Sonication experiments using ex vivo skulls were performed at our institute, in collaboration of the Department of Neurosurgery and the Department of Anatomy. Five human skulls were prepared and modified to measure acoustic fields that could penetrate human calvarias : all brain tissue was removed, including dura mater. The skulls did not undergo embalming processes to reduce the bias caused by the chemical interaction. All fresh skulls were recruited by voluntary donation for the purpose of medical practice, after obtaining consents of the donors and their families. The ultrasound source used was the Exablate 4000 transducer (InSightec, Haifa, Israel) with a radius of 150 mm. The skulls underwent removal of the remaining air inside, and was then placed within a transducer filled with degassed water. A needle hydrophone of the HNC-0400 model (ONDA, Sunnyvale, CA, USA) was inserted into the skull to measure the acoustic pressure in a specific location. The needle hydrophone could be moved and controlled by an attached robotic arm upon the transducers. The acoustic fields were explored in two ways, including two-dimensional (2D) scan from all elements and an acquisition of acoustic correction table (ACT) from each of the elements. The former was to measure acoustic pressure at every point during movement of the hydrophone across the 2D grid when all element of the transducers were acting together. In contrast, when every element was acting separately, the ACT was used to record amplitude and phase of every element, using a hydrophone fixed in one location. The skull was fixed to a mounting frame by weaving in four holes, followed by attachment to the placement frame to experiment at various target points inside the skull. The placement frame was designated with uniformly spaced holes, 1-cm distance between the holes (Fig. 1). As the target position in the skull shifted, various measurements were conducted including energy transmission, phase shifts, and acoustic field shape, in each case. The acoustic measurements were performed using both 230 kHz and 680 kHz transducers. Mid frequency device was used in four skulls and low frequency in five skulls. One skull was not able to undergo experiments in low frequency transducer due to time limitations of the non-embalmed skull condition.

Retrospective review was conducted using sonication data of 46 patients who underwent MRgFUS ablation from February 2013 to August 2017. They were classified with four disease entities : ET, OCD, PD, and MDD. Eighteen of 46 patients underwent unilateral MRgFUS thalamotomy for ET, targeting the left ventrointermediate nucleus of the thalamus (Vim). Nine PD patients experienced unilateral MRgFUS pallidotomy to control their severe dyskinesia, targeting the left posteroventral globus pallidus interna (Gpi). In addition, psychosurgical procedures were performed in 19 patients (16 OCD and three MDD) at the bilateral anterior limb of the internal capsule (ALIC) using noninvasive sonication. Detailed target coordinates are described in Table 1.

The entire MRgFUS procedure was performed in a 3-T MRI scanner (GE Medical Systems, Milwaukee, WI, USA) using the Exablate 4000 device (InSightec). After stereotactic frame fixation to the skull and assembly of the MRgFUS system, diagnostic MRI scans were acquired and the coordinates of target points were set to focus ultrasonic energy. Prior to producing permanent thermal lesions, the size and location of a real thermal spot was evaluated using subthreshold lowpower sonication. Then, acoustic power and energy were steadily increased to reach a sufficient temperature to generate a permanent lesion under the guidance of MRI and MR thermometry. The neurological status of the patients was repeatedly assessed during the cooling time between each sonication.

Other calculations regarding the skulls were performed in an automatic program, mainly using computed tomography (CT) scanning. Skull density ratio (SDR) was calculated by CT Density Analysis Tool (InSightec) and skull volume or thickness was performed by using a three-dimensional image software (Aquarius version 4.4; TeraRecon, Foster City, CA, USA).

Based on the experimental conditions identified in the cadaveric experiments, the clinical patients were classified into several groups, according to sonication parameters and clinical characteristics of each patient’s skull. We performed comparative analysis of the sonication procedures between ex vivo skulls and in vivo skulls, focusing on effective conditions for ultrasonic penetration through the human skull. Written informed consent was obtained from all included patients prior to the procedures. This study received full approval from the Institutional Review Board of Yonsei University College of Medicine (1-2018-0087).

All statistical analyses were performed using SPSS version 23 (IBM Corporation, New York, NY, USA). Linear regression methods were used to determine the correlation between skull- or sonication-related factors and maximal temperature (Tmax) or energy applied for Tmax rise. The threshold for statistical significance was set at p<0.05. Any comparisons among the three different target groups were selectively analyzed by analysis of variance or Kruskal-Wallis test, according to the normality of the data. For the post analysis, the statistical significance level was corrected using Bonferroni’s method (p<0.017).

Skull experiments were performed in five adult cadaver skulls that were not processed by chemical agents for preservation. The mean skull thickness was 6.6±0.9 mm and the volume was 302.1±41.3 cm3. The average SDR was 0.47±0.13.

The clinical studies constituted 46 patients, including 18 patients with ET, nine with PD, 16 with OCD and three MDD. Their mean age was 51.4 years (range, 21–75). There were more men than women (29 : 17). The characteristics of in vivo skulls were as follow : mean thickness of 5.6±1.2 mm, skull volume of 317.1±52.3 cm³, and SDR of 0.57±0.10. The detailed demographics related to skull and MRgFUS procedure of the clinical patients are described in Table 2.

