Comparison of two co-registration methods for real-time ultrasonography fused with MRI: a phantom study

To test the accuracy of spatial registration of real-time ultrasonography (US) fused with MRI in a phantom.

An US prototype system (LOGIQ, GE Healthcare) with incorporated software for fusion imaging was used to test two methods of co-registration in a phantom: co-registration from specific points, where common reference points identifiable on both MRI and US images were marked, and plane registration, where common planes identifiable on both MRI and US images were marked. In two series we performed co-registration from points and in one series we performed co-registration from planes. The accuracy of the co-registration was measured at 3 measuring points, defined before initiation of the study, and it was calculated as the root mean square deviation (RMSD), which corresponds to the standard deviation. It was measured in millimeters. Two observers each performed 30 co-registrations for each series, totaling 180 co-registrations. The difference between the methods and the observers was calculated using analysis of variance (two-way ANOVA).

Co-registration was significantly more accurate when using the measuring points as co-registration points than when using points covering a different area of the phantom (p < 0.0001). The mean calculated RMSD when using the measuring points as co-registration points was 1.3 mm (95 % CI: 1.1 - 1.5 mm), when using points away from the measuring points: 4.0 mm (95 % CI: 3.2 - 4.8 mm), and when using planes for the co-registration: 3.8 mm (95 % CI: 3.2 - 4.4 mm).

Image fusion involving real-time US has high accuracy and is easy to use in a phantom. Working within the area given by the co-registration points optimizes the accuracy. Image fusion is a promising tool for clinical US, since it provides the potential of benefiting from different imaging modalities in one examination.