A team of Korean experts proposed a neural network-based real-time acoustic simulation framework for ultrasound treatment of brain diseases.


Transcranial Focused Ultrasound Technology

Ultrasound therapy has emerged as an non-invasive treatment option for transcranial procedures. This involves focusing ultrasound energy on a few millimeters of the brain, including deep regions, to treat neurological disorders without the need to open the skull. It results in minimal impact on the surrounding normal brain tissue with reduced side effects like infections and complications. Because of this, focused ultrasound technology has been applied to the treatment of intractable diseases like Alzheimer's disease and depression.

In predicting the location of the invisible acoustic focus, medical images taken prior to treatment are utilized for navigation systems. This provides information about the relative position of the patient and the ultrasound transducer. However, the use of this procedure has been limited due to the difficulty in reflecting the distortion of ultrasound waves caused by the differences in skull shapes of patients.

Different simulation techniques have been developed to compensate for this, but they still need significant computational time, which makes them difficult to apply in actual clinical setting. Until now, the clinical use of AI simulation models in non-invasive focused ultrasound therapy has not been validated.


READ ALSO: Light and Sound 3D Imaging That Fits on a Needle Could Revolutionize Endoscopy

Real-Time Acoustic Simulation Framework

At the Korea Institute of Science and Technology (KIST), a team of researchers led by Dr. Kim Hyungmin of the KIST Bionics Research Center has developed a technology which uses generative AI to predict and correct the distortion of the ultrasound focus position. The result of their research was discussed in the study "Real-Time Acoustic Simulation Framework for tFUS: A Feasibility Study Using Navigation System".

In this study, the research team designed a real-time focused ultrasound simulation technology using an AI model based on a generative adversarial neural network (GAN). This deep learning model is widely used for image generation in the medical setting.

The new framework reduces the update time of 3D simulation information which reflects changes in ultrasound acoustic waves from 14 seconds to 0.1 second. At the same time, it shows a focal position error of less than 6 millimeters and an average maximum acoustic pressure error of less than 7%. These errors are both within the error range of existing simulation technologies.

A medical image-based navigation system was also developed by the team to verify the performance of the framework. The new system can offer real-time acoustic simulations at the rate of 5 Hz depending on the position of the transducer. It also succeeded in predicting the position of the ultrasound energy and focus within the skull during the ultrasound therapy.

Because of the long calculation time, the ultrasound transducer used to be precisely positioned in a pre-planned location for the utilization of results. With the new simulation-guided navigation system, it is currently possible to adjust the focus based on the acoustic simulation results collected in real time. In the future, it can improve the accuracy of focused ultrasound and offer safe treatment for patients by quickly responding to unexpected situations which can occur during the treatment process.

RELATED ARTICLE: 'Focused Ultrasound' Neurosurgery Treats Essential Tremor Using Light Beams Without Brain Incisions


Check out more news and information on Ultrasound in Science Times.