Researchers have developed deep learning-based methods to create holograms using focused ultrasound in real time.
Holography is a photographic technique that records light scattered from an object, and then presents it in a three-dimensional way. We often see the use of this technology in science fiction movies like Star Wars and Iron Man. Various types of holograms have been developed over the years, including transmission holograms, which allow light to shine through them and the image to be seen from the side; and rainbow holograms, which are used for security purposes.
But now, a group of researchers led by Professor Hwang Jae-yoon of the DGIST Department of Electrical Engineering and Computer Science has developed a deep learning-based ultrasound hologram generation framework technology that is freely configurable. the form of focused ultrasound in real time based on holograms. . It is expected to be used as a basic technology in the field of brain stimulation and treatment that requires precision in the future.
The problem with the use of ultrasound is that it is difficult to selectively stimulate the relevant parts of the brain where many areas interact with each other at the same time because the current technology focuses the ultrasound on a small point or a large circle for stimulation. To solve this problem, a technology that can freely focus ultrasound on a desired area using the hologram principle has been proposed, but there are limitations, such as low accuracy and long calculation time to generate hologram.
The team proposed a deep learning-based framework that can incorporate free and accurate ultrasound targeting in real time by learning to create ultrasound holograms to overcome the limitations. They show that it is possible to focus the ultrasound on the desired shape more precisely in a hologram creation time that is close to real time, and up to 400 times faster than the existing hologram creation algorithm method. on ultrasound.
The deep learning-based learning framework learns to generate ultrasonic holograms through self-supervised learning. This is a method of learning to find the answer by finding a rule on its own for data that does not have an answer. The research team proposed a learning method for generating ultrasonic holograms, a deep learning network optimized for ultrasonic hologram generation, and a new loss function, while verifying the validity and efficiency of each component through simulations and actual experiments.
Reference: Moon Hwan Lee et al, Deep Learning-Based Framework for Fast and Accurate Acoustic Hologram Generation, IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control (2022). DOI: 10.1109/TUFFC.2022.3219401