Blood-brain barrier disruption in humans using an implantable ultrasound device: quantification with MR images and correlation with local acoustic pressure

OBJECTIVE One of the goals in this study was to set up a semiautomatic method to estimate blood-brain barrier disruption obtained in patients with glioblastoma by using an implantable, unfocused, ultrasound device. Another goal was to correlate the probability of significant ultrasound-induced signal enhancement (SUISE) with local acoustic pressure in the brain. METHODS Gd-enhanced MR images acquired before and after ultrasound treatments were analyzed prospectively. The image sets were segmented, normalized, and coregistered to evaluate contrast enhancement. The volume of SUISE was calculated with voxels labeled as gray or white matter, in a cylindrical region of interest, and with enhancement above a given threshold. To validate the method, the resulting volumes of SUISE were compared to qualitative grades previously assigned by 3 clinicians for 40 ultrasound treatments in 15 patients. A parametric study was performed to optimize the algorithm prediction of the qualitative grades. The 3D acoustic field in the brain was estimated from measurements in water combined with simulations accounting for ultrasound attenuation in brain and overlaid on each MR image to correlate local acoustic pressure with the probability of SUISE (defined as enhancement > 10%). RESULTS The algorithm predicted grade 2 or 3 and grade 3 openings with areas under the receiver operating characteristic curve of 0.831 and 0.995, respectively. The probability of SUISE was correlated with local acoustic pressure (R-2 = 0.98) and was 3.33 times higher for gray matter than for white matter. CONCLUSIONS An algorithm for evaluating blood-brain barrier disruption was validated and can be used for future clinical trials to further understand and quantify this technique in humans.