Page 135 - GTM-4-1

Global Translational Medicine MRgFUS sonication parameters prediction

1. Introduction and oxygen bubbles along the ultrasound path can absorb
ultrasound energy, affecting the efficiency of heating.
Magnetic resonance imaging-guided focused ultrasound Conductivity is also impacted by tissue conditions along
(MRgFUS) treatment has been approved by the Food and the ultrasound path.
Drug Administration and the European Medical Agency
2
1
as a non-surgical alternative for treating movement Incorporating additional clinically relevant factors
disorders. This technology concentrates ultrasound at would enhance the construction of a temperature
specific tissue points in tissue, offering therapeutic potential model, enabling more accurate calculations of the target
in various clinical conditions. 3-13 Clinical studies have temperature based on power, time, and the bone tissue
demonstrated its efficacy in treating essential tremor, 14-16 coefficient.
Parkinson’s disease (PD), 17-23 neuropathic pain, 24-27 and This study aimed to determine the most accurate
psychiatric conditions. 28-32 At present, researchers are temperature calculation model using various models,
exploring the use of MRgFUS in incurable diseases, such as including linear models and a neural network, considering
Alzheimer’s disease. and other dementias, amyotrophic tissue conductivity and other variables.
33
34
lateral sclerosis, glioblastomas and brain metastases, and
35
Huntington’s disease. 2. Methods
Initially, focused ultrasound treatment systems used We obtained log data of 152 MRgFUS treatments between
ultrasound imaging for treatment, which has limited May 05, 2020, and July 01, 2023, using the Insightec
guidance accuracy and lacked real-time temperature Exablate workstation. According to nosological forms,
detection. Combining magnetic resonance (MR) imaging patients were distributed as follows: PD with predominant
with focused ultrasound has significantly improved tremor phenotype (84 patients), PD with predominant
imaging and enhanced thermometry accuracy to assess the akinetic-rigid phenotype (four patients), essential tremor
degree of exposure. MR-thermography is used to control (45 patients), and various forms of dystonia (19 patients).
the target area temperature. 36
The first model was based on power, energy, time,
MRgFUS uses three interrelated parameters: energy and bone ultrasound conductivity coefficient. A random
(joules), power (watts), and time (seconds). 37,38 During initial sample of 782 sonications was used for model training,
sonication, known as ALIGN, the neurosurgeon sets the while 369 sonications were used for testing.
power to 150 watts for 10 s, resulting in 1,500 joules of energy.
We hypothesized that the response to the initial sonication On the test set, the root mean square error (RMSE) was
energy can predict the outcomes of subsequent sonications. 3.68, the coefficient of determination (R²) was 0.62, and the
mean absolute error (MAE) was 2.78. Training set RMSE
When adjusting power during MRgFUS treatments, was 3.55 with an R² of 0.64 and an MAE of 2.67. Figure 1
caution is essential to prevent rapid overheating, which shows the observed and predicted temperatures plotted
can have irreversible consequences. Excessive power against each other. In the second model, all the parameters
can quickly overheat an area, potentially causing patient were included to identify those that significantly affect the
discomfort. In addition, higher power levels may increase temperature. Power (P < 0.05), sonication duration, early
headaches, causing the patient to stop or decline the cessation, bone conductivity coefficient, initial sonication
procedure. Insufficient power may prolong sonication parameters, and results, age, and sex are the key factors
unnecessarily, causing temporary tissue swelling and that significantly impact temperature in the model.
altering the neurological condition.
The third model was constructed by eliminating
Time plays a critical role in the heating process. Longer redundant parameters using the command:
durations allow for more energy delivery and intense
heating, while shorter duration leads to insufficient energy lm(formula = max_avg_temp ~ power + stopped + scull_
to heat tissue. Excessive heating time, on the other hand, score + first_measured_energy + first_max_avg_temp +
can lead to patient discomfort and headaches. Reducing actual_duration + age + male_sex, data = ds [idx == 1, ])
the treatment time may improve patient tolerance. It is
important to note that different patients require varying After training, the median of residuals was 0.05 with
durations to heat an area to a specific temperature. P < 2.2 × 10 , which is close to zero as required. However,
-16
Bone tissue properties, such as the skull density ratio the model tended to underestimate the temperature. The
(SDR), are known to affect energy loss. Patients with parameters of the linear model are shown in Table 1.
varying conductivity may require different durations to Recent advances in machine learning enable the
reach the same target temperature. In addition, the water development of sophisticated algorithms to construct

Volume 4 Issue 1 (2025) 127 doi: 10.36922/gtm.5419