Page 134 - GTM-4-1

Global Translational Medicine

ORIGINAL RESEARCH ARTICLE
Predicting magnetic resonance imaging-guided

focused ultrasound sonication parameters
beyond skull density ratio

Alsu Narkisovna Khatmullina * , Diana Shamilevna Avzaletdinova 1,2 ,
1
Dinara Ilgizovna Nabiullina 1 , Sergey Nikolaevich Illarioshkin 3 ,
Guzaliya Minvazykhovna Sakharova 1,2 , Naufal Shamilevich Zagidullin 1,2 ,
Nadezhdina Ekaterina Andreevna 1,2 , Shamil Makhmutovich Safin 2 , and
Rezida Maratovna Galimova 1,2

1 Intelligent Neurosurgery Clinic, Ltd., V.S. Buzaev International Medical Center, Ufa, Bashkortostan,
Russia
2 Department of Surgery, Bashkir State Medical University, Ufa, Bashkortostan, Russia
3 Institute of the Brain, Research Center of Neurology, Moscow, Russia
(This article belongs to the Special Issue: Special Issue of Global Translational Medicine in the
Fourth RCCCDT-2024)

Abstract

Precise temperature regulation is essential for effective and safe magnetic
resonance imaging-guided focused ultrasound (MRgFUS) treatments. Several
variables influence the target temperature during sonication, with the energy
delivered being a pivotal physical determinant. The skull density ratio (SDR) is
*Corresponding author: utilized to evaluate the feasibility of treatment, with values below 0.3 – 0.4 generally
Alsu Narkisovna Khatmullina considered contraindications for treatment. This study aimed to develop a robust
(info@buzaevclinic.ru)
predictive model for sonication parameters that can accurately achieve the desired
Citation: Khatmullina AN, temperature within the target tissue region. We obtained 152 treatment log data from
Avzaletdinova DS, Nabiullina DI,
et al. Predicting magnetic the Insightec Exablate workstation. Variables, including power output, sonication
resonance imaging-guided focused duration, stop sonication button activation, SDR, age, sex, and initial sonication
ultrasound sonication parameters (ALIGN) parameters, were used as predictors (x), with the achieved temperature as
beyond skull density ratio. Global
Transl Med. 2025:4(1):126-135. the response (y), to construct the predictive models. RStudio was used to build linear
doi: 10.36922/gtm.5419 models, and the TensorFlow library was employed for the neural network models.
Received: October 22, 2024 The linear and neural network models predicted tissue temperature with a mean
Revised: December 28, 2024 absolute error of 2.78°C and 1.93°C, respectively, and a coefficient of determination
Accepted: February 7, 2025 of 0.71 and 0.76, respectively. The neural network model outperformed the linear
Published online: March 10, 2025 model, demonstrating a smaller residual dispersion and a lower root-mean-square
Copyright: © 2025 Author(s). deviation. While the neural network model is more accurate and reliable for
This is an Open-Access article predicting MRgFUS temperatures, the linear model is easier to use. Key factors, such
distributed under the terms of the
Creative Commons Attribution as sex, age, SDR, and the initial tissue response to the first sonication – the energy
License, permitting distribution, delivered during this initial treatment and the corresponding temperature – are
and reproduction in any medium, crucial for optimizing subsequent sonication parameters such as power, energy, and
provided the original work is
properly cited. duration.
Publisher’s Note: AccScience
Publishing remains neutral with Keywords: Magnetic resonance imaging-guided focused ultrasound; Skull density ratio;
regard to jurisdictional claims in
published maps and institutional Sonication
affiliations.

Volume 4 Issue 1 (2025) 126 doi: 10.36922/gtm.5419