure for calibration is shown in the diagram below.

Motions (A), (B) and (C) must be performed in order to constrain the six positional parameters of the B-scan relative to the position sensor transmitter (three angles and three translations) and the two scale parameters (in the plane of the B-scan).

However, the location of the plane being imaged by the probe is also unknown and must be estimated by the spatial calibration process. Repeating motions (A), (B) and (C) in three different locations will provide sufficient constraints for the location of this plane to be determined.

An additional rotation about the vertical axis between each of these three measurement locations (as shown in the diagram above) is not strictly necessary, but will ensure that the calibration process will not be degraded by a poor choice of orientation of the position sensor transmitter with respect to the receiver.

You may need to adapt this sequence to satisfy constraints imposed by your position sensor or water bath, but you must ensure that you incorporate all the motions in several different positions on the calibration plane.

The full set of unknowns that the system solves for are as follows:

1. Z translation parameter of transformation from datum of position sensor to phantom plane. This value is discarded.
2. Elevation parameter of transformation from datum of position sensor to phantom plane. This value is discarded.
3. Roll parameter of transformation from datum of position sensor to phantom plane. This value is discarded.
4. Azimuth parameter of transformation from the ultrasound scan plane to the sensing element of the position sensor.
5. Elevation parameter of transformation from the ultrasound scan plane to the sensing element of the position sensor.
6. Roll parameter of transformation from the ultrasound scan plane to the sensing element of the position sensor.
7. X translation parameter of transformation from the ultrasound scan plane to the sensing element of the position sensor.
8. Y translation parameter of transformation from the ultrasound scan plane to the sensing element of the position sensor.
9. Z translation parameter of transformation from the ultrasound scan plane to the sensing element of the position sensor.
10. Scaling parameter of the ultrasound scan plane in the X direction.
11. Scaling parameter of the ultrasound scan plane in the Y direction.