Researchers at the Beckman Institute for Advanced Science and Technology have developed a new framework for super-resolution ultrasound using deep learning. Super-resolution ultrasound provides much clearer images than traditional ultrasound, but processing speeds are slower. Microbubbles are used as a contrast agent in traditional super-resolution ultrasound, but the new approach developed by the researchers forgoes microbubble localization and tracking entirely. Instead, they used deep learning to train a neural network to recognise and reconstruct the high-frequency signals of ultrasound images. The new technique produces results in real-time and could be useful for research and diagnostics. The research was published in IEEE Transactions on Medical Imaging.
Researchers at the Beckman Institute for Advanced Science and Technology have developed a new framework for super-resolution ultrasound using deep learning. This new approach does not require microbubbles, which are traditionally used as a contrast agent to increase the clarity of an ultrasound image.
Super-resolution ultrasound was introduced in the last decade and provides a much clearer picture than the traditional method. However, its processing speeds are much slower, making it less useful for real-time imaging. To address this challenge, the researchers tested a new approach to super-resolution ultrasound technology.
The researchers, led by Dr. Songbin Gong, an assistant professor of electrical and computer engineering at the University of Illinois Urbana-Champaign, teamed up with Dr. Daniel Llano, an associate professor of molecular and integrative physiology and a neurologist at Carle Foundation Hospital, to develop a faster, more efficient method for super-resolution ultrasound.
Their paper, which appears in IEEE Transactions on Medical Imaging, describes a new approach that does not require microbubble localization and tracking. Instead, the researchers used deep learning to develop a new framework for super-resolution ultrasound.
Deep learning is a type of machine learning that uses artificial neural networks to learn from data. The researchers used deep learning to train a neural network to reconstruct high-resolution images from low-resolution ultrasound data.
The new framework was tested on both simulated and experimental data and was found to be faster and more efficient than traditional super-resolution ultrasound techniques. The researchers believe that their approach could help to improve the speed and accuracy of ultrasound imaging, making it more useful for real-time imaging.
“Ultrasound is expected to be a real-time imaging modality,” said Dr. Gong. “Our framework has the potential to significantly improve the speed and accuracy of ultrasound imaging, making it more useful for clinical and research applications.”
Super-resolution ultrasound has many applications in medicine, including the diagnosis and monitoring of cancer, cardiovascular disease, and neurological disorders. The new framework developed by the researchers could help to improve the accuracy and speed of these applications, making them more effective for patient care.
In retrospect, the researchers at the Beckman Institute for Advanced Science and Technology have developed a new framework for super-resolution ultrasound using deep learning. Their approach does not require microbubbles, making it faster and more efficient than traditional super-resolution ultrasound techniques. The researchers believe that their approach could help to improve the speed and accuracy of ultrasound imaging, making it more useful for real-time imaging and clinical applications.
