Comparing calibration approaches for 3D ultrasound probes

Objective By adding a tracking sensor to a 3D ultrasound (US) probe and thus locating the probe in space, new applications within the fields of image guided surgery and radiation therapy are possible. To locate the US volume in space, a calibration is necessary to determine the mathematical transformation for mapping points from the tracking coordinate system to the US image coordinate system. We present a comprehensive comparison of two different approaches to perform this calibration for 3D US. Methods For both approaches a phantom is scanned and located in the images by means of segmentation and registration techniques. Calibration is then performed by either relating the tracked phantom's (TP) spatial location to the calibration scans, or by solely correlating scans taken from multiple perspectives when using hand-eye calibration methods (HE). Depending on which approach is utilized, a minimum of one or three images, respectively, need to be acquired for the calibration process. Results We evaluated both approaches for calibration and reconstruction precision. Regarding the latter, the performed tests led to mean target localization errors of 3.5 mm (HE) and 3.3 mm (TP) for real data, and of 1.4 mm (HE) and 0.9 mm (TP) for simulated data. Conclusion Our results indicate that taking additional scans leads to a significant improvement in the calibration. Furthermore, the obtained calibration and reconstruction precisions suggest the use of a TP.