BACKGROUND

Histotripsy is a non-invasive, non-ionizing, non-thermal focused ultrasound technique. High amplitude short acoustic pulses converge to create high negative pressures that cavitate endogenous gas into a bubble cloud leading to mechanical tissue destruction. In the United States, histotripsy is approved to treat liver tumors under diagnostic ultrasound guidance but in initial clinical cases, some areas of the liver have not been treated due to bone or gas obstructing the acoustic window for targeting. To address this limitation in visualization, cone-beam computed tomography (CBCT) guided histotripsy was developed to expand the number of tumors and patients that can be treated with histotripsy.

PURPOSE

The purpose of this work is to improve the accuracy of CBCT guided histotripsy by calibrating the therapeutic bubble cloud location relative to the histotripsy robot arm.

METHODS

The calibration correction involves creating a bubble cloud sized treatment (a few mm) in an agar-based phantom consisting of 11 layers with alternating high and low x-ray attenuation. The layers were spaced ∼3 mm apart to allow visualization of mixing after mechanical disintegration from the histotripsy treatment. Bubble cloud treatments were localized using an automated algorithm that minimized a cost function based on the intensity difference within the treatment region on the pre- and post-treatment CBCT. The actual treatment location can be compared to the theoretical bubble cloud location (focal point based on the CAD model of the transducer assembly) to calculate a 3D offset (X, Y, Z), which is used as the calibration correction between the therapeutic bubble cloud location and the histotripsy robot arm. The phantom and algorithm were analyzed to determine parameters that maximized bubble cloud treatment detection (treatment duration, localization accuracy of the phantom, number of bubble clouds) and were tested on four different histotripsy transducers.

RESULTS

Bubble cloud locations were accurately identified with the automated algorithm from post-treatment CBCT images of the multilayer agar phantom. Treating the phantom for 20 seconds was associated with the greatest change in CBCT intensity. The phantom and algorithm were able to localize changes in bubble cloud location with mean residual errors (MRE) between the measured and planned translations of 0.3 ± 0.3 mm in X, -0.2 ± 0.6 mm in Y, and 0.1 ± 1.0 mm in Z. A multi-bubble cloud calibration approach with four adjacent bubble clouds provided a statistically significant lower mean absolute deviation (MAD) in measured 3D offset (0.1, 0.0 and 0.2 mm in X, Y, and Z, respectively) compared to using a single bubble cloud (MAD of 0.2, 1.1 and 1.2 mm in X, Y, and Z, respectively). The calibration correction method measured statistically significantly different 3D transducer offsets between the four histotripsy transducers.

CONCLUSIONS

Creating and analyzing four adjacent bubble clouds together produced more accurate and reproducible 3D offset measurements than analyzing individual bubble clouds. The presented histotripsy bubble cloud calibration correction method is automated, accurate, and can be easily integrated in the current histotripsy workflow to improve accuracy of CBCT guided histotripsy.