SDR indicates the ratio of Hounsfield unit value on CT between the marrow and cortical bone, which is originally calculated as <1 (Fig. 2). A lower SDR indicates a relatively higher cortical bone density than bone marrow, which signifies a larger difference of acoustic impedance between two types of tissues. This difference can affect the degree to which ultrasonic waves are transmitted.

In the experimental setting, five skulls showed a positive correlation between SDR and relative intensity (RI). The RI of ultrasound is represented as the square value of integral calculus, consisting of the relative amplitude of ultrasound, both in plain water and after skull penetration.

This can be interpreted as the definition of transmission efficiency. Although the RI of mid frequency sonication only became one third of the results of low frequency experiments, similar tendencies were maintained in association with SDR (Fig. 3).

In the clinical results, SDR was significantly positively correlated with peak temperature (p<0.001, r²=0.313, y=43.156+24.326x) and the magnitude of temperature increase above 54℃ (p<0.001, r²=0.317, y=-6.046+21.584x). Although none of the skull factors were correlated with SDR, independent skull characteristics showed an inverse correlation with the maximal energy delivered to reach peak temperature (p=0.002, r²=0.198, y=41512.5–37132.5x). In the clinical setting, the magnitude of temperature increase per unit energy was analyzed as a function of transmission efficiency, since it is difficult to place the hydrophone into the human brain and measure the actual RI of the ultrasonic energy. The high SDR was correlated with increases in the magnitude of temperature increase per unit energy (p<0.001, r²=0.285, y=-0.002+0.008x). Thus, high SDR could make it easy to achieve sufficient temperature and subsequent lesions, using relatively low energy. Results from both ex vivo and in vivo skulls both indicated that high SDR is more efficient at energy transmission and at raising temperatures to make permanent lesions in the brain.

Other potential factors of the skull that impeded ultrasonic energy transfer included skull volume, thickness, and skull area. Among these factors, Tmax was negatively correlated with skull thickness (p=0.006, r²=0.162, y=65.792–1.520x) and skull volume (p=0.043, r²=0.090, y=65.501–0.026x). Skull area had no significant relationship with Tmax (p=0.161). In terms of maximal energy requirement to achieve Tmax, only skull volume was closely related with energy delivery required to reach Tmax (p=0.045, r²=0.088, y=3001.7+43.8x) (Fig. 4).

Theoretically, energy reflection and transmission of ultrasound are correlated with the respective acoustic impedances of the different materials and the incidence angle of ultrasound wave. In terms of the incidence angle, the mean values of the angles are higher for targets located laterally than for those in the center position. In the skull experiments, lateral targets showed a much more unfocused ultrasound than those in the center position. However, there was no difference in the ability of energy to focus at a single point, when targets were optimally corrected (Fig. 5). In addition, we set 25 degrees as a standard value because the ultrasound beam in the Exablate system showed steeply increased reflection and decreased transmission with incidence angles >25 degrees. Considering that incidence beam angles within 25 degrees are efficient for transmission through the skull, as the target location moves further from the center, the fewer number of cumulative transducer elements can be used for successful sonication (Fig. 6A). When comparing clinical data across three different targets, it was confirmed that Vim exhibited the most ultrasound elements that could be effective for therapeutic sonication, and was sequentially followed by ALIC and Gpi (Fig. 6B).

As the incidence angle became larger, the energy transmission through the skull was less effective in both settings at low and mid frequencies, which appeared more clearly in skull conditions with low SDRs (Fig. 7A and B). Particularly when the outer incidence angle was >25 degrees, amplitude transmission sharply decreased in circumstances where SDR was less than 0.6. These experimental results of energy transmission were consistent with clinical results. The average of incidence angle was 12.9±5.5 degrees for Vim, 14.9±7.0 degrees for ALIC, and 16.5±5.9 degrees for Gpi. Notably, the ultrasonic energy required for Tmax increase was significantly different among the groups : 15977.5±6712.2 J for thalamotomy, 23907.7±9690.5 for capsulotomy, and 20001.1±7024.2 for pallidotomy (p=0.022). In particular, post hoc analysis (Bonferroni’s method) comparing two groups showed thalamotomy consumed much less energy to reach peak temperature and lesioning, compared with capsulotomy (p=0.015). It demonstrated a similar result when compared with pallidotomy, but this was not statistically significant (p=0.698) possibly due to the small number of cases of pallidotomy. This has reconfirmed that the efficiency of the clinical focused ultrasound can vary depending on the target location.

Interestingly, cases with SDR ≥0.6 showed slightly improved amplitude transmission at incidence angles >25 degrees (Fig. 7C), indicating that high SDRs are less affected by the influence of the incidence angle. In addition, low frequency ultrasound showed better amplitude transmission than mid frequency ultrasound in all SDR conditions and/or larger incidence angles. These results will play an important basic role for selecting clinical target and calibrating the technology.