Automatic Doppler Volume Fusion of 3D Ultrasound using Point-based Registration of Shared Bifurcation Points

Shinya Onogi, Tuan Hung Phan, Takashi Mochizuki, Kohji Masuda
Vol. 4 (2015) p. 27-34

We previously proposed the use of acoustic microbubble delivery in blood vessels as a therapeutic application of microbubbles to improve the efficiency of high-intensity focused ultrasound and the efficacy of acoustic targeted drug therapy. Among the technical requirements for this technique is detailed visualization of the blood vessel network for navigation around a target such as a tumor. For this purpose, three-dimensional (3D) Doppler volumes, which can be acquired by matrix array imaging probes, are quite convenient because they allow the blood vessel structure to be extracted without segmentation. However, the acquirable volume is limited and incomplete because the Doppler signal depends on flow direction. To compensate for these issues, an ultrasound volume fusion technique is required. In this study, we propose a blood vessel volume fusion method by automatic registration among shared bifurcations. In addition, we propose a novel 3D ultrasound calibration method, which is needed to determine the initial transformation. Several optical markers are used as fiducial markers in this calibration. To examine the feasibility of the proposed methods, calibration accuracy and volume fusion accuracy assessments were conducted using an artificial blood vessel and in human subjects. Regarding calibration accuracy, the target registration error of the proposed method was 2.2 mm. Regarding volume fusion accuracy in the artificial blood vessel, the mean distance between the shared bifurcations was reduced from 2.4 mm (initial transformation by tracking data) to 0.5 mm (registration of shared bifurcations). Regarding the volume fusion of blood vessels in human subjects, the distance was also reduced from 10.4 mm to 0.3 mm. The results demonstrate that the proposed methods are accurate for constructing large and complete blood vessel networks for navigation of microbubble delivery. Moreover, the methods may also be useful for extracting intraoperative blood vessel network to support minimally invasive surgery or therapy.