The Focused Ultrasound Foundation has dedicated significant resources over the past few months to exploring the convergence of focused ultrasound and the ever-evolving world of artificial intelligence (AI) and machine learning (ML).
With the aim to identify patterns and fundamentally understand how AI/ML is being applied to improve focused ultrasound treatments, our team reviewed more than 130 papers that may contain insights for improving clinical solutions.
Three Fields of Interest
Our extensive analysis revealed a surprisingly diverse range of applications for AI/ML across a number of medical indications and treatment particulars. But despite the breadth of these findings, three areas emerged as prominent fields of interest: neurological, gynecological, and oncological applications.
The neurological applications were the broadest group, representing about one-quarter of the use cases identified. Here, ML tools were primarily used to help clinicians monitor treatments and interpret treatment results. Techniques like reducing the noise (or imperfections) in treatment images and assessing biomarkers during focused ultrasound blood-brain barrier opening procedures were prevalent. These tasks rely on convolutional neural networks (CNNs), a class of neural networks frequently used in other medical image processing applications.
The second area of interest for ML applications was the gynecological field, which was the focus of approximately 20% of our findings. In this space, ML is used heavily in treatment monitoring, like we saw in the neurological space, but we also saw it used upstream in the treatment planning process. It can assist in predicting treatment outcomes for focused ultrasound ablation of uterine fibroids, and clinicians use image segmentation techniques for precise spatial targeting before and during treatments.
In oncology, the data cover a wide variety of malignancies. Together, they represented another 20% of the ML applications. In this area, ML could play an important role in helping predict disease progression after focused ultrasound treatment and assessing biomarkers for outcome evaluations, along with other uses.
In all, the three areas of medicine we have discussed represent approximately two-thirds of the AI/ML applications in focused ultrasound in recent publications. However, we have also seen emerging uses of ML in cardiovascular care, cosmetology, and many other areas, underscoring the evolution of the field. AI/ML will have a significant clinical impact in focused ultrasound for years to come.
Creating a Community
The insights above were distilled from the past few years’ worth of published papers, and the Foundation is committed to staying abreast of these developments over time. To help exploration and research, we’ve compiled articles, conference abstracts, and graduate theses referencing the use of AI in combination with focused ultrasound, which is now available as a Google Scholars report. We will monitor shifts in these patterns and report back on emerging trends as they surface. One of our primary goals is to ensure that our community remains well-informed and ahead of the AL/ML curve. To that end, we have also established a strategic partnership with the University of Virginia School of Data Science to explore shared interests and collaborate further. I hope to share more about that in the months ahead.
I encourage you to join our Community of Practice and participate in the User Forum to engage, share, and help shape the future of focused ultrasound and AI/ML. Your insights, your questions, and your enthusiasm will drive this field forward and collectively push the boundaries of what is possible.
You can also reach out to me and our staff at techteam@fusfoundation.org. We are happy to answer any questions and to hear your ideas. As always, I appreciate your dedication, passion, and commitment to innovation, keeping in front of mind the goals of improving patient outcomes and ensuring real-world impact.
Rick Hamilton is the Foundation’s managing director and chief technology officer. He is recognized as one of the most prolific inventors in world history, with over 1,060 issued US patents.